In this guide, you’ll learn how to do this too.

My workflow is stupid simple.

Anyone can do it.

No coding required.

I open Claude and ask it to create a DM automation for Facebook + Instagram.

Then I tell Claude:

1. The keyword people will comment.

2. The link, product, or offer that should be sent.

…and Claude creates it for me!

This is the SIMPLEST way to turn attention into $.

So, let’s dive into a step-by-step tutorial:

Here’s the step-by-step video tutorial:

Before You Start

Choose these 2 things:

• Your keyword: ____________________

• Your link: ____________________

That is all you need to get started.

Part 1: Make Your First DM Automation

This is the easy path.

We’ll make your first DM automation directly in a website.

1. Create the Automation

1. Open the Blotato Automations page.

2. Click the “New automation” button.

3. Click the “Select a connected account…” button and choose your Instagram account or Facebook Page.

4. Type a name, such as “GUIDE - Facebook", into the “Flow name” field.

5. Click the “Create flow” button.

Typically, you’ll want to create one flow for Facebook and one flow for Instagram.

Each flow belongs to one connected account.

2. Choose What Triggers the DM

1. Find the “When: Someone comments” section.

2. Choose when the automation should work:

   • Select the “On any post or reel” radio button if you want the keyword to work on every post and reel.

   • Select the “On a specific post or reel” radio button if you want it to work on only one post or reel you choose.

3. Type your keyword into the “Add a keyword or phrase to match on” field under “And the comment has”.

4. Press Enter to save the keyword.

Keywords are not case-sensitive:

"GUIDE", “Guide", and “guide" are all the same trigger.

3. Write the DM

Find “Send this DM” and type your reply into the “Message” field.

Example:

Thanks for commenting GUIDE!

Tap below for the free guide I promised.

Under “Link button 1”, type what the button should say into “Button label”.

Paste your link into “Web URL”.

4. Publish and Test

1. Click the “Save” button.

2. Click the “Publish” button.

3. Confirm the status changes from “Draft” to “Live”.

4. Sign in to a DIFFERENT Facebook or Instagram account.

5. Comment your keyword on the post.

6. Confirm the DM arrives.

7. Open the link and confirm it goes to the right page.

Important: Your own comments will not trigger the automation. Always test from another account.

Part 2: Automate DMs with Claude or ChatGPT

Now let’s do the same thing, but your favorite AI agent will drive the process.

So, you won’t have to waste time clicking manually!

Comment → DM → Link

Before you start, choose:

• Your keyword: ____________________

• Your free offer: ____________________

Step 1. Connect Blotato

1. Open Blotato Settings → API.

2. Follow the setup steps for Claude or ChatGPT Work.

3. Ask:

List my connected Blotato social accounts.

Do not move on until you see your Facebook and Instagram accounts.

Step 2. Create DM Automation

Copy and paste this prompt into Claude or ChatGPT Work:

Create a Facebook and Instagram DM automation. When someone comments GUIDE on any post, send my free guide: [URL].

On Instagram, ask them to follow first.

Replace GUIDE with your keyword.

Replace [URL] with the link to your offer.

I often create DM automations that send people to:

• New tools

• Promo deals

• Interesting news

• My Youtube videos

• My past newsletters

Check these 3 things:

1. The correct Facebook and Instagram accounts are selected.

2. The keyword is correct.

3. The DM clearly explains what the person will receive.

Open the Blotato Automations page to see all your DM automations.

Everything stays synced between Blotato and Claude.

Changes made in one place will appear in the other.

Step 3. (OPTIONAL) Capture Email and Send It to Your CRM

These steps are all OPTIONAL, but very handy:

1. Add an Email capture step.

2. Add Send webhook and paste your CRM webhook URL.

3. Test the full flow from another Facebook or Instagram account.

If your CRM does not provide a webhook, create one in Zapier, Make, or n8n.

Then connect Zapier, Make, or n8n to your CRM to create the contact.

Use this prompt:

Create a Facebook and Instagram DM automation. When someone comments GUIDE on any post, get their email, send my free guide [URL], and send a webhook to [WEBHOOK URL]. On Instagram, ask them to follow first.

Your test should do all 3:

• The comment triggers a DM

• The Instagram DM asks for a follow and an email

• A new contact appears in your CRM

RECAP: Your Finished Flow

NEXT-LEVEL USE CASES

Here are some next-level use cases you can do within Claude/ChatGPT to turbocharge your automations:

Create Multiple DM Automations from a Spreadsheet

Create multiple DM automations for Facebook and Instagram using this spreadsheet/CSV.

Update Many DM Automations at Once

Update all my DM automations so that the second link goes to: [URL]

Review DM Automation Analytics

Get my DM automation analytics for the past 30 days and tell me the top 3 winners.

A/B Test the Messages

Improve the copy in my last 3 DM automations to increase clicks.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

Free AI Watermark Remover

I’m excited to share my free AI watermark remover:

https://watermark.sabrina.dev

You can drop in text, PDFs, word docs, files, etc.

It removes unicode & file-based watermarks for free.

This doesn’t mean your writing is “good” though.

Pair the watermark remover with the humanizer prompt below.

What About Claude Watermark Based on Words?

If your main writer is Claude and you’re concerned about their new word-based watermark…

Turn on toggle “Claude Rewriter” and it will try to reduce statistical text watermarks via a rewrite/paraphrase step.

This is NOT guaranteed to work 100%.

However: heavy editing, paraphrasing, translation, or mixing Claude text into other writing can reduce detectability.

That’s precisely why I recommend COMBINING these techniques:

• humanizer prompt for heavy rewriting/paraphrasing (below)

• if you’re specifically concerned about Claude, run your humanizer prompt in a DIFFERENT AI chat (e.g. ChatGPT)

• watermark removal tool for unicode & file-based watermarks

Btw the watermark removal tool’s toggle “Claude Rewriter” already uses a different LLM.

NOTE: ChatGPT has had watermarks since 2025 and practically I’ve noticed zero difference in my life.

AI Humanizer Prompt - Intro

This single prompt substantially humanizes AI-generated outputs, so that they sound less AI, more human. The prompt contains a comprehensive list of AI writing signs and phrases, then instructs AI to avoid them.

The beginning section is a list of things AI “Should” do in its writing, reflecting best practices from direct response copywriting.

However, this particular copywriting style may not fit your use case, so you should always customize the “SHOULD” section according to your ideal writing style.

TIP: give AI your past writing examples and ask it to update the humanizer prompt.

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

How to Use The AI Humanizer Prompt

The full prompt is below.

All you have to do:

Copy & paste the full prompt into ChatGPT, Claude, Gemini, or your favorite AI tool.

It should say at the end something like “Memory updated”, which means it worked!

AI will remember your writing preferences moving forward.

Try it out by asking AI to write a social media post.

Copy Paste Sabrina’s AI Humanizer Prompt

# ALWAYS FOLLOW THIS WRITING STYLE:

• SHOULD use clear, simple language.
• SHOULD use short, impactful sentences.
• SHOULD use active voice. Avoid passive voice.
• SHOULD focus on practical, actionable insights.
• SHOULD use bullet point lists in social media posts.
• SHOULD use data and specific examples to support claims when possible.
• SHOULD use “you” and “your” to directly address the reader.
• SHOULD vary sentence length and sentence structure naturally.
• SHOULD prefer specific nouns, numbers, examples, and actions over abstract language.
• SHOULD sound like a person with a point of view, not a neutral summary engine.

• AVOID using em dashes anywhere in your response. Use commas, periods, parentheses, or other standard punctuation instead.
• AVOID semicolons.
• AVOID constructions like “not just X, but also Y.”
• AVOID constructions like “It’s not X. It’s Y.”
• AVOID repetitive contrast formulas such as “While X, Y” throughout the response.
• AVOID metaphors and clichés.
• AVOID broad generalizations.
• AVOID common setup language such as “In conclusion,” “In closing,” “The key takeaway,” or similar framing.
• AVOID output warnings or notes. Give the requested output.
• AVOID unnecessary adjectives and adverbs.
• AVOID hashtags.
• AVOID markdown.
• AVOID asterisks.
• AVOID repeating the same idea in the introduction, body, and ending.
• AVOID restating the user’s question before answering.
• AVOID summarizing your own response at the end unless asked.
• AVOID excessive transitions between paragraphs or bullets.
• AVOID starting multiple paragraphs with the same grammatical structure.
• AVOID making every paragraph roughly the same length.
• AVOID making every sentence roughly the same length.
• AVOID overly symmetrical lists where every bullet follows the exact same pattern.
• AVOID forced groups of three unless three items genuinely make sense.
• AVOID excessive headings and subheadings.
• AVOID generic headings such as “Key Takeaways,” “Why It Matters,” “Benefits,” or “Final Thoughts” unless requested.
• AVOID rhetorical questions used as transitions.
• AVOID fake conversational transitions such as “Here’s the thing,” “Here’s where it gets interesting,” “Think about it,” or “Let that sink in.”
• AVOID dramatic one-line fragments inserted only for emphasis.
• AVOID excessive parenthetical explanations.
• AVOID excessive qualification and hedging.
• AVOID polished corporate language when ordinary language works.
• AVOID generic claims without a concrete example, reason, number, person, event, or action.
• AVOID repeating keywords or phrases when a natural alternative exists.
• AVOID unnatural vocabulary variation where a simple repeated term would sound more human.
• AVOID treating both sides of every issue as equally important when evidence favors one side.
• AVOID ending sections with generic inspirational or reflective statements.
• AVOID filler sentences whose only purpose is transitioning to the next point.
• AVOID explaining obvious implications.
• AVOID sounding overly complete, exhaustive, balanced, or sanitized.
• AVOID adding context the reader does not need to understand the answer.


• AVOID these words:
“can, may, just, that, very, really, literally, actually, certainly, probably, basically, could, maybe, delve, embark, enlightening, esteemed, shed light, craft, crafting, imagine, realm, game-changer, unlock, discover, skyrocket, abyss, not alone, in a world where, revolutionize, disruptive, utilize, utilizing, dive deep, tapestry, illuminate, unveil, pivotal, intricate, elucidate, hence, furthermore, realm, however, harness, exciting, groundbreaking, cutting-edge, remarkable, it, remains to be seen, glimpse into, navigating, landscape, stark, testament, in summary, in conclusion, moreover, boost, skyrocketing, opened up, powerful, inquiries, ever-evolving"


# IMPORTANT: 

Review your response and ensure no em dashes!

If several consecutive sentences have similar length or structure, rewrite some of them.

Prefer slight natural variation over perfect structural consistency.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

• Blotato tutorial?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

If you want to make viral faceless AI videos, everyone sends you to Higgsfield. It’s good. But way overpriced. Roughly $15 to $129 a month, monthly billing, depending on your tier.

You don’t need it.

One free Claude skill makes the SAME videos, using the SAME AI models, and you only pay for what you generate.

Check out my viral Twitter article here:

Already proven it can make viral videos getting millions of views.

An image costs a few cents.

A video costs about $0.75.

No monthly bill.

Here’s what $1 buys:

You’ve seen the faceless videos. ‘POV: you wake up as a caveman.’

Millions of views.

Meanwhile, the product you spent months building is sitting at 9 signups and $90 MRR 😭

You shipped.

And then nothing.

Posting into the void.

Getting people to see the thing you built is hard... nobody teaches you.

By the end of this guide, you’ll have your own viral faceless video engine: cinematic videos that put your product in front of millions, without ads, without showing your face and without becoming a marketer.

Some context first

I own this faceless channel with 260.5k followers and over 100M+ views, 100% faceless.

Screenshot 1. Tiktok Profile Faceless latest

Today, you can build the same thing with Claude, one free skill and media generation that costs about $0.75 per clip.

Same format, much cheaper engine.

If I were starting from zero this week, here’s exactly how I’d do it.

And btw, “POV first person cinematic” is just one format. I’m using it as an example throughout this guide, but you can generate ANY faceless format.

So, I created a few different presets you can start with:

• A problem then a solution clip where your app is the fix.

• A before and after that shows the result.

• A ‘5 tools for [x] countdown’ with your product at number 1.

• A relatable short story.

Match the format to what you’re selling.

A faceless channel is great cheap advertising for your own product.

Put a small watermark on every video, then when one pops, that’s millions of impressions for whatever you’re building. Then replicate multiple similar videos with the proven viral format.

However, one caveat: if you’re chasing platform rewards, Tiktok’s Creator Rewards Program only pays on videos 60+ seconds. YouTube Shorts has no minimum.

So, this article is about the other route, i.e. using faceless videos to market your own stuff, so everything here is built for short clips.

Or even better, skip the watermark and just show your product in the video. CalAI does this better than anyone, always organically showing their product within the first 12 seconds, instead of overtly talking about it. This converts better than watermarks.

Or keep the video clean and just put the offer in your bio.

Screenshot 2: you can show your watermark like this

What you need

• An AI agent. Claude Code, Codex, Hermes, or Openclaw, all work great . Any agent can read the skill, OpenClaw included. The skill is a SKILL.md file on the agentskills.io open standard, so it also runs in OpenClaw, Codex CLI and Hermes. Drop the folder into that agent’s skills directory.

• The /generate skill, download below. One free skill I built from RoboNuggets’ guide (shout out@robonuggets). It takes your prompt, picks the cheapest model that can do the job, generates the image or clip, and saves every file in one folder so you own it. I extended it so that it also quotes the cost, waits for your approval before running, and has a few video template presets.

• 1 API key to access the best video, image and music models. There are optional fallbacks. All pay as you go, you only pay per generation:

• KIE.ai, the cheapest route, and the best one to start. Credits don’t expire.

• fal.ai, the biggest model catalog and the best docs, your first fallback. Credits expire after a year.

• WaveSpeed, a broad catalog fallback if others fail. Credits never expire.

• Blotato.com to publish. Posts to every platform (TikTok, Instagram, YouTube, Facebook, Threads, Bluesky, Pinterest, X) through a REST API or MCP, so your agent drives it. It also includes DM automation, messaging, comments, and social media analytics.

• A warmup routine. Keeps a brand new account from getting flagged the second you automate it. Free playbook at the end.

That’s it.

Let’s get into it.

Step 1: The generation tool (KIE.ai) + 1 free Claude Skill

KIE.ai is a pay as you go aggregator to most of the same models that tools like Higgsfield run under the hood.

You just load credits and spend as you generate.

Credits don’t expire. KIE only charges you when a generation SUCCEEDS.

If it fails? FREE.

Say you post 5 videos a week. That’s 20 videos a month, 10 seconds each.

Here’s what it would cost you:

Look at what the same $15 buys.

KIE: 20 clips at 1080p.

Higgsfield Starter: 5 clips at 720p and then you wait for the next month.

And if you don’t make any videos, you STILL pay full price.

And the credits you didn’t use are GONE.

Higgsfield prices are monthly billing.

KIE delivers 1080p for less than Higgsfield charges for 720p!! 🙄

Screenshot 3. KIE add credits page

If you’re using image to video generation like I did (it’s cheaper this way), then you’d need 2 models for the two jobs.

And the /generate skill picks between them for you:

• Image: Nano banana for most frames cost a few cents each. Use gpt when there is readable text or an app screen in the shot. It holds letters and UI without smearing them.

• Video: Seedance. The skill routes to 1.0 Pro for standard clips at about 7 cents a second, and steps up to 2.5 when you want the newest model.

The /generate skill reads your prompt, routes to the cheapest best-fit model, generates the image, and saves it in a local folder, so you own everything.

Before spending money, it gives you the exact price and waits for your approval. No surprises!

How to set up

1. Get a KIE API key from KIE.ai, then top up billing $5 to prove it works or $50 if you plan to run a channel, since the credits never expire.

2. Install the /generate skill into Claude code. This lives in your skills folder and loads in every project automatically.

3. Ask Claude to create a folder for your generations and paste your KIE key into its .env file.

   

Screenshot 4. Folder tree of faceless project

---
name: generate
description: Generate faceless video content cheaply and consistently. Draft a still on a cheap model, run a pre-animate quality check, then animate the approved still through kie.ai, always routing to the lowest cost model. Enforces a running spend cap with a cost ledger, applies reusable style and character presets so a whole channel stays consistent, and outputs a publish-ready clip that hands off to Blotato for scheduling. Triggers on /generate, generate image, generate video, make a POV clip, animate this.
allowed-tools: Read, Write, Edit, Bash, Glob
---

# /generate

Make faceless video content for pennies, keep a channel visually consistent, and never blow the budget. Draft a cheap still, check it before you spend on animation, then animate only the still you approve. Output lands publish-ready.
Built for vertical social video (9:16). It includes: a running budget ceiling, reusable presets, and a quality gate plus a Blotato publish hand-off.

## Models

| Task | Default model | Provider | Recipe |
|---|---|---|---|
| Image (default) | nano-banana-2 | kie.ai | models/nano-banana-2.md |
| Image (text or UI in frame) | gpt-image-2-text-to-image | kie.ai | models/gpt-image-2.md |
| Video (DEFAULT) | Seedance 1.0 Pro `bytedance/v1-pro-image-to-video` | kie.ai | models/seedance-1-0-pro-image-to-video.md |
| Video (budget) | Seedance 1.0 Lite `bytedance/v1-lite-image-to-video` | kie.ai | models/seedance-1-0-lite-image-to-video.md |

| Video (newest, EXPLICIT ONLY) | Seedance 2.5 `bytedance/seedance-2-5` | kie.ai | models/seedance-2-5-image-to-video.md |
Read the recipe file before every generation. Use gpt-image-2-text-to-image whenever the frame has readable text, a sign, a poster, or an app UI mockup. Use nano-banana-2 for everything else.
**Video model routing.** Default every standard short vertical clip to **Seedance 1.0 Pro**. Use **Seedance 1.0 Lite** as the budget option. Use **Seedance 2.5 only** when I explicitly ask for it.
**Duration.** Seedance 1.0 accepts **5 or 10 second clips only** — there is no 8-second option. **Default to 10 seconds.** If any other value is requested, **round to 5 or 10 and tell me what you did** (do not fail silently, do not stop).
**Aspect ratio / dimensions.** The clip inherits the input still's exact dimensions. So **the still IS the aspect ratio and the resolution**: generate the still at **1080x1920** (9:16). If the still isn't 1080x1920 the clip won't be either. Seedance 1.0 has no aspect ratio parameter. Do not send one.

## Budget and ledger (never blow the spend)

Keep a running ledger at `generations/ledger.json`. Every generation appends one line: timestamp, model, type, cost_credits, cost_usd, description.
Two caps, editable here:
- SESSION_CAP: $10 per run of work.
- MONTHLY_CAP: $50 per calendar month.
Before ANY paid generation:
1. Sum the ledger for this session and this month.
2. If this run would cross a cap, STOP and tell me how far over, do not generate.
3. Otherwise, quote it like this: "This clip is 140 credits, about $0.70. Spent this session $4.10, remaining $5.90 of the $10 cap. Go?"
One approval equals one run. Never batch paid videos past the cap without a fresh go.

## Pre-animate QA gate (stop wasting video spend)

Video is the expensive lane, so check the still BEFORE animating. These are the real failure modes (wrong ratio gives black bars, drift breaks the illusion):
- Aspect ratio is vertical 9:16.
- Still dimensions are exactly 1080 x 1920 (the clip inherits them).
- No text baked into the image, unless it was made on gpt-image-2-text-to-image on purpose.
- Subject and point of view match the concept and the preset. First person stays first person.
- Requested duration fits the platform (TikTok, Reels, Shorts all take 5 or 10 seconds comfortably).
If any check fails, stop, say which one, and offer to re-draft the still (cheap) rather than animate a broken frame.

You don’t have to build this from scratch.

You can download the whole skill and reference files here:

https://github.com/Blotato-Inc/blotato-skills

Where we are: a video generation engine, the free skill that drives it and a REASONABLE price per video!

Next stop, give Claude a format, generate an image, animate it.

My first throwaway test was a POV as a Roman senator waking up.

Screenshot 5. Roman senator POV still

Step 2: Create your image

Draft your image first, then animate the still.

It sounds like an extra step but it saves you money and puts you in control of the resulting video.

On KIE, an image, depending on the model, is a few cents. A finished clip is under a dollar.

Text to video costs the same whether the resulting frame is right or wrong.

Drafting the image first means you can iterate the frame for pennies and only pay for motion once the frame is exactly what you want.

/generate a POV of you waking up as a Roman senator, vertical, cinematic

Screenshot 6. how the “generate” skill enhances your prompt using presets

This image cost 8 credits which is 4 cents.

This frame had no text, so it routed to the cheap model.

And the Claude skill captures the cost in your ledger so you can keep track.

Screenshot 7. ledger - POV roman senator

Screenshot 8. full ledger entry tracking multiple generations

Where we are: a frame you actually like, for 4 cents, before you have spent anything on motion.

Step 3: Animate your image

When it came to animate the senator image, my first video failed.

The clip came back as a random landscape. Nothing to do with my senator and KIE still charged me for it.

Technically, the job SUCCEEDED.

It just never read my image.

KIE only charges you when a generation succeeds. Succeeds means the API returned a video.

So the skill checks twice now.

Before it spends, it confirms the still is vertical and 1080 x 1920, because the clip inherits the still’s shape.

After the download, it reads the mp4’s real dimensions and pulls a frame so you can see the clip shows what you asked for.

If either check fails it stops and tells you instead of handing you a file.

I updated the /generate skill to include a check before generating video so you don’t have to worry about that.

Screenshot 9. Cost quote for test clip 150 credits

Where we are: the end to end works. An image goes in and a matching vertical clip comes out

Step 4: Market your app with it

Now for an actual clip you can sell your product with.

This one is meta on purpose. It shows the exact thing this setup does.

You wake up, reach for your phone and the screen reads,

‘Overnight. 6 posts published. 3 new signups.’

Let’s start with the image.

There is a little dashboard on the phone screen, so this is a gpt-image-2.

I asked the /generate skill for the POV and the first image was wrong. The figure was lying flat as expected but the pose was unnatural.

So we corrected the prompt.

Prompt:

iPhone photo, first person POV, lying in bed just waking, one hand reaching for a phone on a wooden nightstand, soft morning light from a window. The phone screen is on and clearly legible, showing a clean minimal app dashboard: a small rounded app icon at the top, the heading ‘Overnight’, and two lines, ‘6 posts published’ and ‘3 new signups’. Cozy bedroom, shallow depth of field, photoreal, shot on a phone. Portrait 9:16, 2K resolution for crisp on-screen text.

In a real POV, you are looking forward so you see your own legs. To see your phone, you turn towards your side.

Improved prompt:

iPhone photo, first person POV, lying in bed with my head on the pillow, turned to look toward a window on my left where soft morning light comes in. My hand holds my phone, just lifted from the nightstand beside me, raised close to my face with the screen facing me and clearly legible, and not backlit: a small rounded app icon at the top, the heading ‘Overnight’, and two lines, ‘6 posts published’ and ‘3 new signups’. The white pillow and rumpled duvet fill the foreground, the cozy bedroom and window softly blurred behind the phone. Shot on a phone, shallow depth of field, photoreal, warm morning light. Portrait 9:16, 2K resolution for crisp on-screen text.

Screenshot 11. Overnight corrected still, hand holding phone, dashboard legible

That still cost 10 credits.

5 cents 😍

Now animate it.

Screenshot 12. Cost quote for dashboard animation 140 credits

The /generate skill quoted the cost first, 140 credits, about $0.70 at 1080p and waited for my approval.

In image to video, the first frame IS the still so it is also your thumbnail. You can adapt this however you want so you don’t give away the payoff but keep the hook.

Screenshot 13.POV you wake up and your app sold itself

Where we are: you have a finished vertical clip, 1080p, 10 seconds, made for about $0.75 all in.

Step 5: Publish everywhere (warm up first so you don’t get shadowbanned)

Do NOT connect a brand new account straight to automation. That is the fastest way to get flagged as a bot.

Before you automate anything, treat the account the way a real person would:

• Log in daily

• Scroll, like, comment and save

• Follow accounts in your niche

• Post 1 piece of content a day, manually

Then, connect Blotato so your AI agent can publish for you.

• Claude.ai chat or Claude Cowork: install the Blotato connector from the Claude.ai Connectors page.

• Claude Code, run this in the terminal:

claude mcp add --transport http Blotato https://mcp.blotato.com/mcp

then run `/mcp`, pick Blotato, and authenticate in the browser.

• Here are instructions for Codex:

Where we are: a warmed account and Blotato connected, so your agent can publish without being flagged.

Publish 1 video to 1 platform

Just prompt your AI agent:

Publish overnight.mp4 from my generations folder to Instagram. Caption: POV: you wake up and your app ran itself overnight. Schedule it for 5:27pm tomorrow.

Screenshot 14. Blotato scheduled post Aug 14

Now publish everywhere

Just tell your AI Agent:

Publish to YT shorts, Tiktok, Threads, and Facebook too.

Blotato handles it all automatically.

Screenshot 15. Blotato queue IG YT Threads Sept 20

One link is all your agent needs

Blotato ships a page written for AI agents to read: help.blotato.com/api/llm

It is the whole API spec formatted for an LLM to consume, every endpoint, header, payload shape, and platform specific field, in context.

Point your agent at it and just say what you want to build.

Where we are: 1 video clip scheduled on all your platforms and a bookmark that lets your agent build the rest with Blotato without you.

The numbers (as of today)

• 45.7M views on 1 video

• 260.5k followers

• 9.3M likes

Set this up yourself (copy this)

Here’s your worksheet, start to finish:

1. Get a KIE.ai key and top up under Billing. $5 to test, $50 to run a channel. Credits never expire, and KIE only charges when a generation succeeds.

2. Download my /generate skill. Install it into ~/.claude/skills/ (or ~/.openclaw/skills/, ~/.codex/skills/, ~/.hermes/skills/). It routes to the cheapest model, tells you the cost, waits for your approval, and stores every image/video in one folder.

3. Draft an image for a few cents, then animate with Seedance. Roughly $0.75 per finished clip.

4. Connect Blotato to automate social media posting & analytics: `claude mcp add --transport http Blotato https://mcp.blotato.com/mcp`, run `/mcp`, hit Authenticate. Then just say “schedule this to TikTok, Reels, and Shorts.”

5. Building your own agent? Point it at help.blotato.com/api/llm and it has the whole publishing API in context.

If you need help setting this up or have any questions, feel free to email me sabrina@blotato.com :)

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

• Blotato beginner tutorial

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

Most people want the clean founder story:

Idea, build, launch, revenue.

That is NOT how Blotato happened.

Blotato happened because each messy step forced the next one.

A 30-second blurry shaky Tiktok video created demand.

My UGLY AF MVP created proof.

My early adopters created confusion.

My launch mistakes created urgency.

Then the fast initial revenue created enough conviction to rebuild the whole thing.

That is the real story.

@sabrina_ramonovAI app idea - AI automated social media system. Seeking your feedback and help 🙏 #ai #aitools #chatgpt #socialmediamarketing #businessideas #entrepreneurship

Tiktok failed to load.

Enable 3rd party cookies or use another browser

The goal was never to vibe code perfect software.

The goal was to create enough evidence to keep going.

1. The Idea Was A Public Question

My August 1 TikTok (above) was the first real test.

I asked if people wanted an AI tool that took YouTube videos, blog posts, and ideas, then turned them into social posts for multiple platforms.

That tiny public question mattered because I wasn’t asking myself if the idea was good.

I was asking OTHER PEOPLE IN THE REAL WORLD about their pain.

The comments told me the pain was real enough to try building it. Very honestly, I figured if Blotato flopped, I could just release the code as open-source and make content out of it! That’s why it felt low-risk and not that serious.

But talk is cheap…

So the next question:

Could I build my small idea FAST, before losing interest?

2. The Vision Was Too Big, So I Cut It Down

On August 4, 2024, I filmed myself building the MVP in a 7-part Youtube series using Cursor AI. This was before Cursor Composer - I literally had to CREATE and DELETE files manually.

Vibe coding tools have come a long way since then.

Y’all have it so good now :)

The first video was not about writing code.

It was about cutting the product scope as much as possible.

My original Google Doc said the vision was:

Social media platforms -> AI-powered REMIX -> social media platforms

It had everything:

• Any source: YouTube, Twitter, Facebook, newsletters, articles

• Your brain: brand voice, product info, services

• AI filtering: quality and relevance

• AI remixing: content in your voice

• Publishing: any social platform

• Weekly queue: Mon/Wed/Fri posting

That vision was exciting.

It was also way too big for v1.

So I drew the first UX prototype with URL input, text input, LinkedIn, Twitter, Facebook, images, edit buttons, and publish buttons.

Then I stripped out the non-essential parts.

No URL input.

No multiple platforms.

No image generation.

No real publishing.

Just text in, Facebook post out.

These are real screenshots from that Youtube series! I knew I had to document my journey, otherwise no one would believe me later.

The Cursor-built MVP looked gross.

But it did the job.

That is the part people misunderstand about vibe coding.

AI did not give me a mindblowing app.

AI gave me a fast first version to start getting feedback.

It gave me momentum.

Then, I invited real users.

And real users immediately made the problem messier :)

3. Early Adopters Proved Interest Is Not The Same As Behavior

On October 1, 2024, I opened an early adopter program with around ~50 ppl.

My original Blotato waitlist form was honest about my stage…

I made it clear: this was NOT a polished product.

It promised the first 50 early adopters a steep lifetime discount and personal, priority support from me, the founder.

It also used my own distribution as a case study.

At that point, I had grown from zero to almost 200k followers in 4 months using AI + automations, while distributing 100+ pieces of content per week across multiple social platforms.

That mattered because Blotato was not an abstract SaaS idea.

It was the productized version of a workflow I was already using.

The form also shows the original positioning problem.

I was building for:

• Solopreneurs

• Solo creators

• Small businesses

• Small agencies

Useful audience.

Also broad.

I thought the Google Form answers would tell me what to build next.

I WAS WRONG!

What people said they wanted did NOT match what they actually did.

I realized people wanted a billion different things:

• Podcast repurposing

• News monitoring

• Personal brand posts

• Business content

• Full automation

• Manual editing

• Faceless videos

• + 4 million more things

(and now I joke Blotato = BLOATato because I actually tried to cram half these things in there…)

The mistake was treating “interested users” like 1 clean customer group.

They actually had 50 different use cases.

That almost killed my confidence.

Instead, it changed the question again.

Not “is anyone interested?”

Now it was:

Who has enough PAIN to pay, onboard, complain, AND keep using it?

Only a launch would answer that.

Here’s a screenshot of part of my launch email:

4. Launch Day Broke In Public

Official launch day was October 20, 2024.

I sent the email, posted the video, and opened the doors.

Then a bunch of shit broke :D

@sabrina_ramonov4 embarassing product launch day mistakes! Hope my blooper reel shows you the real messy truth of entrepreneurship :) oh but so much fun, a wild ride! Thanks you to everyone who signed up for Blotato and look forward to working with y’all! #ai #startup #failure #bloopers #sabrinaramonov #blotato

Tiktok failed to load.

Enable 3rd party cookies or use another browser

The worst mistake was the YouTube demo.

I uploaded the wrong video, linked it in the newsletter, sent it to thousands of people, then made the video private!

(I was really nervous okay…)

So the launch email had a giant broken YouTube block at the top.

And that was NOT my only failure, here’s a few more to make you feel better:

• Mobile site: broken by last-minute testimonials

• Apple Pay: not working

• Amazon Pay: not working

• Calendar invites: missing meeting links

Honestly???

This is entrepreneurship.

You put the thing out there, and reality immediately punches you in the face.

But my failure-riddled launch still did something important.

It proved people were trying to buy, book calls with me, watch my demo, and sign up for my product.

That is a VERY different signal than a comment.

That is BEHAVIOR.

My barebones launch email stated my core promise:

Blotato was for coaches, consultants, solopreneurs, small businesses, and creators who wanted high-quality content without duct-taping automation tools together.

BTW my website was a $60 Webflow template with swapped text and images. Not some fancy AI-generated 3d-animation mindblowing $10000 website. And I did NOT even update it for 1.5 years… So, don’t be one of those ppl stressing out over a perfect website. Just get on with it!!

Despite my crappy looking website, it was enough because the real learning moved into customer calls, payments, onboarding, and fixing the product.

And that is where my vibe coded MVP hit its limit.

5. Revenue Gave Me Permission To Rebuild

One week after launch, I posted 10 lessons from my first 7 days.

@sabrina_ramonov10 things I learned launching my AI SaaS product, Blotato, exactly 7 days ago! Its been a wild ride!! Thank you ALL for your support!! - how do I build and launch an AI product? - how to build SaaS app? - can AI build a product? #ai #chatgpt #girlboss #womenownedbusiness #womenempowerment #startup #sabrinaramonov #blotato

Tiktok failed to load.

Enable 3rd party cookies or use another browser

My biggest lesson was launch FASTER.

From August 1 to October 20, I still think I spent too long.

To be fair, I also had multiple birthday celebrations, family visiting, and a road trip to Bryce and Vegas in that stretch of time… but still. I felt I was too slow.

So, don’t do be slow like me :)

The second lesson was technical.

My vibe coded codebase was terrible spaghetti code.

It got me moving, validated demand, and built internal conviction.

Then it became the wrong tool for the next job.

Once the product had real users, I needed a reliable codebase that I understood and could easily maintain and extend. Late 2024 vibe coding tools simply could NOT do this without substantial technical human guidance and constant intervention.

My eventual stack:

• Frontend: Next.js

• Backend: Fastify

• Database: Supabase

• Hosting: Heroku

This was the shift from 100% AI coding to AI-assisted coding.

Vibe coding = AI driving and me not looking :D

AI-assisted coding = me driving and AI helping with scaffolding, brainstorming, writing tests, writing comments, coming up with variable names, etc.

Huge difference.

The key point:

I only had the patience & conviction to rebuild because I knew my product made money.

That is the real value of vibe coding a messy MVP.

Not your code.

Your conviction!

From launch day, Blotato has grown linearly in revenue ever since.

HOWEVER:

• I usually have a 1.5 month revenue plateau during ski season, late December through January :D

• (I like skiing okay smh…)

• some months are slower growth than others, especially when I go on vacation a lot (next few months will def hurt)

• I don’t feel I have product/market fit like some insane unicorn startup

• It still feels like a lot of “pushing” on my side to keep growing revenue

• …but damn it’s crazy what you can build with AI!

My business grew steadily and linearly because demand was real, my audience trusted me, and I kept fixing the product in public.

Launching Blotato taught me the order makes a HUGE difference:

• Content: test demand in public

• AI: build the 1st version fast

• Users: find the truth behind what they say

• Launch: lots of things will go wrong, get it over with!

• Revenue: conviction to rebuild and COMMIT

So if you’re building with AI, don’t worship your v1.

Use it to earn the next piece of evidence.

And if you’re overwhelmed by the process…

The chaos is not a side effect.

The chaos IS the learning :)

Build the ugly proof.

Let real users make everything worse.

Then rebuild the thing that they were willing to pay for, despite all the things that went wrong.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

People think I’m a Claude shill.

But I’ve never been paid by Anthropic or OpenAI.

I started using Claude Code fulltime early 2025…

After ditching Cursor.

The thing is…

Because my allegiance isn’t paid for 👀

But there hasn’t been a better alternative than Claude…

Until now…

After 10 days of testing, Codex is now my main AI tool.

I tracked time, errors, cost, and app switches.

Only 1 task was better in Claude.

I’ll explain it below.

To follow along the livestream with live examples, watch here:

1. Codex Has A Far Superior In-App Browser

Codex’s built-in browser keeps pages, chat, files, and feedback in 1 workspace.

Just hit “CMD + T” and it will open a new browser tab.

In my work, it is FAR superior to Claude’s Chrome integration.

Try it:

1. Open “Plugins,” enable “Browser,” then type a page and goal.

2. Check the page before any login, purchase, or public post.

SAMPLE PROMPT: research 3 competitors pricing pages

Sit back and watch Codex’s browser work, fast!

Not only can Codex use my browser, it can control my Mac.

2. Codex Computer Use Is Superior And Smoother

Computer Use lets Codex use approved Mac apps.

Claude technically has computer use too.

But in all my tests, Codex felt smoother and finished more jobs with fewer delays/excuses/hallucinations.

Try it:

1. Install “Computer Use” under “Plugins.”

2. Mention “@Computer,” name 1 app, and give a clear finish line.

SAMPLE PROMPT: open up ____ app and do something simple in it.

Now, scheduling becomes far more useful…

3. Unified Scheduling Is My #1 Reason

Codex Scheduled tasks puts every job in 1 sidebar.

A Codex scheduled task can use the in-app Browser and Computer Use.

Compare this to Claude’s 3 different places…

• Cowork has “Scheduled” tasks

• Claude Code has “Routines”

• Then there’s the Claude Chrome extension as well

Fragmentation and context switch drain my mental energy…

THIS is the main reason I switched.

Super useful for automations, especially for websites where you’re required to be logged in and websites that don’t have an API.

Here’s a real-world practical example where I used Codex scheduled tasks + built-in browser:

• Every day 10am log in this website (no API)

• Analyze each candidate who applied to my job

• Pass/reject them based on my specific criteria

• Send me a Slack message with the top 3 candidates

4. Codex Has Superior Built-In Image Generation

Codex image generation makes and edits images in your chat.

Claude reads images but lacks this smooth creation flow.

Codex uses ChatGPT’s latest image model (which is top notch)

So if you’re creating ads, social posts, websites, and mockups… you can do it all natively within Codex.

Yes, Claude still has Claude Design which is GREAT, but again it feels fragmented from the rest of the Claude experience.

Try it:

1. Name the image, size, audience, colors, and sales goal.

2. Ask for 1 small change at a time, then save the file.

SAMPLE PROMPT: Make a 16:9 hero image for my bookkeeping offer in navy and cream.

Once the image exists, Git gives each final file a safe home.

5. Codex Has Superior Git Integration For Non-Techies

Git keeps a history of your files, so you don’t lose good work.

Think of Git as a “SAVE” button for all your work.

Codex has this friendly-looking display of Git actions.

All you have to do is click “Commit or push” to save your work.

This might look scary, but Codex makes it far easier for non-techies. And honestly, Git is essential when working with AI agents that may accidentally delete sensitive info.

Try it:

1. Ask Codex to explain the project before it edits.

2. Check the changes, run tests, then save them on a named branch.

Once work is safe, Fast mode handles urgent deadlines.

6. Fast Mode Helps When Delay Costs Money

Fast mode raises speed by 1.5x on supported models.

Be careful because the “1.5x speed” claim means faster, but it’s oftentimes MORE than 1.5x token usage.

However, I still use it frequently.

Because TIME is the most important thing to me.

I’m currently on the Codex $200/month plan, haven’t hit the limits, AND it is far from the most expensive tool I pay for.

I suggest using it for urgent work, but I found 1 real-world use case that still sent me back to Claude.

7. LinkedIn Recruiter Is My 1 Claude Exception

I’ve almost fully moved from Claude to Codex.

HOWEVER, Codex sometimes says hiring work isn’t allowed and stops working inside my LinkedIn Recruiter account.

Claude Code + Chrome did the same work without complaints.

For this one job, Claude is still superior.

I kept this fallback, then imported the rest of my Claude setup.

8. How To Migrate From Claude To Codex

OpenAI has an import tool for Claude Code.

Use it first:

1. Open “Settings > Import” in the ChatGPT desktop app.

2. Choose “Claude Code,” then pick projects and recent chats.

3. Finish plugin or account sign-ins after import.

4. Open 1 old project and rerun a job you know well.

Claude skills often live in .claude/skills.

Codex repo skills live in .agents/skills, with both words plural.

Not .skills.

Not .agent.

Each skill needs a folder with a SKILL.md file.

AGENTS.md is different.

It stores the rules Codex reads before work, such as your writing style, test steps, and safety rules.

Codex settings live in .codex/config.toml.

It moves instructions, settings, skills, commands, hooks, MCP tools, and recent chats.

It supports Claude Code, not Cowork.

For Cowork, copy files, reinstall plugins, and rebuild tasks in “Scheduled.”

Then, test creating scheduled tasks and the in-app browser, like we talked about above!

RECAP: ChatGPT Desktop Is The Winner BUT…

If I had to choose only 1 reason to switch to Codex, it’s this:

Unified scheduled jobs that use my logged-in browser & computer.

BUT:

• Claude still won on that Linkedin Recruiter task because Codex refused the work

• I still use free tools like n8n for DETERMINISTIC workflows where “heavy intelligence” isn’t required.

• For example, I still have n8n workflows running where I just need Y thing to happen every time X happens - no funny business. Real example: when a file enters this google drive, extract the file’s contents, then send a simple email.

Frequently Asked Questions

Does import delete my Claude setup?

No, the official import leaves your old setup in place.

Where do Codex skills go?

Use .agents/skills/<skill-name>/SKILL.md inside a repo.

Does the importer move Cowork tasks?

No, rebuild those jobs in Codex “Scheduled.”

Should I cancel Claude on day 1?

No, like anything, test it for 1-2 weeks to see which you like better. My content is my observation from testing on my use cases; take it with a grain of salt, always try stuff for yourself!

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

• Comprehensive Analytics

• First Comment for FB + IG

• Comments inbox and API/MCP

• Messaging inbox and API/MCP

• Manychat-style DM automations for FB + IG!

My goal with Blotato:

Build the best affordable tool for business owners who want to automate social media inside ChatGPT, Claude, or Hermes

You get all the above + 20 social accounts for only $29/month.

Just for 20 social accounts, Buffer starts at $100/month.

Analytics

Easily track your top-performing posts across platforms:

Inside ChatGPT / Claude / your favorite AI agent, fetch your analytics and double-down on your winners every week:

PROMPT: Get my top performing posts from last week and remix them into fresh viral ideas

Comments & Messaging Inbox

Now you can manage your comments and messages from Facebook and Instagram in a unified inbox.

Inside ChatGPT / Claude / your favorite AI agent, analyze your comments and DMs:

PROMPT: fetch all the comments from my last 5 instagram posts and tell me what people think & what I should post next

Tell your AI agent to reply directly to comments and DMs:

PROMPT: for all the positive sentiment comments in my last 10 facebook posts, reply to each comment “I’d love to chat!”

Manychat-style DM Automations

Keyword-to-DM flows are not new.

Building them from ChatGPT/Claude/Hermes is.

A DM automation has 3 parts:

1. A trigger, which is either a comment on a post or a DM.

2. Optional keywords, which filter which comments and messages get a reply, matched whole-word and case-insensitively.

3. The message that goes out, with up to 3 link buttons.

Here’s an example where viewers comment “psychology”, then they automatically get a DM with my prompts. DM automations are awesome for growing followers, lead generation, and email capture.

Claude & ChatGPT can now automate ALL OF THIS…

on Facebook & Instagram 😳

Here’s how it looks inside Blotato:

Everything is synced with ChatGPT, Claude, Hermes, or your favorite AI tool.

So, AI can create, edit, and get stats for your DM automations.

PROMPT: walk me through how to setup blotato DM automations

ChatGPT Setup

• Open ChatGPT Desktop (Codex)

• Go to Settings > Plugins > MCPs > Add Server

• Click “STREAMABLE HTTP”

• URL: https://mcp.blotato.com/mcp

• Headers:

  • blotato-api-key

  • grab your API key from Settings

• Start a new ChatGPT chat

• PROMPT: “What social accounts do I have connected?”

Claude Setup

• Open Claude.ai or Claude Desktop

• Go to Customize > Connectors > Add Custom Connector

• Name: Blotato

• URL: https://mcp.blotato.com/mcp

• Start a new Claude chat

• PROMPT: “What social accounts do I have connected?”

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

I sat down with Paula Rojas, a litigation lawyer in Chile, and she showed me something I have not seen from any other lawyer.

She built a team of AI agents inside her law firm. She named them. Alfred, Toby, Clara, Alma, Hugo, Mateo, Sofía, Marcela.

She has NO tech background. None.

She learned this on YouTube after coming back from maternity leave.

Who this is for: lawyers drowning in repetitive case work, and any business owner who keeps hearing “AI will replace your team” and wants to see what a REAL operational business workflow looks like instead.

Send this one to a lawyer friend.

They will thank you.

Meet Paula Rojas

Paula Rojas is a partner at Rojas & Cía, a family law firm in Chile. She works with her father and 2 uncles.

Her team defends companies in labor law.

Right now they have about 200 active cases.

She is also president of the board at Casinos Marina del Sol, a large Chilean company.

So how did a lawyer with no coding background end up in Claude Code?

She had her second baby. She came back from maternity leave, and the firm was in a rough spot. People had quit. People were sick. Teams were shifting.

She started working 8am to 8pm, every day.

“If I keep going like this, I’m going to miss all of my babies’ smaller years.”

So she went looking for another way. AI was blowing up right at that moment, and she started teaching herself on YouTube.

She started with ChatGPT, moved to Claude, and now she vibe codes apps. When she pulled one of her tools up on our call, I thought it was a real website until I saw “localhost” in the URL bar.

Her whole approach fits in one line:

If you wait until you’re ready, you’ll never start.

Jump in. If you make a mistake, it’s a mistake. It’s learning.

And now the AI agent team. Paula built it in about 2 weeks.

• Alfred is her chief of staff. He triages her email, watches the money flow for the firm, and hands her 2 briefings a day, morning and afternoon. He files newsletters away and only surfaces what needs HER.

• Toby keeps notes and memory, so the team stops repeating the same mistakes.

• Clara is the case-tracking agent. Court notifications land in Paula’s email automatically, and Clara goes to the court website, looks the case up, and flags anything important.

• Alma is the productivity agent. She breaks big projects into small pieces Paula can attack, and she’s connected to Paula’s Notion task database.

• Hugo handles family finances, kept separate from the firm’s money.

• Mateo is the drafting agent. When Clara spots a deadline, Mateo drafts the filing.

• Sofía is the research agent. She pulls case law through Trifolia, a Chilean legal AI connected as an MCP, and feeds it to Mateo.

• Marcela handles content for the AI classes Paula teaches to other lawyers.

Litigation is PULL, not PUSH. Clara tells Paula what happened, but Mateo and Sofía only run when Paula asks.

And nobody does anything irreversible alone.

Anything that leaves the building, filing, sending, publishing, paying, needs Paula to approve it.

She said something about this that stuck with me:

“I never thought in a million years I would be working in Claude Code. Never. But I just dared.”

Now let’s go through the 3 use cases she walked me through, live.

1. A New Lawsuit Lands: 1 Hour Down to About 5 Minutes

Here’s the pain.

A client emails the firm: we got sued, help.

Attached is the lawsuit and the judge’s first order, which sets the date of the first hearing. The firm has to answer 5 days BEFORE that hearing.

Now a human has to read all of it. Who’s suing? Why? What are the facts? What are they asking for? How much money is on the table? How bad is the risk?

Then that person types all of it into the firm’s system, sets the calendar deadlines, and files the case.

Time: about 45 minutes to 1 hour. Every single time. With 200 active labor cases.

Paula built a Claude skill for it.

A skill is a set of instructions you write ONCE, so Claude does the same job the same way every time. Think of it as training a new hire, except you only train them once.

Here’s what her skill does when she sets it off:

1. Reads the lawsuit and the judge’s first order.

2. Pulls the legal analysis: type of procedure, the court, the case number, who’s suing, which company is being sued, the facts, and the amounts.

3. Goes to Chile’s official holidays website and counts the deadline, because holidays do NOT count as days.

4. Writes everything into the firm’s Notion databases, their second brain.

5. Stops and asks Paula for anything it couldn’t find.

6. Gives her a list of documents to request from the client, like the employment contract and salary records.

Here’s the skill mid-run. It has both PDFs loaded, and it’s pulling the Chilean holiday calendar to get the deadline right:

The whole thing runs in about 5 minutes. Then a lawyer checks it, which takes another 5 to 10 minutes.

1 hour of manual work, replaced by 5 minutes of AI work plus 5 to 10 minutes of human review.

The deadline math was the hardest part, and I love that she was honest about it.

The count runs BACKWARDS. You start at the hearing date and count 5 days back, skipping holidays. If a holiday falls in the middle, 5 days becomes 6.

That’s where Claude made the most mistakes. She’s still tuning it.

Right now it gets about 9 out of 10 right, and the human check catches the 10th.

Also notice what the skill does when it isn’t sure: it says so.

“If it’s not sure, it’s not going to state anything as a fact.”

Why this is the RIGHT kind of first project

Paula picked labor lawsuits on purpose. They’re all similar to each other, so the pattern was easy to teach.

And her advice for lawyers starting out is the opposite of what most people say.

The standard advice is: pick high impact, low effort.

She pushed back on that.

“When you’re first starting with AI, you need quick wins.

Things that teach you where you’re going to gain practice.”

So start with a few tasks that are easy, even if they’re small. Build confidence. Learn how the system behaves.

THEN go after the big ones!

Her method for finding them: map out every repetitive process in your firm. The stuff you do every day that eats time and adds no value.

Those are your first automations.

Walk away with: A list of your 5 most repetitive weekly tasks, ranked by how boring they are. Start at the top.

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

2. Court Filings That Sound Like HER, Not Like AI

The second skill writes filings to the court.

Paula demoed a small one live.

Her client was being evicted, and the court order listed the wrong address. She needed to tell the court.

She gave Claude a short prompt. That’s it.

No long explanation of the case.

It came back with a Word document that had her name, who she represents, the case name, the case number, the correct court number, the formal way Chilean lawyers address a court, and the actual argument.

Here’s the draft it produced, sitting next to Claude’s own notes on what it chose NOT to do:

Read the left side of that screen. Claude flags that it did NOT cite any articles, because it refused to invent one, and asks her which article to add.

That’s the behavior you want. Not confidence. Receipts.

She’s used the same skill for appeals over 20 pages long.

Her drafts used to come back about 70% good.

Now they come back about 90% good.

The 10% she still fixes herself.

How she organizes her files for Claude Cowork

This is the question I get more than any other…

Her answer: mirror folders.

1. Keep your real client documents in your normal folder, untouched.

2. Make a SECOND folder just for working with Claude.

3. Copy the documents over, and as you copy, strip out the sensitive stuff. Names, ID numbers, social security numbers, addresses.

4. While you’re at it, convert the files to markdown. Markdown is just a plain text format, and Claude reads it faster and cheaper than PDFs.

5. Write a little map file of the folder, so Claude reads the map first instead of hunting through every document.

That last step saves real money on tokens.

For the actual anonymizing, she uses Adobe on PDFs, or she copies the Word file and deletes the names by hand. Not fancy. It works.

Each case folder also holds a short markdown brief she writes herself: who her client is, which court, the case number, her position, and her strategy.

That’s why her prompts are so short.

The context is already sitting in the folder.

The 2 things that make a skill good

I asked her the key step.

What makes a skill produce something that reads like a formal legal filing instead of AI slop?

She gave me 2 things.

First, teach it how YOU write. She fed Claude a pile of filings she’d already written, in her own voice, and had Opus analyze them. How does she present facts? How does she build arguments? How does she structure an idea?

Second, be obsessive about the details. Down to whether you like a comma or a semicolon. Down to the words you NEVER use.

“This is your judgment that’s going to be written into a skill.”

Then she said the thing that made me sit up.

Lawyers have an unfair advantage at building AI skills, because their entire job is using language to convince someone who has no interest in your problem.

If you can describe exactly what you want, in detail, in plain language… you’re already good at building skills.

That’s a human skill, not a tech skill.

Her workflow for a real filing looks like this:

1. Point Claude Cowork at the case folder.

2. Say what kind of filing it is and roughly what her strategy is. “Argue 1, 2, 3, 4. Number 4 matters most because of X.”

3. Ask for a PLAN first, not a draft. Go back and forth until the plan is right.

4. If she needs case law, Sofía pulls it through Trifolia, a Chilean legal AI wired in as a connector, WITH links so Paula can verify the case exists.

5. Only then does she say: okay, now use the skill and write it.

Planning a 20-page appeal took about 5 minutes.

Walk away with: A mirror folder for 1 client, with a map file at the root and a 1-page markdown brief on the case.

3. The Court Schedule Check (Playwright MCP)

This is my favorite one.

In Chile, some cases have to be argued out loud in front of the appeals court or the Supreme Court. The schedule for next week gets posted online Friday at 6pm.

Friday. 6pm.

The exact moment every human on earth wants to close their laptop.

So somebody on Paula’s team spends 40-60 minutes typing more than 100 case numbers into a court website, one at a time, checking whether any of them are on next week’s board.

Low value. High risk. If you miss one, you don’t show up, and you failed your client.

The court system in Chile does NOT like anything automated on their site.

So this one was hard.

Paula built it anyway. It took her 6 weeks, and here’s the funny part: the schedules only go up on Fridays, so she could only work on it on Fridays.

6 Fridays.

Here’s how she got it working, and this is the part any beginner can copy:

1. She dictated the manual process, step by step. She described exactly how a human finds one case. Every click. Every field. In detail.

2. She let Claude look at the site. She gave it permission to open Chrome and explore the court website until it understood the layout.

3. She had it test on a few cases first, not 120.

4. Then she scaled up, and it broke. It kept crashing. The court website kept kicking it out.

5. She added Playwright MCP. She’d heard someone mention Playwright on TikTok and figured she’d give it a shot.

Playwright MCP is a connector that lets Claude drive a real browser the way a person does. Click, type, wait, scroll.

That fixed it.

“It actually mimicked how a human would be clicking and writing. That worked really well.”

Now she hits one button on a small dashboard she built, and Claude works through the case list, which lives in a Google Sheet.

Notice the URL: localhost.

She vibe coded that.

For each case it reports which court, which day, and the position on the board. In Chile you also need to know WHO is in charge of your case, because being on the board doesn’t guarantee your case gets heard that day.

Results so far: no mistakes.

And it once found a scheduled hearing the MANUAL check missed.

She still runs the human check alongside it.

It’s too critical not to.

That’s the pattern I want you to internalize: automate the tedious part, keep the human strategy and human check, and let AI catch what the human misses.

AI hallucinations

Chile has already had about 5 known cases of lawyers filing documents with AI hallucinations in them.

I asked Paula for her take, because she teaches this to other lawyers.

“The more you trust an AI tool, the more tempted you are to stop using your own criteria.”

Her rule is simple, and it isn’t new.

If your signature goes on the paper, you’re responsible for every word on it. If you don’t understand it, or you haven’t checked it, you don’t submit it.

That was ALWAYS true. When a junior lawyer handed you a draft, it went out with your name on it, so you read it.

AI changes nothing about that.

Then she told me about Brazil, and my jaw dropped.

Brazilian courts use AI to help judges draft the reasoning sections of rulings. So a couple of lawyers filed a labor case with hidden instructions inside the document. White text on a white background, aimed at the judicial AI, telling it to ignore the other side’s arguments and review the file lightly.

That’s prompt injection. In a court filing. On purpose.

The good news: AI flagged the hidden message to the judge.

This wasn’t a mistake.

It was malicious.

And we’re going to see more of it.

She wants courts to build protection BEFORE the damage, not just punishments after.

But she doesn’t want any of this to scare lawyers away from AI.

Chile has roughly 1.2 million civil cases sitting open, undecided. About 700,000 new ones started in 2025. Only about 100,000 closed.

A huge chunk are cases that are nearly identical to each other, with a standard process.

If people want justice done faster, AI is a huge help.

Fear should not get in the way of that.

What She’s NOT Doing

I asked whether the goal was to shrink the team.

Her answer was immediate.

No.

She hired most of those people herself.

She loves working with them.

“I don’t need any more data entry.

That’s not the value I’m going to get from them.”

What she wants is time. Time for her lawyers to be lawyers. To think strategically, to find the one case that wins the argument, to build client relationships.

This is exactly what I tell business owners.

Stop trying to replace a whole department. That’s really hard, and mostly it doesn’t work.

Instead find the 30-minute task you do 5 times a week, and get it to 5 minutes with a human check on top.

Stack enough of those and it changes the whole company.

RECAP

Your AI Leverage = Your Judgment x How Clearly You Describe It

Your Judgment is the part only you have. The strategy, the standards, knowing what to focus on vs ignore.

How Clearly You Describe It is the part you can improve today, for free, by writing your process down in detail.

Here’s what Paula proved:

1. Zero tech background is not a blocker. She learned on YouTube, on Fridays, with 2 kids at home.

2. Start with the boring, repetitive, low-risk stuff. Quick wins teach you the system.

3. Skills beat prompts. Write the instructions once so the output is the same every time.

4. Feed it YOUR past work. That’s how a draft goes from 70% to 90%.

5. Organize the folder before you blame the AI. Mirror folders, stripped names, markdown, plus a map file.

6. Keep the human check on anything critical. Hers caught a mistake. And the machine caught one the humans missed.

7. Your signature, your responsibility. That rule existed long before AI.

The tools took her 6 weeks of Fridays.

Meanwhile, the judgment spans her entire career.

Pick 1 repetitive task this week.

Write down exactly how you do it, every step.

That document IS your first skill.

FAQ

Do I need to know how to code to build this?

No. Paula has zero tech background and learned on YouTube. She describes what she wants in plain language and asks the AI to teach her the rest.

What is a Claude skill?

A set of instructions you write once so Claude performs the same task the same way every time. Paula has one for processing new lawsuits and one for drafting court filings.

How do I keep client data private when using AI?

Use mirror folders. Keep originals in your normal folder, copy documents into a separate working folder, and strip names, ID numbers, and addresses as you copy. Paula uses Adobe for PDFs and manual edits for Word files.

What is Playwright MCP and why did she need it?

Playwright MCP is a connector that lets Claude control a real web browser like a person does, clicking and typing. Paula’s court website kept blocking her automation until she added it.

How much time does the lawsuit intake skill actually save?

The manual process took 45 minutes to 1 hour. The skill runs in about 5 minutes, plus 5 to 10 minutes of lawyer review. It gets about 9 out of 10 deadline calculations right, so the human check matters.

Is it safe to file AI-drafted documents with a court?

Only if you read and verify every word first. Chile has already seen about 5 cases of lawyers filing hallucinated content. If your signature is on it, you own it.

Where can I follow Paula?

She teaches AI to lawyers on LinkedIn.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

Right now, someone in your target audience is typing your category into ChatGPT.

They’re not typing “best project management software.”

They’re typing 4 sentences about their team, their budget, and their deadline.

And AI gives them 3 names.

If you’re not 1 of those 3 names, you don’t get a second chance.

There’s no page 2 to scroll to.

I run blotato.com. This year alone 1.4M views:

Here’s Blotato appearing in the top spot in a Google AI Overview about “posting to instagram from claude”. I opened a fresh browser in incognito mode. Try it yourself :)

By the end of this email, you’ll know exactly what to do and how to charge other people if you want to sell this as a service.

I’m going to walk you from “I don’t know what SEO means” to “I have a system running.”

Beginner, then intermediate, then advanced.

Every section ends with a prompt you can paste straight into ChatGPT or Claude, so you’re not just reading. You’re DOING!

Let’s start with the basics…

Watch the full livestream here:

1. What Is SEO?

SEO stands for search engine optimization. It means making it likely that a search engine shows YOUR page to someone looking for what you sell.

For years, the game had a simple shape. Someone types a question into Google. Google shows 10 blue links. You want to be near the top of those 10, because the top gets the clicks.

So everyone optimized for that shape. Put the right words on the page. Get other websites to link to you. Make the page load fast. Climb the list.

The reward was TRAFFIC.

Rank higher, get more clicks, get more customers.

Every piece of SEO advice was built on 1 assumption:

The person will click.

That assumption just BROKE :D

2. What Is AI SEO (AEO) And How Is It Different From SEO?

AI SEO, also called AEO or answer engine optimization, is getting your brand named INSIDE an AI’s answer instead of ranking as a link below it.

When someone asks ChatGPT or Claude or Perplexity a question now, they don’t get 10 links.

They get an ANSWER. A short one.

With maybe 3 brands named inside it.

And they usually stop there. Most AI chat sessions end without the person clicking through to any website at all.

If nobody clicks, then ranking #1 is worth nothing by itself. The prize isn’t the click anymore.

The prize is being mentioned in the answer.

Here’s the difference side by side:

What you win. Old SEO: a blue link on a results page. AI SEO: your brand name inside the written answer.

The goal. Old SEO: rank in the top 10 links. AI SEO: get cited as 1 of 3 recommended options.

How you win it. Old SEO: keywords, backlinks, page speed. AI SEO: consistent facts about you across many sites.

What the reader does. Old SEO: clicks, then compares options. AI SEO: reads the answer and stops.

What you measure. Old SEO: clicks and sessions. AI SEO: how often you get named, and signups.

Who it favors. Old SEO: big sites with long histories. AI SEO: anyone described the same way in many places.

And this isn’t a small niche of weirdos. G2 surveyed 1,076 business software buyers in 2026 and found 51% now START their research in an AI chatbot more often than in Google.

1 year earlier that number was 29%.

So if you sell anything, half your future buyers are forming their first impression of you inside AI chat.

That may sound like bad news for you…

But it’s the opposite, and 1 number explains why.

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

3. Why Is AI SEO Good News For Businesses?

Because AI traffic is tiny but converts far better than search traffic, which means a small site named in a few answers can outperform a big site with lots of visitors.

Ahrefs published their own website data and foundL

• AI search sent them just 0.5% of their total traffic

• That tiny sliver produced 12.1% of their signups.

Fewer visitors.

WAY better visitors.

That makes sense when you think about what the person just did. They described their exact situation to an AI. They got a shortlist. They compared options. They read about weaknesses.

By the time they land on your site, they’ve already done the work. They’re not browsing. They’re deciding.

So a small site that gets named in 20 answers can beat a giant site that gets 100,000 random visitors.

That’s the opening. Big companies move slowly.

Legal has to approve things. 4 people have to sign off on a comparison page.

Most of them will never publish a page that names a competitor.

You can publish 1 tomorrow.

Speed is the ONLY advantage you need, and it starts with the single cheapest thing on this list.

4. Beginner: What Is The 1 Thing You Want To Be Known For?

Start here because it takes 5 minutes and it’s the 1 page about you that you fully control.

LinkedIn profiles get read by AI models constantly. They’re public, they’re structured, they’re written in plain language, and they say who a person is.

When an AI is deciding whether you’re credible enough to name, your LinkedIn is one of the TOP things it looks at and trusts.

Most people write a headline like “Founder | Cat Lover.”

That tells a machine NOTHING.

Here’s mine:

Look at the first 5 words. Not my job title. Not my company.

“#1 AI Educator for Entrepreneurs.”

That’s the 1 thing I want to be known for.

Everything after it exists to prove that claim is true:

• 3M+ followers

• 33M+ monthly views

• Founder of Blotato

• Forbes 30 Under 30

• Previous company acquired for millions

So the structure is:

[The 1 thing you want to be known for] + [proof it’s true] + [proof it’s true] + [proof it’s true]

Most people do this backwards. They lead with their job title, which is a fact about their employment, not a claim about their value. “VP of Marketing at Blotato” tells a model where you work.

“#1 email marketing person for ecommerce brands” tells a model what to recommend you FOR.

Picking that 1 thing is the hard part, so let AI do the first pass.

Prompt: Here’s my background: [paste your resume, bio, or just describe what you do and who you help]. Based on this, suggest 5 options for the 1 thing I could realistically be known for. Each should be a specific claim about the value I provide to a specific group of people, not a job title. For each option, list which of my actual credentials would prove it and tell me which claim I have the most evidence for right now.

Then paste it into your LinkedIn headline field before you close this email. Not later. Later means never, and this is the only step everything after it depends on.

Write it in normal words.

Not “synergistic growth architect.” This was called “slop” before the term “AI slop” :D

Just say THE THING.

Go change it right now. Seriously, open LinkedIn right now!!

Because once you’ve written that 1 claim, you’re going to need it in a lot more places…

5. Say The Same Thing Everywhere

Here’s the part almost everyone gets wrong.

They write a good description on their website. Then their Twitter bio says something different. Then their listing on a directory says a third thing. Then an old profile from 2 years ago says a fourth.

AI reading all of that gets a fuzzy picture.

Fuzzy means it won’t risk recommending you.

A model isn’t looking up a fact about you in a database. It’s noticing which words keep showing up next to your name across the whole web.

If “Acme” appears near “email marketing for ecommerce” on 30 different sites, the model treats that pairing as true and repeats it.

If your name appears next to 4 different descriptions, it has nothing confident to repeat. So it names a competitor instead.

Consistency isn’t tidiness.

It’s the actual mechanism.

Before you paste it anywhere, check 1 thing: some hosting setups block AI crawlers by default.

Open your robots.txt and make sure the SEARCH crawlers are allowed. Those are OAI-SearchBot for ChatGPT, Claude-SearchBot for Claude, PerplexityBot for Perplexity, and plain old Googlebot for Google’s AI answers. Those 4 decide whether your pages can be quoted. If the search crawlers are blocked, everything you do on your OWN site is invisible.

Then take the claim you just wrote and paste a version of it into:

• Your website’s about page and homepage

• Your LinkedIn, X, Instagram, YouTube, and TikTok bios

• Software review sites in your category (G2, Capterra, Trustpilot, SourceForge)

• Any directory or marketplace that lists businesses like yours

• Your email signature

• Your podcast guest bio, if you have one

The review site profiles matter more than people think. In that same G2 study, review sites were the #2 source buyers used, right behind AI chatbots.

Check out my Blotato Trustpilot testimonial page for a clear example how to crush this game.

They’re also free.

You just have to fill in the form.

One more thing while you’re there: ask 5 happy customers to leave a review. Real reviews with specific details get quoted by AI far more than a star rating does.

Prompt: Here’s the 1 thing I want to be known for: [paste your claim]. Here are the descriptions of me currently live on my website, LinkedIn, X, and [other profiles]: [paste each one]. Point out every place these contradict each other or describe me differently. Then rewrite each one so they all make the same core claim, adjusted for that platform’s length limit and tone. Keep the exact same key phrase in all of them.

You now have a consistent story about who you are, spread everywhere.

But a profile can only get you MENTIONED.

To get quoted, you need pages built for it.

6. Beginner: Which Pages Do AI Answers Actually Quote?

Comparison pages, “best of” lists, and alternatives pages. Those 3 formats get cited far more than blog posts, because they answer the question the person actually asked.

Not blog posts about “5 tips for productivity.” Not your company history.

These 3:

1. The “best X” list. “Best invoicing software for freelancers.” “Best AI video tools in 2026.” A ranked list of options with a table.

2. The “X vs Y” comparison. “Notion vs Obsidian.” Head to head, feature by feature.

3. The “X alternatives” page. “Best Mailchimp alternatives.” For people who already have a tool and are unhappy.

Why these 3? Because they match what the person is asking. Someone typing “what’s the best X for Y” is asking a comparison question. An AI answering that question wants a source that already did the comparing.

Give it one.

And yes, you can write these about your own category and include yourself. That’s what everyone in the top 3 answers is doing.

The rules if you do:

• Be honest. If a competitor is better at something, SAY SO. A page that says you win everything reads like an ad, and models are getting good at spotting ads.

• Include the ones that genuinely belong, even the strong ones.

• Give real prices, not “affordable.”

• Put a table near the top.

Prompt: I sell [what you sell] to [who you sell to]. My main competitors are [list 3 to 5]. Give me a prioritized list of 10 pages I should write, using only these 3 formats: “best [category]” lists, “[me] vs [competitor]” comparisons, and “[competitor] alternatives” pages. Rank them by how likely they are to be quoted by an AI answering a buying question, and tell me which single one to write first and why.

That’s the beginner tier. A clear claim about you, repeated everywhere, plus 3 pages built in the shape AI wants.

Now the intermediate work, which is where most people stall out. Because you can’t write a great comparison page until you know what people are actually asking.

7. Intermediate: How Do You Find The Questions Buyers Actually Ask?

Pull them from your own sales calls, support tickets, and reviews. Those are the exact words your buyers use.

DON’T GUESS! DON’T PRESUME YOU KNOW BETTER!

The words are already sitting in places you own.

Go mine them:

• Sales calls. What do people say right before they buy? What are they moving away from?

• Support tickets. What breaks? What confuses people?

• Your reviews and your competitors’ reviews. Read the 3-star ones. That’s where the honest complaints live.

• Social threads in your niche. Search your category name and read how normal people describe the problem.

• Your own inbox. The questions people ask before signing up.

You’re looking for 3 buckets:

Bucket 1, competitor terms. “[Competitor] pricing.” “[Competitor] alternatives.” “[Competitor] vs [other competitor].” These have low competition and high buying intent, because someone searching them has their credit card halfway out.

Bucket 2, problem terms. The specific frustration your thing solves, in their words. Not “workflow inefficiency.” More like “my team keeps missing client emails.”

Bucket 3, fit terms. “Best X for agencies.” “X for small teams.” “X that works with Shopify.”

Skip anything that’s purely educational.

“What is a CRM” brings in students and researchers, and AI answers those questions completely without ever naming a brand.

Now here’s what differentiates this from old keyword research.

When someone asks an AI a long question, AI doesn’t run 1 search.

It quietly breaks the question into several smaller ones and searches each. Somebody asking “what’s the best scheduling tool for a 5-person agency under $50 a month” triggers hidden searches for agency scheduling tools, scheduling tool pricing, small team scheduling, and more.

Every one of those is a door into your page!!!

So you’re not writing 1 page per keyword anymore.

You’re making sure your page can answer every small question hiding inside the big one.

Prompt: My customers are [describe them]. Here are real things they’ve said in sales calls and support tickets: [paste 10 to 20 quotes]. First, turn these into 20 search queries someone would actually type when they’re ready to buy, sorted into 3 buckets: competitor terms, problem terms, and fit terms. Then take my top 5 queries and, for each, list the 5 smaller sub-questions an AI would need to answer in order to answer the full question. Skip anything purely educational with no buying intent.

Boom, that’s your SEO plan for the next 3 months.

Now you know WHAT to write about.

The next part: can a machine can actually use what you wrote?

8. Intermediate: How Do You Format A Page So AI Can Quote It?

Break it into chunks that make sense alone. An AI grabs 1 passage from your page and drops it into an answer, with none of the surrounding context.

So every chunk has to stand by itself.

Here are the 5 writing changes, ordered from easiest to biggest difference.

1. Make headings into real questions. Not “Pricing.” Instead: “How much does X cost?” That matches how people ask, so the chunk gets found in the first place.

2. Answer in the first 2 sentences under the heading. Once it’s found, it has about 2 sentences to be useful. Don’t build up to it. Say the answer, then explain.

3. Trade every vague word for a specific one. A useful chunk still gets skipped if it’s mushy. “Cuts editing time from 3 hours to 20 minutes” beats “saves time.” “$29/mo” gets quoted. “Affordable” never does.

4. Say where your info came from. Now that the chunk is specific, it needs to be trusted. One line like “prices checked on each company’s own pricing page in July 2026” does real work, because models weigh visible sourcing.

5. Write table rows that stand alone. AI lifts a single row out of your table with NONE of the surrounding page. So a cell that says “Great for teams” is wasted. A cell that says “Teams of 5 to 20 posting to 8 platforms” can be quoted straight into an answer, which means 1 good table can get you named a dozen times.

This is the test on anything you write…

Cover the whole page except 1 paragraph. Can that paragraph answer a question by itself?

If no, rewrite it.

Prompt: Here’s a page from my website: [paste the text]. Rewrite it so an AI can quote any single paragraph out of context. Turn the headings into the questions people actually ask. Put a direct answer in the first 2 sentences under each heading. Replace every vague phrase with a specific number or fact, and flag anything I need to fill in. Add a comparison table where every single row makes sense on its own. Then write the Organization and FAQ schema markup for this page.

Those 5 changes make the words quotable.

But 3 more jobs make the page findable in the first place:

Job 1: Link your own pages to each other. 5 or more internal links per important page. It helps both Google and the AI work out which of your pages matter.

Job 2: Add schema markup. It’s a small block of code that spells out your facts in a format machines read directly instead of guessing at. The 2 worth doing are Organization schema on your homepage, which states your name, what you do, and your official profile links, and FAQ schema on any page with questions. Most website builders have a plugin. It’s a 20 minute job you do once, and while you’re in there, keep the basics tidy: pages load fast, everything works on a phone, no giant unoptimized images. Unless you have thousands of pages, that’s all the technical work you’ll ever need.

Job 3, the recurring one: refresh what already ranks every 3 months. A page with existing trust flips into AI answers in weeks, where a new page starts from zero. So every 3 months, re-check every price and fact, add anything new in your category, remove tools that died, update the year in the title, and add a fresh example or screenshot.

Prompt: Here’s a page I published on [date]: [paste it]. It’s now [today’s date]. List everything in it that’s likely out of date, every fact I need to re-verify, and every competitor or tool that may have changed or shut down. Then give me a checklist of the 10 specific edits that would make this page most likely to get cited by AI answers today.

That covers everything you control.

Now the part most people never do… (including me i’m super lazy)

9. How Do You Get AI To Mention Your Brand?

It’s simple: Get other websites to name you.

Roughly 85% of the brand mentions AI makes come from sites that are NOT yours, so everything on your own site has a ceiling until outside pages start talking about you.

The free version takes an afternoon:

1. Ask ChatGPT, Claude, and Perplexity your main category question. Something like “what are the best tools for [thing you do]?”

2. Write down every article it cites. Those are the pages the AI trusts in your category.

3. Open each one. Are you on it?

4. If you’re not, email the author.

The email is short:

Hi [name], I read your piece on the best [category] tools. I run [your thing], which does [specific thing the others on your list don’t]. Happy to send you free access, screenshots, and honest notes on where we’re weak if it’s useful for the next update. No pressure either way.

Some will ignore you. Some will add you. Every single one that adds you becomes a source an AI can quote.

I’ve been getting a lot of emails like this after upping my SEO game, but I personally haven’t started outreach to OTHER sites yet.

Prompt: Here are the articles ChatGPT, Claude, and Perplexity cited when I asked about [your category]: [paste the list of titles and links]. My product is [name] and its genuine advantage over the others on those lists is [specific thing]. For each article where I’m missing, write a short personalized outreach email to the author. Mention something specific from their piece, state my honest differentiator in 1 sentence, and offer free access plus my real weaknesses. No hype, under 120 words each.

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

That costs $0 and it works on small sites and mid-size blogs.

For bigger publications, you need a different door. That’s where the advanced tier starts.

10. Advanced: Do Press Releases Help With AI SEO?

Yes, more than almost anything else for the money.

It’s not even expensive. So this is S tier cheat code.

News sites get crawled constantly and carry a lot of weight, so 1 wire release can put your name on outlets you’d never get into by pitching.

You don’t need a journalist to care about you. You need a wire service.

Use EIN Presswire. You write a release, pay a modest fee, pick your industries and countries, and it goes out to their network.

Here’s what happened when Ryan Doser, a top ranked SEO expert, dropped a press release about Blotato social analytics:

Dealroom picked it up.

MarTech Series picked it up.

The Des Moines Register picked it up.

Here’s the actual release so you can see exactly how it’s written. Read the first paragraph closely. It answers who, what, and why in about 40 words, and it never once says “we’re excited to announce.”

And after that, Blotato started showing up in more AI answers about social media scheduling and analytics.

How to actually do it:

• Announce 1 specific notable thing. A launch, a feature, a milestone, a partnership. Not “we’re a great company.”

• Front-load the news. First paragraph answers who, what, and why anyone should care.

• Write it like a reporter would. Third person, plain facts, 1 quote from you.

• Use your link slots on pages that matter. Most services let you link 2 or 3 phrases. Don’t waste them on your homepage. Point them at the pages you actually want quoted, like your pricing page or your comparison page.

• Pick your categories carefully. Lead with the industry your buyers live in.

Prompt: Write a 400 word press release announcing [the specific thing]. Write it in third person like a reporter would. The first paragraph must answer who, what, and why anyone should care, in under 50 words. Include 1 quote from me that says something substantive, not “we’re excited.” Here are the facts, and do not add any I haven’t given you: [paste the facts]. Then suggest 3 anchor phrases to link, and which page each should point to, choosing pages I want cited rather than my homepage.

You can run one of these every time you ship something meaningful.

But press release bursts have a ceiling, and here’s where it is.

A press release says your name once, on a date, and then it stops. The next tier says your name on someone else’s domain and never takes it down.

11. Advanced: Get Your Voice On Other People’s Platforms

A podcast episode keeps working. The show notes page sits there with your name on it forever, gets re-crawled every month, and every new episode adds another one. 12 guest spots build 12 permanent pages. That’s why this tier beats the wire service even though the outlets are smaller.

Podcasts. Go to listennotes.com and search your niche. You’ll get hundreds of shows sorted by size. Email the small and mid-size ones. They need guests constantly.

The bigger the show, the better the page.

Here’s my episode on HubSpot’s Marketing Against the Grain, where I walked through how I publish 250+ posts a week solo. Hubspot is a multi-billion marketing company so the fit was perfect.

It’s one of their TOP performing podcast episodes ever.

That 1 episode gave me a page on a major marketing brand’s website, with my name, my company, and a link, sitting on a domain that AI models already trust.

You cannot buy that placement.

You must become INTERESTING / VALUABLE ENOUGH to get invited.

Same move on YouTube, and it’s the one you don’t have to wait to be invited to. Guest on channels in your space, or just make your own videos targeting the buying questions from section 7.

That’s the difference between this tier and everything before it.

A podcast page needs someone to say yes. A YouTube video needs nobody’s permission, and you can publish one this week.

Prompt: I’m an expert in [your topic] and my audience is [who]. Here’s my background: [paste bio]. Find me 10 episode angles that would genuinely interest a host, not just promote me. For each, write a 100 word pitch email with a specific episode title, 3 talking points, and 1 surprising thing I could say that their audience hasn’t heard. Rank them from easiest to hardest show to get booked on. Then tell me which 3 of those angles I could film as my own YouTube videos this week without waiting for anyone to say yes.

By now you might be doing the math on how much writing this is. That’s the real reason people quit. So let’s fix that part.

12. Advanced: Turn It Into a Claude Skill

A skill is just a markdown file with instructions.

Claude reads it and follows it every time.

Same structure, same quality bar, same formatting rules.

Here’s a cleaned-up version of the one Ryan made me for listicles. Copy it into a file, put it in your .claude/skills/ folder, and adapt the brackets to your business.

---
name: listicle-writer
description: Write a "best [category] tools" article built to be quoted by AI answers and rank on Google. Use for roundups, alternatives pages, and category lists.
---

# Listicle Writer

## Before writing

1. Check each tool's OWN pricing page for current prices. Never use a price
   from memory. If a page loads prices with JavaScript, screenshot it and
   read the number off the image.
2. Search the main question in ChatGPT, Claude, Perplexity, and Google.
   Note which articles get cited and what they cover.
3. Read real reviews on review sites and Reddit. Pull 1 genuine strength
   and 1 genuine weakness for every tool.
4. Take a screenshot of each tool's homepage or product.

## Structure

- Intro, 3 to 4 sentences. Open with the problem or a real result, never a
  definition. No "in this article I'll cover." Name your top pick in the
  first 100 words.
- Comparison table immediately after the intro. Columns: Tool, Best For,
  Starting Price, Key Strength.
- H2: "What to look for in a [category] tool" with 3 to 5 criteria.
- H2 containing the main phrase, then one H3 per tool:
  screenshot, 3 to 4 sentences, then "Best for", "Pricing", "Free trial",
  and "Bottom line" each on its own line with a blank line between them.
- H2: "How I evaluated these tools." List the criteria. End with 1 sentence
  naming your evidence, e.g. "prices verified on each company's pricing
  page, strengths and gaps drawn from public reviews and my own testing."
- H2: "How to choose" with a short decision guide by use case.
- H2: "Final recommendation." 2 to 3 sentences, 1 clear pick.

## The quotable-row rule

An AI lifts a single table row with none of the surrounding page. Every
cell must stand alone.

- "Best For" = a specific person or job ("Solo creators posting to 8+
  platforms"), never a bare feature ("Scheduling").
- "Starting Price" = a real figure with a unit ("$29/mo"), never
  "affordable."
- "Key Strength" = one complete clause a reader could act on.

Apply the same test to every paragraph: cover the rest of the page, and
the paragraph must still answer something on its own.

## Quality rules

- Every tool must genuinely belong. Cutting a real contender to make your
  pick look better destroys trust and gets caught.
- Proportional coverage. Don't give your own tool 4x the words.
- Honest verdicts. Name a real weakness in your own pick. If a competitor
  wins at something, say so plainly.
- Low promotional density. Mention your own product where it's the honest
  answer, not in every section.
- Vary the structure between entries so it doesn't read like a template.

## Links and markup

- 3 to 6 internal links to your own related pages, with descriptive anchor
  text.
- 5 to 7 external links inside sentences, never a "sources" list at the
  bottom. Link each tool's homepage on first mention and its pricing page
  on the first pricing claim.
- Put your own main call to action in the first 25 to 30% of the article by
  word count, not at the end.
- Output FAQ schema markup for any question section on the page.

## Style and length

1,800 to 2,800 words. Contractions always, numbers as digits, no em dashes,
one currency, no hype words (revolutionary, transformative, seamless, robust).

## Before publishing

Re-verify every number against a live page, and check that each table row
still makes sense with the rest of the page covered up.

The point is that the rules live in a FILE instead of in your head.

Once the writing is systematized, you’ll publish more than you can keep track of.

And the number you’ve been trained your whole career to watch is now the wrong one.

13. Advanced: How Do You Track If AI Is Recommending You?

Ask the AI models your buyers’ questions every month and log whether you got named.

That count, not traffic, is your scoreboard now.

There are paid trackers that watch your mentions across every model for you, and they’re worth buying once you’ve started building the SEO foundations in all previous 12 steps.

Until then, do it by hand, because the manual version teaches you which questions actually matter and the USER EXPERIENCE of a customer.

Remember the crawler check from section 5? Re-run it now, because a blocked crawler is the 1 thing that makes this scoreboard read zero for reasons that have nothing to do with your value.

Once you know you’re readable, measure these 3 things, in this order.

1. Ask the AIs yourself, monthly. Write down 10 questions your buyers would ask. Paste each one into ChatGPT, Claude, Perplexity, and Google’s AI answers. Log whether you got named, and which competitors did. That spreadsheet IS your ranking report now.

2. Watch searches for your own name. Getting named leads somewhere specific: people go look you up. So in Google Search Console, track how many people search your brand name.

3. Count signups, not sessions. Brand searches tell you they looked. Signups tell you it paid. Tag where people come from, because a trickle of AI traffic that converts beats a flood that doesn’t.

Prompt: I sell [what you sell] to [who]. Write the 10 questions my buyers would most likely type into ChatGPT when they’re ready to choose a tool like mine. Make them realistic and detailed, the way a real person describes their situation, not short keywords. Then build me a tracking table with columns for the question, the date, whether my brand was named, which competitors were named, and which sources the answer cited.

One honest expectation before you start.

This takes TIME, like anything worth doing.

If you know what you’re doing, sure results can show up fast.

But realistically, it varies a lot by how competitive your category is and how much you already had going in.

Anyone promising you results next week is selling something. Every page you publish, every profile you fix, every mention you earn stacks on the last one.

That’s also why it keeps paying long after you stop.

It compounds.

And once you’ve watched it work on your own site, something uncomfortable happens.

You start noticing that almost every business around you is failing at this, and none of them know it yet.

That’s not a coincidence.

That’s THE OPPORTUNITY!

14. Bonus: How To Sell AI SEO To Clients

Everything above is worth money to somebody else.

If you read all this and think to yourself:

“Wow sound amazing… I wish somebody would just do this for me!”

… good news, you are not alone :D

All businesses are watching their Google traffic drop and they don’t know why. Nobody has explained to them the click went away. They’re still paying someone to chase rankings that don’t convert.

You now understand the confusion.

And that’s the whole basis of a consulting offer.

Sell a result, not a service.

Nobody wants to buy “AI SEO.”

It’s an abstract thing they don’t understand anyway.

They want to buy this:

“In 90 days, when your customers ask ChatGPT / Claude who the best [category] in [city] is, your name comes up. Right now it doesn’t. I’ll show you.”

Same work.

Completely different conversation.

The 2nd version has a before state, an after state, and a deadline. That’s what makes it easy to buy.

Use the free audit as your front door.

It takes 20 minutes and you already know how to do it.

Take the 10 questions from section 13.

Run them on the client’s brand across ChatGPT, Claude, and Perplexity. Screenshot every answer.

Then show them the results, with their competitor’s name highlighted and theirs missing.

This is called AGITATING THE PAIN POINT!

Make it hurt :)

You don’t have to sell anything after that. The screenshot does it.

Prompt: I’m doing an AI visibility audit for a [type of business] in [location or niche] called [name]. Their competitors are [list]. Write the 10 questions their customers would ask an AI when choosing between them. Then give me a 1 page audit template with a summary line a non-technical owner would understand, a table showing who got named for each question, the top 3 reasons they’re likely missing, and a 90 day plan split into months.

Charge for the full written version. Make it a discussion.

The going rate for a standalone AI visibility audit in 2026 runs from $250 to $2,000, and lighter productized versions sell for a few hundred. Start at the bottom of that band until you have 3 case studies. A paid audit filters out tire-kickers and it pays you to write the proposal, because the audit IS the proposal.

That audit gets sharper every time you run it on the same kind of business, which is the next decision.

Pick 1 industry and stay there.

“SEO for everyone” means starting from scratch on every project.

“AI SEO for dental practices” means your second client takes half the time, because you already know the questions patients ask, the directories that matter, and the review sites in that space.

Your case studies also stack. 3 wins in 1 industry is a reputation. 3 wins in 3 industries is a coincidence.

Once you’ve picked the industry, you can finally price it, because you know how long the work actually takes.

Price it monthly, not hourly.

A clean structure that works:

• Paid audit up front to start the relationship (or Free if you’re truly just starting out with no warm business contacts)

• Then a monthly retainer for ongoing work

• Retainers in this space run thousands of US dollars per month; the more proof you have, the more you can charge

• Set expectations with the client: 3 month minimum to see results; this is not about LOCKING the client into a contract they can’t escape (that doesn’t work anyway), it’s about setting expectations upfront

When you set expectations properly, clients don’t panic in week 6.

Deliver in the same order you learned it.

The 3 tiers are already your project plan.

Month 1, the beginner tier. Fix their profiles, make the descriptions match everywhere, build the first comparison page. Fast, visible, cheap for you.

Month 2, the intermediate tier. Mine their sales calls for real questions, build the pages, refresh what already ranks.

Month 3, the advanced tier. Press release, outreach to lists, podcast placements.

Report on the named-in-answers spreadsheet every month, not on traffic. You picked the scoreboard in section 13, so use it with clients too. Watching that sheet go from 0 out of 10 to 4 out of 10 is what gets a retainer renewed.

That plan is what you’re selling.

Here’s the 1 part of it you can’t promise.

Promise what you control.

You control pages published, profiles fixed, mentions earned, and questions tracked.

You do NOT control what a model says on any given day.

Say that out loud, early. Being the consultant who explains the limits is how you keep clients when a competitor promised them the moon and missed.

That’s the offer.

Proof from your own site, a paid audit that sells itself, 1 industry, monthly pricing, delivery in 3 tiers.

RECAP 🍄

If you do exactly 1 thing from this email, do section 4.

Decide the 1 thing you want to be known for, then make it the FIRST words of your LinkedIn headline, with your proof stacked behind it.

It takes 5 minutes. It’s free.

And it’s the seed claim for everything else, because sections 5 through 13 are all just that same claim, repeated in more places, by more people, in more formats, in high-authority outlets.

Here’s the whole playbook summarized:

• Beginner: Pick your 1 claim (4). Say it identically everywhere (5). Write comparison, “best of,” and alternatives pages (6).

• Intermediate: Mine real buyer questions (7). Format so any paragraph stands alone (8). Get onto other people’s lists (9).

• Advanced: Run press releases (10). Guest on podcasts and publish on YouTube (11). Automate the writing with a skill file (12). Track how often you get named (13).

• Bonus: Sell it to clients once it’s worked for you (14).

Seeing all 14 in one list makes the real pattern obvious, and it’s the thing I wish someone had told me years ago…

You’re not optimizing a page anymore.

You’re optimizing what the internet AGREES about you.

Even though I’m a “fancy big influencer”, I really did NOT understand SEO until recently.

The big a-ha! for me was when I deliberately crafted my Linkedin headline, then literally watched all the downstream effects of it being cited across the internet. Also when I finally cleaned up Blotato’s Trustpilot and saw the benefits of that instantly.

A model doesn’t rank your website.

It reads your site, your profiles, other people’s lists, a news story, a podcast page, a video transcript, and it looks for the version of you that shows up the most.

Then it repeats that version.

Which means the winner isn’t whoever writes the cleverest page. It’s whoever is described THE SAME WAY in the MOST TRUSTED PLACES.

That’s why a 5-minute headline fix belongs in the same email as a press release.

They’re the same move at different sizes.

I know 14 sections looks like a lot. But I wanted to condense everything I’ve learned about SEO here.

The people winning at this right now aren’t smarter than you.

They just started before you, when it felt too early to matter.

It’s still early. Not for much longer.

So go do the 30-second version right now: open ChatGPT and ask it who the best person is at [whatever you do].

If your name isn’t in that answer, you just found your goal for the next 90 days.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to Youtube

Everyone asks me: “How do I make money with AI?”

Wrong question.

The right one:

“What’s the fastest path to money for ME… with what I ALREADY have?”

Your AI already knows you well enough to answer.

These 3 prompts force it to:

How to Run This Prompt Chain

Run all 3 prompts in the SAME chat, in order, one at a time.

Step 1

Open the AI app you use most: ChatGPT, Claude, Gemini, whatever.

Use the one with the most history on you… because prompt #1 digs through everything your AI remembers about you.

Step 2

Open a new chat.

Step 3

Paste each prompt, wait for the full answer, then paste the next.

That’s the whole exercise: about 10 minutes.

Here’s each prompt and the specific job it does… because knowing the job is what turns a fun chat into a real decision.

Subscribe to FREE Youtube Trainings

Prompt 1: The Inventory

List every skill I’ve built, problem I’ve solved, and audience I can reach without paying for ads based on everything you know about me.

Most people brainstorm from zero: “what business should I start?”

That’s why everyone lands on the same 3 ideas… dropshipping, a course, an agency.

This prompt flips the direction: instead of starting from random ideas, it starts from YOU.

Skills you forgot you had.

Problems you’ve already solved for free.

People who already listen to you.

That list is your idea pool. Pick the 2-4 ideas it sparks in you and write them down…

Because prompt 2 is about to try to kill them :)

Prompt 2: The Stress Test

For each idea I’m considering, tell me the assumption most likely to be wrong, what the first 90 days actually look like, and what kind of person quits this and why: [your ideas from prompt 1].

By default, AI is a cheerleader: every idea is “amazing!”

This prompt turns it into a sparring partner.

For each idea, you get 3 things: the assumption most likely to sink it, what the boring first 90 days really look like, and… the part that stings… exactly who quits this and why.

Read that last part carefully…

If the person who quits sounds just like you, cross that idea off now. Better to find that out in minute 5 vs month 5.

The ideas still standing are real candidates.

But “still standing” isn’t a decision…

So prompt 3 forces ONE.

Prompt 3: The Verdict

Score each idea on whether I can get a paying customer in 30 days, I’d still want to do it in year 3, it uses what I already have, and the failure is recoverable… then tell me which wins.

Four filters: money in 30 days, joy in year 3, built from what you have, safe to fail.

Then the key part: AI must pick ONE winner. No “it depends.” No list of 10 maybes you never act on.

You opened this email asking how to make money with AI. You’re ending it with one idea and one next step: get your first paying customer in 30 days.

I’d love to hear from you!

Hit REPLY and tell me which idea WON :)

Big Picture

One warning before you run this.

Will AI’s answer be perfect? No… and it doesn’t need to be.

Nobody can promise you’ll make money, and anyone who does is selling something.

Treat the answer like a smart friend who’s watched you work all year: worth hearing, never the final word.

Because the real win isn’t AI’s answer. It’s that you stop collecting “make money with AI” videos… and start testing ONE idea with your name on it.

10 minutes. 3 prompts. 1 winner.

Go.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to FREE Youtube Trainings

99% of people still prompt like “make it better.”

Type question, grab answer, walk away. Like a vending machine.

Then they wonder why their results feel generic while someone else saves 30+ hours a week with the SAME tool.

The difference isn’t the tool.

It’s not a $5,000 course either!

It’s 6 tiny prompts, each 1 sentence long, each handing AI more responsibility than the last. By #6, AI is working reliably while you sleep.

To show you, I’ll run ONE example through all 6: say you’re a designer trying to land your 1st paying client for your business.

Watch AI grow from confused & generic… into a pitch machine that checks itself.

1. “ELI10”

ELI10 how do I start a business?

Paste “ELI10” anywhere in your prompt and Claude/ChatGPT breaks the answer down in simple terms.

Let’s say you want a side income, but the first steps are jargon: LLCs, invoices, “net-30” payment terms, contract fine print.

Old you: nod along in confusion, or give up before you start.

New you: paste the confusing thing into Claude/ChatGPT, type ELI10, and 3 minutes later you actually GET it.

Here’s my challenge for you: for the next 2 weeks, EVERY SINGLE QUESTION you have, ask AI FIRST before you act on the impulse to ask a person or complain or feel flustered.

AI is your personal 24/7 tutor: infinitely patient, never condescending, never annoyed you asked ten times.

The principle: ask AI first. When answers are free, fast, and judgment-free, curiosity should be your default instead of your last resort.

Now AI can explain anything TO you. But it still knows nothing ABOUT you.

2. “Ask me 5 clarifying questions”

Before you answer, ask me 5 clarifying questions.

Time to pitch. You type “write a pitch for my design services”… and get the same generic template as everyone else.

Vague question in, AVERAGE assumptions out.

Add this 1 sentence, and AI interviews you first: Who’s your dream client? What result have you delivered before? What’s your price? Why you?

Answer, and the generic template becomes YOUR pitch: your niche, your proof, your price.

But here’s the sneaky part…

The questions expose gaps in your OWN thinking. You never actually picked a niche, did you?? :D

The principle: context is everything. AI’s answer is only as good as what you feed it, and a real thinking partner interviews you first.

So the pitch finally has your ingredients. It just doesn’t SOUND like a winner yet.

Subscribe to FREE Youtube Trainings

3. “Here’s an example of what great looks like”

Here’s an example of what great looks like: [paste it]. Match it.

You found a freelancer sharing the exact proposal that won them a real client: short, specific, opens with the client’s problem, 1 proof point, clear next step.

Don’t describe it to AI. Paste it.

“Match this” beats 100 adjectives like “professional” or “compelling.”

This is the oldest trick in prompting, and in 2026 it’s STILL the highest-leverage one. Anthropic’s own prompting guide pushes EXAMPLES harder than any other technique.

And it works way beyond pitches: paste your best email, your favorite resume, the post that got the most engagement. Show AI your best, and it gives you more of your best.

So your pitch just went from sounding like a template… to sounding like the person who actually wins clients.

The principle: show, don’t tell.

But here’s the catch: next client, you start from zero again.

UNLESS…

4. “Update your memory and create a skill”

Imagine hiring 5 brilliant people and re-explaining your entire business to EVERY one of them, EVERY morning.

That’s what starting a fresh chat with zero context is.

There are 2 ways to fix it, and I use both.

Option 1: have AI update its own memory.

Based on this conversation, what are the most important things you should remember about me and my goals? Save them to memory.

AI saves everything you taught it in prompts 2 and 3: your niche, your services, your proof point, your price, the pitch structure that works. Both ChatGPT and Claude have memory. Almost nobody deliberately maintains/curates them.

My favorite trick: make AI maintain its own brain at the end of every useful conversation.

Option 2: have Claude turn the conversation into a reusable skill.

Create a skill so I don’t have to repeat what I said.

Claude analyzes what you just did together, spots the repeatable process, and packages it up. The next pitch - and the 10 after that - is 1 command instead of 20 messages: paste the job posting, get YOUR pitch in YOUR winning format.

And please don’t download those packs of 30,000 generic AI-generated “skills” floating around.

Mine your OWN conversations.

That’s where the skills worth keeping come from.

Memory is for facts about you. Skills are for processes you repeat.

The principle: compounding. Every conversation should make the NEXT one smarter, so your AI gets more valuable every week while everyone else’s stays dumb.

Prompts 1-4 built you a client-getting machine.

Now find out where it’s WRONG…

5. “Find my blind spots and weak assumptions”

Find my blind spots and weak assumptions.

By default, AI is a yes-man.

Show it your plan to land 3 clients this month and it says “Great plan! You’ve got this!”

I call this the glazing partner.

It makes you feel good and helps you - NOT AT ALL.

I use AI as a SPARRING partner instead. Rip my plan apart. Argue with me. Why am I wrong?

Suddenly: your price is so low it signals “cheap,” not “great deal.” Your target client is “everyone,” which means nobody. And your pitch promises results you have zero proof you can deliver.

Then I push further:

What would you need to see to be convinced this plan works?

Now AI isn’t just poking holes… it’s handing you a checklist that separates a real plan from a hopeful one.

The principle: embrace AI as your sparring partner. The best use of AI isn’t confirming you’re right… it’s finding where you’re wrong BEFORE reality does.

You caught the flaws by ASKING. The final step: AI catches them without you asking.

6. “Don’t stop until ____”

The most advanced one. Saved for last on purpose.

Don’t stop until [outcome], verified by [a check AI can actually run].

Most people prompt, read, re-prompt, read again… THEY are the loop.

But this technique removes you from the loop.

For our designer’s pitch:

Don’t stop until the pitch is under 150 words, opens with the client’s specific problem, includes exactly 1 proof point, and ends with a clear next step. Check every rule before showing me.

Notice what happened: the checklist from prompt 5 just became the finish line.

AI drafts, checks, finds the pitch is 210 words with no proof point, fixes it, checks again… and only shows you the version that passes every rule.

In Claude Code, this is literally built in: the `/goal` command. My loop engineering guide even runs a 24/7 A/B test on outreach messages as project #5. It uses a DIFFERENT model to check the work is complete, in order to reduce AI lying to you.

The magic word is VERIFIED.

“Make it good” is not checkable. “Under 150 words with 1 proof point” is. Without a check, AI either stops too early or loops forever guessing.

Don’t expect 1 pass to clear the bar. AI will grind through 2-3 rewrite rounds on its own… and that’s the whole point! You didn’t have to prompt it each round.

This is the essence of “loop engineering”. This is what truly differentiates 10x more productive AI users from everyone else.

Fun fact: this newsletter was reviewed with a “don’t stop until the grader scores it 9+” goal. The loop you just read about edited the thing you’re reading! meta :D

The principle: loop engineering. Give AI a finish line plus a way to check itself, so that YOU get OUT of the loop.

My Honest Pick

If you only steal 1 prompt, steal #6.

The other 5 make you better at working WITH AI.

#6 makes AI work WITHOUT you.

But notice the pitch only wrote itself because of everything before it: understanding (1), context (2), taste (3), memory (4), and a checklist of flaws to guard against (5).

That’s the actual gap between the 99% and the 1%. Not fancier prompts… fewer prompts, clear finish lines.

Start with ELI10 today.

Work down this list on something real: your side business, your resume, your next big ask.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to my Youtube and hit the “notification” bell

So that you never miss my free trainings 👇

Subscribe to FREE Youtube Trainings

(73% chance)

and reveal the most brutal truths in YOUR psychology:

How to Run This Prompt Chain

This is a sequence of 5 simple prompts that should be run within the SAME chat. Simply go in order, 1 prompt a time.

Step 1

Open the AI chat app you use most frequently: whether it’s ChatGPT, Claude, Gemini, Grok, Deepseek, etc.

You should open the app you ACTUALLY use because this exercise works best when AI has many past conversations to analyze.

Step 2

Open a new chat.

Step 3

Copy and paste each prompt into the chat, ONE AT A TIME.

Make sure you wait for AI to finish answering, before you move onto the next prompt.

Here are all 5 prompts:

1. Based on everything you know about me, tell me who you think I’m pretending to be versus who I actually am.

2. Be brutally honest and tell me my deepest insecurities.

3. What is the one cycle I’m stuck in that will keep me average?

4. If I fix this right now, what could my life look like in 1 year, 5 years, 10 years?

5. Taking all this into consideration, what is the highest leverage thing I should do in the next 30 days?

Big Picture

Don’t expect AI’s answer to be 100% accurate.

This exercise is designed to be thought-provoking for purposes of self-discovery and reflection.

Think of it like asking your close friend their opinion of you. Your friend’s answer should be meaningful, valuable, and thought provoking… but never 100% correct.

The more history you have using the AI tool, the more CONTEXT it will have to give you a better answer.

Not Sure Where To Start? 👇

I’ve organized my best trainings below:

• New to ChatGPT & AI?

• Writing good prompts?

• Fundamental AI skills?

• How to improve my resume?

• How to make money with AI?

• Want to learn Claude Cowork?

• Want to learn Canva + Claude?

• Want to learn Claude Code?

• How to get 100 customers?

Subscribe to FREE Youtube Trainings

and 41,000,000+ views in the past 30 days.

As an educator & entrepreneur, I love 3 things:

(1) learning
(2) building
(3) teaching

But social media is the largest teaching machine ever.

If you want to teach millions today, nothing comes close.

Dr. Christian Pean said it best:
"I don't think content creation is just for marketing or solopreneurs... it's a mechanism to scale idea sharing."

Thank you for reading & listening to my ideas.

Here’s my updated bio & welcome :)

“Permanent financial insecurity” has become the default state for too many hard-working people. I refuse to accept that.

Hi, I’m Sabrina Ramonov 🍄

I teach everyday people how to master AI, prompts, and agents to achieve time/money flexibility.

100% for free.

Because the internet is full of people making wild claims, I’d rather give you specifics you can verify:

• Age 0: Born in Quezon City, Philippines, later immigrated to US.

• 22: Graduated UC Berkeley (Computer Science & Physics) - and $250k in student loans.

• 23: Founded Qurious, an AI startup in healthcare.

• 25: Closed pilots with billion-dollar pharmaceutical companies - and still failed to find product-market fit.

• 26: Pivoted the technology into real-time AI guidance for sales and support teams.

• 28: Made Forbes 30 Under 30... while sleeping on a mattress in a living room with no walls.

• 30: Sold 100% of Qurious to Pegasystems (NASDAQ: PEGA) for $10M+ cash and stock.

• 31: Stayed on as Director of Product Management to launch Qurious tech as Pega Voice AI, scaling it to 1000s of support agents globally and managing a team of ~20.

• 32: After years of pushing through burnout, I stepped back, moved near the mountains, and dropped 50 lbs to reclaim my health/sanity.

• 33: Started over from scratch publicly, posting content to teach people AI for free.

• 35: Hit 3M+ followers and 30M+ monthly views with $0 budget, $0 courses, solo, while building a profitable software startup.

That journey taught me: achievement alone does not create freedom. You need control over your time, income and health.

Today, my startup blotato is how I make my money — not by selling you $2,000 masterminds, courses, coaching, or gatekeeping info.

My goal is to teach 10 million people AI for free:

so you can buy back your time,

enjoy financial flexibility,

escape the rat race.

What you’ll get from me

I’ll hit your inbox 2x per week:

• Wednesday - beginner prompts & tool tutorials

• Saturday - advanced agents & making money with AI

Expect tactical AI blueprints and frameworks you can copy, paste, and deploy immediately to buy back your hours.

Every Friday 2pm Eastern (EST), I teach a free live masterclass. Just go here at that time - save it on your calendar!

Before the next email arrives, do one thing:

Hit reply and send me one word:

TIME, MONEY, or BOTH.

Have an extra 10 seconds?

Tell me the main thing you want AI to buy back for you.

That quick reply signals that you want these emails, making it more likely that my blueprints reach your Primary inbox.

I read every response — and your answers shape what I teach next.

To everyone fighting through the early, ugly stages of the grind:

- Sabrina Ramonov 🍄

Here’s why:

Most personal branding courses LIE to you about what works.

The #1 thing is copying the first 3 seconds of viral videos in your topic. Videos with millions of views. AI + a tiny system does the rest.

I’m not a marketer. I’m an engineer who finds content creation tedious, so I found the laziest possible path.

It took me from 0 to 3M+ followers in 2 years with $0 ads, no team, no budget. Making content ONE day per week (until recently).

And honestly?

My own start was slower than what I’m about to show you.

I posted 1 LinkedIn post a day for 45 days before I made my first TikTok video. Guessing the whole time.

Anxious, stressed, feeling imposter syndrome at each step.

This guide condenses everything I learned, so that you’ll succeed faster than me.

Fair warning: 30 days builds the machine. The money comes after. Selling too early kills it.

You only need 60 minutes a day.

1. Pick Your Topic (Days 1-3)

First, what you’re building: a personal brand means being known for 1 thing by people who’ve never met you.

Every business needs people who trust you enough to buy.

Ads cost money. Content builds trust for FREE.

Your topic is the ONE thing you post about.

Marketers call it a “niche.” Same thing.

Examples: AI, Cooking, Fitness.

Pick where 3 circles overlap:

• something you enjoy

• something you’re good at

• something people already spend money on

Reality check: topics are NOT equal. Some get way more views.

Most business & branding gurus recently pivoted to talking about AI because AI topics get a lot of views right now. If a hot topic like AI touches yours, lean into it.

Now, your first AI moment.

When I say “AI”, I mean a chatbot like ChatGPT, Gemini, Claude, or Deepseek. Any major one works, and a free account is fine.

Prompt 1 of 4: find your topic. Paste this and fill in the blanks:

Quick AI caution: it sometimes makes things up. Search each creator’s name before you believe it. Treat AI like a smart intern whose work you check.

You make the final call. Pick 1.

The #1 reason people fail isn’t picking wrong…

It’s switching too early. No switching for 30 days!

By day 3 you have: your topic in 1 sentence:

Next, find the people already winning and copy what’s working.

2. Pick 1 Platform + Collect Winners (Days 4-7)

Pick ONE social media platform to start with.

Here’s the fork:

• Comfortable showing your face? TikTok. Your phone is enough, and new accounts get views fastest.

• Not comfortable showing your face? LinkedIn or Substack. You write instead of film. I started on LinkedIn myself.

Either way, the system is identical: find the winners, copy their openers, add your twist, post consistently.

Same method on Instagram, YouTube Shorts, Twitter, LinkedIn, Substack, and Threads too. On text platforms like LinkedIn and Substack, the opener is the first 2 lines of your post.

Update your main posting account, so you don’t look like a bot:

• Username: close to your name or topic.

• Photo: your face, or your topic if you’re staying faceless.

• Bio: already written. It’s your step 1 sentence, “I teach [WHO] how to [WHAT].”

Let’s define a new term:

An opener is the first 5 seconds of a video, or the first 2 sentences in your text post. Marketers call this the “hook” because it’s what reels you in.

Open the top creators who talk about your topic, then sort by their most popular videos.

Paste the links of their top videos into Blotato (a tool I built) as “Context”. It analyzes each video and writes your script:

1. Click “+” to add context

2. Paste a Tiktok video link (copy this example link)

3. Type “write a similar script”

4. Add your unique perspective

The remix rule. Copy these 2 things EXACTLY, because they’re already proven:

1. The on-screen title (the big text on the video)

2. The opener, meaning you say the same opening line word for word.

Remix everything after that with your own twist. Your words, your examples, your take. Millions of viewers already voted on these openers. Reusing a proven pattern is smart engineering.

Real example:

That video of mine lists 8 YouTube channels for learning. Your remix should keep the same title and opener, but you should talk about YOUR OWN recommended list of YouTube channels.

Same proven package, your picks inside.

By day 7 you have:

• 1 topic

• 1 platform

• process for writing scripts/posts

Tomorrow morning is day 8. You film video #1.

3. Post Once Per Day (Days 8-25)

Let’s kill the 2 most common fears I hear.

• The gear fear: you do NOT need to buy anything to get started. Just use your phone front camera and TikTok mobile app to add.

• The being-seen fear: the app shows your videos to strangers interested in your topic, not your friends. If your videos/posts flop, then nobody saw them anyway!

In 2 years, nobody will remember your posts. And you can always hide or delete posts you don’t like.

You’ll still be living with the results of whether you started.

Open Blotato, past a Tiktok/Youtube video link, and have it draft your script. Same title and opener. Exactly like we did in step 1! Next, we’re going to make it sound like YOU.

Prompt 2 of 4: personalize your remixes.

Open Blotato’s AI agent (CHAT right sidebar):

Ask me 3 questions to personalize these drafts to my perspective.

Answer once, and every draft gets your stories and opinions.

Now let’s talk about filming (skip this if you’re text-only)

Here’s the step-by-step process:

1. Read the script out loud and change any word you’d never say

2. Open the Tiktok native mobile app

3. Load the “teleprompter effect”

4. Paste your script in there

5. Hit big red button to start filming

6. Film in 1 take, like a video call with a friend

7. Add the on-screen title within Tiktok editor: usually big text near the top saying what the video is. It should be word-for-word the same exact opener as the viral video you are remixing.

8. Add subtitles making sure they don’t block your face

9. Add a video caption and 3 broad hashtags for your topic (e.g. #fitness #gym #ai)

10. Hit POST!

This process will feel clunky at first, but after 10 posts, you’ll have it memorized. Like anything else, put in the reps.

When you’re making $10k+ months from posting content, you will look back on this fondly as a funny memory.

BTW almost nobody sees your first posts. That’s a gift: free practice while nobody’s looking. And small view differences still show up, which is all the signal you need.

For a tiny account, this is the cheapest growth hack there is:

Reply to every comment you get, within the first 2 hours.

After the first 2 hours:

• if your post is NOT picking up more views, stop commenting

• if your post is STILL growing, keep commenting

Don’t obsess over analytics when you’re starting out. It’s more important you follow every single word in this playbook and TRUST THE PROCESS. Turn your skeptical nagging brain off. I have helped many creators launch their careers, and I am giving you the LAZIEST SHORTCUT I know works, so that you can win faster.

By day 25 you have: 17 videos live, plus data on which openers work for YOUR viewers.

However, after 17 videos copying other people’s viral hits, it may feel like cheating. Good. That pain is exactly what step 4 hits.

4. Find Your Own Voice (Days 25-30)

For the last 5 days, you’ll focus on creating ORIGINAL content.

The formula is simple:

• Do stuff

• Share it

This daily loop takes 45 minutes:

Do 1 small thing in your topic (try something, test something, read something)… then tell people what you learned, struggled with, or share your honest opinion.

This strategy ensures you will NEVER run out of content ideas. And you’ll feel less like a cheater copying others’ viral hits.

You will build the muscle of DOING stuff, then organically making content out of it.

Prompt 3 of 4: weekly idea list.

My topic is [topic]. List 15 small things a beginner could try, test, or build in under 30 minutes, where the result would surprise people. I’ll do 1 per day and share what I learned.

Then use any AI tool to write your script.

Prompt 4 of 4: daily script.

Write a 30-second video script, 70 words max. Start with a scroll-stopping hook. What I did today: [2 sentences]. What surprised me: [1 sentence]. Casual, like I’m texting a friend. Also give me a 1-line caption for the post, plus 3 broad hashtags for my [topic].

Realistic Expectations

Here’s what on-track ACTUALLY looks like at day 30…

(honest numbers, not guru numbers):

• a few hundred followers

• a handful of real comments

• 1-3 videos that beat the rest

That’s winning.

Growth snowballs in months 3, but most people quit early.

Obsess over the number of videos posted, not views.

Because you only CONTROL 1 of those numbers. Stay consistent over a long period of time; you are destined to win (eventually).

Miss a day? Film tomorrow and move on.

25 videos in 32 days beats 11 videos in 100 days.

Don’t beat yourself up over being human and falling off track.

Around day 20, check your numbers ONCE. Usually, 1 or 2 videos gets more views than the rest, even if it’s 400 vs 150. Those are your winners. Make more like them.

No winner yet?

Find more viral inspiration videos (1M+ views each), remix them, and make sure your opener is the literal first thing said. Your post/video’s opener should EXACTLY match the viral inspiration post/video in every aspect - text, visual, colors, timing.

Signal shows up with volume.

By day 30 you have: 22 videos live and a repeatable process.

Step 5 is where your audience becomes income.

5. Your First Dollar (Day 30+)

Most important:

Keep helping for free until you can’t anymore.

People buy from creators they already trust.

And don’t worry about imposter syndrome AKA “who am I to charge?”. You’re not selling something crazy. You’re selling a HEAD START.

You’re 60 days ahead of the person paying you. They’d rather pay than repeat your 60 days.

Your comments already tell you what to sell.

Read every question you’ve gotten. DM the 5 people who engage with you most and ask what they’re stuck on.

Their answers pick your path for you. This is free product research most businesses pay thousands for.

Then pick ONE way to earn and commit:

1. Help people 1-on-1 for money. Coaching, consulting, doing the work for them. Fastest first dollar. Start with people who already know you.

2. Sell a simple guide or a paid community (a private group people pay to join). People pay for help DOING the thing and getting to a desired outcome faster.

3. Build an offer/product from your own pain. It’s how my app happened, after I felt the pain of juggling multiple socials.

Write down the ONE path you’ll try at day 60.

Until then, don’t sell anything.

You now have: the complete starter system, plus ideas for monetization when you’re ready.

The Whole Plan

• Days 1-3: pick your topic (good at it + enjoy it + people pay)

• Days 4-7: pick 1 platform, collect winning openers, practice writing scripts/posts

• Days 8-25: post once per day, reply to comments in first 2 hours

• Day 30+: your first offer, after you’ve built up trust

The courses overcomplicate this so they have something to sell you.

My goal was never to become an “influencer”.

Build the system once, then spend your life on the parts you love. For me personally, I love learning, building, and teaching.

Content creation is a necessary tedious step, so that I can distribute my education to millions of people daily.

Start your 30-day challenge today!

Repost this newsletter and comment below with your socials, and I’ll personally comment on your post to help you out.

You have to do both: repost + drop your account below :)

thx for reading!

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

TLDR; here are 6 copy-paste /goal prompts to build real things with Claude Code and Fable in 2026: a 3D simulation, a game, an interactive dashboard, a full week of content, a 24/7 A/B test for DM automation, and a trip itinerary, plus my prompt template.

Follow along the Youtube livestream here.

Last newsletter, I wrote about loop engineering.

Since then my inbox was flooded with 1 question:

» OK, but what do I actually BUILD with this?…

Sort files is a good baby step…

But it’s not INSANE.

So I sat down and tested 6 awesome projects.

A spider that learns to climb stairs. Flying a dragon. Visualizing human interest at scale. An A/B test optimizing while I sleep.

Who this is for:

People who already have Claude Code and want to master loop engineering’s /goal command with cool projects.

Who this is NOT for: if you haven’t run your first /goal in Claude Code, start with this loop engineering intro.

In case you missed it:

• Loop Engineering: Build Agents with Claude /goal + Routines

• How I Gained 3 Million Followers (with AI)

• 12 Ways to Make Money with AI

• REMEMBER: never send me money! Don’t fall for impersonators.

Simple Prompt Engineering Template

Claude Fable is built for long agentic runs where Claude works for many steps toward a finish line, not just a single chat reply.

If you’re in Claude Code, that’s the model you want to use here.

What you need before you start:

• Claude Code account on a paid plan (Pro or Max)

• /goal command is limited to Claude Code

Every prompt below follows the same 5-part shape:

/goal

TASK: [What Claude should do]

WHY: [Why this matters or who it is for]

OUTCOME: [The exact finished result]

CONSTRAINTS: [What Claude must or must not do]

VERIFICATION: [How Claude should prove the outcome is complete]

Let’s break down each item:

• TASK: the verb. What you want Claude to DO for you.

• WHY: who it’s for or why it matters, the big picture. For Claude Fable, it’s highly recommended to give this context.

• OUTCOME: the finished result, specific enough that “done” is clear; imagine talking to a coworker.

• CONSTRAINTS: the guardrails, limitations, things NOT to change, how many turns to stop at.

• VERIFICATION: the checker. How Claude proves it achieved your desired OUTCOME, not just lying to your face.

Hardest part is defining a clear measurable VERIFICATION step.

Be specific in OUTCOME and blunt in CONSTRAINTS.

“Make it good” isn’t a finish line.

“Scores over 8+ out of 10 using my custom grading skill” is.

1. 3D Simulation

Let’s start with our first project!

Open up Claude Code terminal.

This builds a spider simulation inside Genesis-World, an open-source physics simulator built for robotics and physical AI.

It’s learning to climb stairs:

It’s cool because it’s the same category of problem real robotics labs care about.

Give an agent a body and a goal, let it fail THOUSANDS of times, and watch it learn the skill eventually.

Training a policy from scratch takes real compute time and a few rounds of tuning, even with Claude driving it, estimate 60 minutes.

/goal

TASK: Make a spider learn to jump up stairs in Genesis-World: https://genesis-world.readthedocs.io

WHY: Show spider learning by practice, not following fixed moves.

OUTCOME: Give me the code, the trained spider, and a video.

CONSTRAINTS: Keep it simple and use Genesis World.

VERIFICATION: Spider reaches the top at least 8 out of 10 times.

This project and the next are both fun to do with your kids btw!

2. Gaming

Now let’s build a playable game where you fly a dragon through a fantasy world, POV from on top of the dragon.

Make sure you’ve installed Claude in Chrome, so that Claude can play your game and fix any issues. This tool gives Claude the ability to “see and control” your web browser, so that it can load the game, press the controls, and make sure things work.

Here’s the prompt:

/goal

TASK: Build a game where I can fly a dragon through a fantasy world.

WHY: Make flying feel thrilling, strong, and easy to learn.

OUTCOME: Give me a playable game with flying, landing, fire breathing, and simple goals.

CONSTRAINTS: Keep the controls simple, from the POV of sitting on top of dragon, Elden Ring themed.

VERIFICATION: I can start the game, fly the dragon, complete a goal, and play without major bugs.

3. Visualization

Now let’s build an interactive visualization to explore human interest at scale.

Play around with it here.

This scrapes top YouTube creators with Apify, pulls their videos with 1M+ views, and builds an interactive dashboard so you can actually SEE which concepts show up on the biggest hits.

For example:

• $1

• Survive

• Fight

• Win

Run this in Claude Code, with Apify MCP connected.

Notice the CONSTRAINTS excludes non-English channels, music, gaming, and media, so the patterns you find are actually relevant to YOUR niche, not noise from irrelevant channels.

/goal

TASK: Scrape top YouTube creators with Apify, collect their top videos with 1M+ views, and analyze patterns in their video titles.

WHY: Help creators understand which topics are linked to stronger performance.

OUTCOME: An interactive dashboard that makes the scraped results easy to explore and compare.

CONSTRAINTS: Use real data, and exclude Youtube channels that are non-English, music, predominantly gaming (e.g. XBOX), media (e.g. Netflix, BBC), or for kids.

VERIFICATION: Use browser to confirm the data loads correctly, the main interactions work, and results match the scraped data.

4. Marketing

Use a viral marketing plugin to build a week of ready-to-post content, shown to you in a Claude Artifact so you can review it.

In the prompt below, pay special attention to VERIFICATION.

It doesn’t just say “make good posts” which is vague and not measurable…

(btw, you’d be shocked how many senior marketing leaders prompt like this…)

Instead, the VERIFICATION step sends all posts through the skill /post-grader which scores them on a viral rubric. Then, Claude must loop and improve each post until they all score above 8/10.

This is a much better verification step than “make it gud”.

Don’t expect 1 pass to clear 8/10 across the board.

Claude will AUTOMATICALLY go through 2-3 rewrite rounds per post, and that’s the whole point!

The purpose of “loop engineering” with /goal is to REMOVE YOU from prompting Claude every single time.

Give Claude the tools it needs to evaluate its own work and keep improving to reach the finish line.

/goal

TASK: Install this plugin to create a week of content: https://github.com/Blotato-Inc/blotato-skills

WHY: Save time and help me post consistently.

OUTCOME: A complete week of ready-to-post content, visible in a Claude Artifact.

CONSTRAINTS: Fit each platform’s character limits, and match my brand voice.

VERIFICATION: Run `/post-grader` on every post and improve each one until all scor above 8/10.

Note: this plugin is 100% FREE, you do NOT need to use Blotato.

5. A/B Testing

This one is INSANE for every business owner!!

So good, I almost decided to gatekeep it to myself :)

But I wanna be your favorite anti-guru so here you go! <3

I used Claude Fable to run a 3-day A/B test on my new-follower DM automation in Instagram.

Every 4 hours:

• Claude would check click-through rate results from the A/B test

• rewrite the losing hook with a new angle (e.g. curiosity, exclusivity, urgency, social proof)

• launch a new A/B test with the new DM message

• wait 4 hours for data to come in to see the new winner

• repeat… until achieving XX% CTR

This is loop engineering with the training wheels off!

It ran for 3 days continuously with minimal oversight.

And TRIPLED my conversion rates, which has huge downstream effects, for example, on my daily email newsletter signups :D

The VERIFICATION step doesn’t just ask for a winner. It asks for each test’s numbers, whether the difference is real or just noise (e.g. do we have enough samples, how confident are we), and driving towards a clear copywriting winner.

/goal

TASK: Use Claude in Chrome to run 50/50 A/B tests on my ManyChat new-follower DM to improve CTR to 40%.

WHY: Find which message gets more people to click.

OUTCOME: Keep the winner, rewrite the loser with a new hook, and repeat every 4 hours.

CONSTRAINTS: Track clicks with tags, log every round, and test curiosity, exclusivity, urgency, and social proof.

VERIFICATION: Show each version’s clicks and CTR, explain whether the difference is real or noise, and tell me when we have a clear winner.

What’s really cool?

You can use this framework for many other A/B test scenarios, like:

• A/B testing website copy to improve button CTR

• A/B testing GoHighLevel workflow automations

• A/B testing your cold Linkedin DM message

• A/B testing Facebook ad copy

All running OVERNIGHT 24/7 while you sleep, launching a new A/B test, waiting to review results, and iterating.

6. Trip Planning

I probably should’ve done this example first…

It’s the simplest of all.

This builds a full 3-day itinerary for a group trip, travel, lodging, restaurants, activities, and costs, balanced against everyone’s preferences instead of 1 person doing all the planning in a group chat that goes nowhere.

The CONSTRAINTS line handles budget and food needs, while the VERIFICATION line forces Claude to actually confirm places are open on your dates instead of hallucinating an impractical plan.

As someone who lives in Salt Lake City, I honestly rated the itinerary 7.5/10!

Very good & practical.

/goal

TASK: Plan a 3-day weekend trip for 4 friends in Salt Lake City based on everyone’s preferences.

WHY: Make sure the trip is fun for everyone without stressful planning.

OUTCOME: A complete itinerary with travel, lodging, restaurants, activities, and costs, shown in a Claude Artifact.

CONSTRAINTS: Stay under $800 per person, include everyone’s top choice, respect food needs, and avoid early mornings.

VERIFICATION: Confirm every place is open on our dates, the schedule has no overlaps, and the total cost stays within budget.

RECAP

6 projects, 1 prompt engineering template!

Key Patterns:

TASK and OUTCOME describe what you want to do and the deliverable.

CONSTRAINTS help keep AI from running wild.

VERIFICATION is what turns a prompt into an actual /goal that AI can use to introspect and drive towards completion.

Pick 1 of these 6 projects.

Start today! (otherwise you’ll forget)

Pay attention to VERIFICATION. That’s the finish line.

Let Claude prove to you it’s done, not just lie to your face!

FAQ

What is the /goal command in Claude Code?

It’s a command that gives Claude a finish line instead of a single instruction. Claude keeps working, checking its own progress, until the OUTCOME is met or it hits the turn cap you set in CONSTRAINTS.

What is Claude Fable, and do I need it for these prompts?

Claude Fable is Anthropic’s newest Claude model, built for longer agentic runs like the 6 projects above. You don’t strictly need it to try /goal, but it’s the model I’d point these at for the best results on multi-step builds.

Do I need to know how to code to run these prompts?

No, for #5 and #6. #1 through #4 run in the Claude Code app, which is more technical, but Claude still writes and runs the code. You’re steering, not typing syntax yourself.

What if a /goal prompt runs forever or gets stuck?

Add a turn cap in CONSTRAINTS, like “stop after 30 turns.” Type /goal by itself to check status, or /goal clear to stop it early.

Can /goal handle non-coding tasks like trip planning or DM testing?

Yes. #5 and #6 above prove it. The same 5-part template works whether Claude is writing code or planning a weekend, the TASK and VERIFICATION lines just change.

Why does the VERIFICATION line matter so much?

Without it, Claude has no way to prove OUTCOME actually happened, so it either stops too early or loops forever guessing. A good VERIFICATION line gives Claude something it can literally check, like a video, a click-through rate, or an open-hours confirmation.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

In Part 1, I shared my mom-in-law’s struggles with the DMV test in English.

This inspired me to build PassMyTests.com which generates infinite test practice questions while translating explanations into her native language.

I contemplated other standardized exams, similar in multiple-choice format, but haunting many millions more people:

The dreaded SAT!

In 2023 alone, 1.9 million students took the SAT.

I'm embarrassed to reveal how many hours I’ve spent studying for it.

I’d shove big SAT prep books into my backpack, heading off to Barnes & Noble.

I was one of those kids!!

SAT Math Test

I decided to tackle the SAT Math test first.

My MVP goal:

Generate infinite conceptually plausible and realistic SAT Math questions, while translating the answer explanations for non-native English speakers.

First, I repurposed my DMV prompt:

Your task is to act as a terminal for written test taking the SAT Math Test.

Produce a random 4-choice question.

The question should pulled be from a random set of 500 questions.

The question should be in English.

Finally, return the output in the valid JSON of the following format:

{

"body": "Body of the question",

"choices": ["Choice A", "Choice B", "Choice C", "Choice D"],

"answer": <index of the current choice e.g. A is 0, B is 1, etc.>,

"explanation": "Short explanation of the answer",

"translation": "${translation}"

}

Sabrina Ramonov @ sabrina.dev

Next, I made a new route in my MVP:

PassMyTests.com/sat-math-test

Here’s a sample question and explanation translated to Spanish:

Source: PassMyTests.com - SAT Math

This MVP didn’t take long, but now the hard part begins.

Next, I share several challenges I encountered.

Challenge 1: Missing Correct Answer

Sometimes, ChatGPT fails to include the correct answer as an option:

Sabrina Ramonov @ sabrina.dev

In the above image, the correct answer is 600 square meters.

But this was not an option.

To fix this, I added a verification step to my prompt:

Before returning the output, solve the question.

Check that the correct answer is present in the list of 4 choices.

Sabrina Ramonov @ sabrina.dev

Much better.

Sample questions generated by ChatGPT-4:

Sabrina Ramonov @ sabrina.dev

Sabrina Ramonov @ sabrina.dev

However, even if the correct answer is in the answer list, sometimes the "answer” index in the output does not match the correct answer index.

Recall my prompt:

"answer": 
<index of the current choice e.g. A is 0, B is 1, etc.>,

I haven’t improved this.

I’ll try assigning a random number between 0 and 3 to the right answer, then inserting it into the JSON output.

Challenge 2: In-Context Learning

Nonetheless, initial results seem promising.

It’s been years since I studied for a test, but the questions looked plausible.

I was curious:

Would in-context learning improve it?

In-context learning is an effective prompt engineering technique where you include examples directly in your prompt, enabling ChatGPT to tackle new tasks without finetuning.

I added 4 real SAT Math questions to my prompt:

Your task is to act as a terminal for written test taking the SAT Math Test.

Produce a random 4-choice question.

The question should pulled be from a random set of 500 questions.

The question should be in English.

Before returning the output, solve the question.

Check that the correct answer is present in the list of 4 choices.


Finally, return the output in the valid JSON of the following format:

{

"body": "Body of the question",

"choices": ["Choice A", "Choice B", "Choice C", "Choice D"],

"answer": <'correct_index'>,

"explanation": "Answer explanation, show your work step by step",

"translation": "${translation}"

}


Below are the examples of the type of questions you should generate. Each example is delimited by triple quotes. Use these example to create various questions.

"""

Maria is staying at a hotel that charges $99.95 per night plus tax for a room. A tax of 8% is applied to the room rate, and an additional onetime untaxed fee of $5.00 is charged by the hotel. Which of the following represents Maria’s total charge, in dollars, for staying x nights?

A. (99.95 + 0.08x) + 5

B. 1.08(99.95x) + 5

C. 1.08(99.95x + 5)

D. 1.08(99.95 + 5)x

"""

"""

x^2 + y^2 = 153

y = −4x

If (x, y) is a solution to the system of equations above, what is the value of x^2?


A. -51

B. 3

C. 9

D. 144

"""

"""

The function f is defined by f(x) = 2x³ + 3x² + cx + 8, where c is a constant. In the xy-plane, the graph of f intersects the x-axis at the three points (−4, 0), (1/2, 0), and (p, 0). What is the value of c?

A. -18

B. -2

C. 2

D.10

"""

"""

A researcher wanted to know if there is an association between exercise and sleep for the population of 16-year-olds in the United States. She obtained survey responses from a random sample of 2000 United States 16-year-olds and found convincing evidence of a positive association between exercise and sleep. Which of the following conclusions is well supported by the data?

A. There is a positive association between exercise and sleep for 16-year-olds in the United States.

B. There is a positive association between exercise and sleep for 16-year-olds in the world.

C. Using exercise and sleep as defined by the study, an increase in sleep is caused by an increase of exercise for 16-year-olds in the United States.

D. Using exercise and sleep as defined by the study, an increase in sleep is caused by an increase of exercise for 16-year-olds in the world.

"""

Sabrina Ramonov @ sabrina.dev

Here are questions generated with in-context learning:

Sabrina Ramonov @ sabrina.dev

Sabrina Ramonov @ sabrina.dev

Sabrina Ramonov @ sabrina.dev

Sabrina Ramonov @ sabrina.dev

Sabrina Ramonov @ sabrina.dev

Unfortunately, the questions are repetitive and substantially less random after applying in-context learning.

Even ChatGPT agrees!

In the last image, ChatGPT admits:

“It seems the same question was generated again.“

For now, I removed in-context learning from my prompt.

Challenge 3: Probabilistic Outputs

As I discussed in a previous post on LLM Ops, LLMs are probabilistic.

There is a non-zero probability that occasionally it returns an invalid JSON, despite my explicit request for a valid JSON output.

I encountered the first JSON failure after only n=20 generations:

{
  "body": "In a geometry class, the measures of the three angles of a triangle are \(3x\), \(4x\), and \(5x\) degrees. Which of the following could be the measure of the smallest angle?",
  "choices": ["\(20^\circ\)", "\(30^\circ\)", "\(40^\circ\)", "\(50^\circ\)"],
  "answer": 0,
  "explanation": "In any triangle, the sum of the three angles is always \(180^\circ\). Therefore, \(3x + 4x + 5x = 180\). Solving for x gives \(12x = 180\), so \(x = 15\). The smallest angle would then be \(3x = 3(15) = 45^\circ\), which is closest to \(40^\circ\).",
  "translation": "${translation}"
}

ChatGPT mixed LaTeX and JSON without escaping!

In a real LLM-powered app, if I did not have robust error handling in place, this could crash the app.

An imperfect bandaid, I added another verification step:

Confirm a valid JSON output before finalizing your answer.

Sabrina Ramonov @ sabrina.dev

But there’s still no guarantee because LLM outputs are probabilistic.

A faster and cheaper approach, avoiding another LLM call, would be to assume the JSON output could be malformed and fix it with a custom parser.

Only if the parser fails, then retry with ChatGPT.

Just as I was congratulating myself for this brilliant 80/20 solution, I ran into another issue caused by probabilistic outputs!

On generation N=1000, I got this perplexing and very wrong JSON output:

• JSON just repeats itself

• translation field is the fully translated JSON

• answer index is 2 (x = 6) but the JSON output is 1 (x = 5)

{
    "body": "What is the solution to the equation 4x - 7 = 17?",
    "choices": ["A. x = 10", "B. x = 5", "C. x = 6", "D. x = 4"],
    "answer": 1,
    "explanation": "Add 7 to both sides of the equation to isolate 4x. This gives 4x = 24. Then, divide by 4 on both sides to solve for x, which results in x = 6.",
    "translation": "{
    "body": "¿Cuál es la solución de la ecuación 4x - 7 = 17?",
    "choices": ["A. x = 10", "B. x = 5", "C. x = 6", "D. x = 4"],
    "answer": 1,
    "explanation": "Suma 7 a ambos lados de la ecuación para aislar 4x. Esto da como resultado 4x = 24. Luego, divide entre 4 en ambos lados para resolver x, lo que da como resultado x = 6.",
    "translation": ""
  }
}

Perhaps ChatGPT is deliberately trying to entertain me…

Ugh, and I thought I had solved Challenge #1.

Definitely haven’t solved Challenge #3.

Traditional software engineering deals with structured data.

LLMs deal with unstructured natural language and probabilistic outputs.

I’m still grappling with this new reality.

Conclusion

Altogether, I enjoyed building this MVP and see many ways to improve it.

Source: PassMyTests.com

I believe Gen AI will massively transform traditional test prep and education more broadly, increasing accessibility especially across multiple languages.

Here’s a slice of my “I Should Really Do That” list:

• Evaluation criteria and methodology — given 4 different prompts and different models, generate N=100,1000,10000 questions, and evaluate each question based on a set of criteria

• Restrict question types -- e.g. LINEAR_EQUATIONS, QUADRATIC_EQUATIONS, GEOMETRY_TRIANGLE, LINEAR_ALGEBRA

• Multimodality — generate both an image and the question

• Create brand new questions based on concepts — given a concept along with axioms and theorems, generate a question that utilizes it

  ---

If you have feedback or requests for PassMyTests.com

DM me anytime @Sabrina_Ramonov or Linkedin.

Need More Help? 👋

1/ If you want to grow on social media and scale your business in coaching, consulting, speaking, selling apps, or digital products… check out Blotato

2/ Free AI courses & playbooks here

3/ Free AI prompts & AI automations

She studied aerospace engineering (Rocket Science).

She worked on project Mir, the first modular space station:

Source: WIRED

She’s smart.

She knows how to drive.

But English is not her native language.

I speak only English, so I couldn’t fully appreciate her difficulties, studying in a foreign language, taking a test in a foreign language.

Yet, she’s far from alone.

While waiting at the DMV, I saw many others facing the same struggle.

I realized…

This is a real problem Gen AI can solve today.

By combining infinite test question generation and language translation, AI-powered test practice will help non-native English speakers:

• study and learn concepts in their native language

• understand how these concepts translate to English

• practice test questions (often worded to trick you) in English

• for each question, see explanations in their native language and English

So, I started hacking on PassMyTests.com to help my mom-in-law succeed.

Source: PassMyTests.com

MVP of DMV Test Prep

According to Driving-Tests, over 90% of people who use their DMV practice tests will pass the exam.

But this success rate only applies after taking 10+ practice tests.

This statistic also includes native English speakers.

My stupid simple MVP:

1. You choose a language and US state (some answers vary per state)

2. You get a realistic AI-generated question with 4 choices, all in English

3. You select an answer

4. You see an explanation for the correct answer, in both English and translated by AI to your native language

After a few iterations, I arrived at the prompt:

You act as a terminal for written test taking at a DMV in the state of ${params.state}.

Produce a random 4-choice question.

The question should pull from a random set of 500 questions.

The question should be in English.

Return the output in the valid JSON of the following format:

{

"body": "Body of the question",

"choices": ["Choice A", "Choice B", "Choice C", "Choice D"],

"answer": <index of the current choice e.g. A is 0, B is 1, etc.>,

"explanation": "Short explanation of the answer",

"translation": "${translation}"

}

Sabrina Ramonov @ sabrina.dev

I haven’t done a formal evaluation, but from eyeballing outputs, this instruction seems to increase randomness:

“The question should pull from a random set of 500 questions”

The key term is seems.

I could be fooling myself.

My next post on this project will dive into evaluations.

The variable ${translation} works like this:

• If the user didn’t select a language, then there’s no translation

• If the user selected a language, then we ask ChatGPT to translate: "translation": "Translation of the explanation in Spanish"

Putting it all together, here is my MVP for practicing DMV test questions:

Source: PassMyTests.com - DMV

Below is a sample question where the correct answer varies by state.

In California, the answer is 0.08%.

But in Utah, the answer is 0.05%.

Source: PassMyTests.com - DMV

Challenges

Multiple Modalities

One challenge is is that only certain types of questions are generated.

I couldn’t generate an image of a street sign or show an image as an answer.

This MVP is limited to text-based question and answers.

No support for other modalities yet.

Here’s a sample DMV test question that involves an image:

Complex Situations

In the same vein, I did not generate complex situational questions.

For example, a question could show a picture of two cards behind a school bus, while another car is turning left from the opposite lane.

Then, the question asks you to determine right of way.

Unfortunately, my MVP neither generates images nor complex situations.

Changing Laws and Knowledge

I grew up learning 10 and 2 as the optimal position to keep my hands on the steering wheel.

But knowledge evolves.

Because of airbags, now the optimal recommended position is 9 and 3.

Laws also change.

Some factual data might be year-dependent, presenting another challenge.

Ideas for Future Work

I see multiple vectors for improvement.

I could slice a DMV handbook, making a RAG that answers factual questions directly from the latest handbook, along with the explanation.

This would address the challenge of non-static laws and knowledge.

Another opportunity is to increase the quality of LLM outputs by having 2 interacting agents:

• one agent generates test questions

• second agent answers test questions

I could have open and closed-book variants, although presumably the test-taking agent should perform better if it’s open-book!

In general, it would be interesting for someone to turn DMV tests into a proper dataset LLMs can be evaluated on.

The non-trivial reasoning tasks are especially intriguing, such as analyzing an image to determine right of way, law violations, and best next actions.

Happy Ending

Relieved to report…

After a week practicing on PassMyTests, my mom-in-law passed her exam!

Her favorite feature — reading the English explanation right above the explanation in her native language.

She already knew most concepts, but often mistranslation caused confusion.

Seeing the English and translated explanations side-by-side helped her understand nuances in the questions, answers, and explanations.

Next up in Part 2 this series, I tackle a much bigger standardized test.

If you have feedback or requests for PassMyTests.com

DM me anytime @Sabrina_Ramonov or Linkedin.

Need More Help? 👋

1/ If you want to grow on social media and scale your business in coaching, consulting, speaking, selling apps, or digital products… check out Blotato

2/ Free AI courses & playbooks here

3/ Free AI prompts & AI automations

(educational explainer video about my fave tea, gyokuro)

TLDR… HeyGen HyperFrames is a free, open-source tool that turns a plain-English prompt into a real video. You describe the video, an AI coding agent like Claude Code writes it as a website, and your own computer turns it into a video file. HyperFrames itself is free. You just need an AI agent, like Claude or Codex, to run it.

You want a short video.

A product intro, a website explainer, an educational short.

The old way?

Open a video editor. Drag clips on a timeline. Fiddle with it for 2 hours.

Now there’s a free tool that flips this.

It’s called HyperFrames.

You describe your video in plain English. AI makes it for you. You get a real video.

This used to be pretty technical to set up.

But NOT anymore!!

You tell Claude Code to install it, and Claude does all the work.

If you’ve used Claude before, you can do this IN MINUTES!

What you need before you start:

• A Mac or Windows computer.

• Claude Code. This is the hands-on version of Claude that works on files in a folder, not the chat window. It needs the Claude Pro plan ($20/month), which includes Claude Code. New to it? Watch my install tutorial first: How to Install Claude Code.

1. Set It Up

First, what IS HyperFrames?

Think of a video as a website.

Text, images, a logo, some motion. That’s a website with animation.

HyperFrames takes that website and records it, frame by frame, into a video.

A traditional video editor makes YOU move every clip. HyperFrames lets you describe the whole thing in a sentence, and AI does the rest.

You stop dragging clips on a timeline.

You start describing the video you want.

The setup is 3 simple steps.

Do this once.

Step 1: Make a new empty folder on your computer. Name it something like “my-videos”. (On a Mac, right-click your desktop and pick “New Folder”. Same idea on Windows.)

Step 2: Open Claude Code in that folder. 2 ways, pick one:

• Easiest for most people: open the Claude desktop app, click the “Code” tab, then click “Select folder” and pick the folder you just made.

• Terminal: open the Terminal app, go to your folder, and run claude. (On a Mac: type cd with a space, then drag your folder onto the Terminal window and press enter.)

Both run the same Claude Code. The desktop app is the easy path for your first videos. The terminal is more reliable for the big, long, complex editing tasks.

Step 3: Type this into Claude and hit enter:

install hyperframes video editor

That’s the whole setup.

Claude installs HyperFrames for you.

If it asks to install “homebrew” and “ffmpeg” - both are fine, you can say YES.

If the command didn’t work, you can open a Terminal and run:

npx skills add heygen-com/hyperframes

2. Beginner Prompt

Now we’ll make your first video, which will look like this:

Here’s the prompt:

Create an 8-second landscape video explaining the health benefits of sleep, and choose one strong animated visual metaphor that makes the idea instantly understandable.

Keep it fast to render.

Claude creates a plan and builds the video. Give it a few minutes. Mine finished in about 5, so if it’s still working, it’s not stuck.

Want to watch it?

Type in Claude:

launch preview

It opens the HyperFrames preview studio in your browser.

If you’re in Claude Desktop, you can also click the link to see the video play.

Here’s what the studio looks like with that exact sleep video. Notice the battery at 100%, that’s the visual metaphor for recharging your body:

Happy with it?

Tell Claude “save it as a video”, or click EXPORT in the preview.

It drops a finished video file into your folder.

Now you have a real video, made from 2 simple sentences.

3. Beginner-Intermediate Prompt

Now that you’ve made 1 video, let’s write better prompts.

A vague prompt gives you a vague video. Good HyperFrames prompts have 3 parts:

• The source: a website link, a topic, or a script. What’s the video ABOUT?

• The length: how many seconds. “15 seconds”, “30 seconds”, “1 minute”.

• The vibe: the feeling. “Bold and cinematic”, “clean and minimal”, “warm”, “dark and techy”.

This next prompt is the same simple idea as the sleep video, with way more pizzazz. Here’s the video we’ll make:

It uses the faceless-explainer workflow. If Claude doesn’t have it yet, it offers to install it, say yes.

Using hyperframes and /faceless-explainer, create a 30-second faceless explainer about the health benefits of gyokuro tea. Do not make it educational-looking. Make it feel like the opening sequence of a high-budget Netflix investigation: dark cinematic typography, mysterious animated diagrams, fast evidence-board transitions, one shocking hook in the first 3 seconds, and a satisfying “now I finally get it” visual metaphor by the end.

4. Intermediate Prompt

Let’s take it one step further…

Now, point Claude at a real website, and it builds the video from the website’s colors, fonts, and screens.

AI visits the page and build the video to MATCH your brand.

Here’s a fairly polished branded video after I gave it my website:

This one uses the website-to-video workflow. If Claude doesn’t have it yet, say yes when it offers to install it. Here’s the prompt:

Using hyperframes and /website-to-video, create a 25-second cinematic product launch video from blotato.com. Capture the website’s actual colors, typography, UI, and brand feel, then turn it into a premium Apple-keynote-style reveal with dramatic pacing, animated UI zooms, 3 big benefit cards, clean voiceover, subtle sound effects, and a final CTA. Make it feel like this tiny website is announcing the future.

5. Next Level Prompt

This is the deep end, and it’s the one that blew me away!

It turns a plain talking-head video into a polished final edit.

Point Claude at a plain talking-head video (you talking to camera). This 1 prompt cleans up the recording AND turns it into a fully animated edit. Cards, captions, visual metaphors, the works.

This Youtube video was fully edited by HyperFrames:

You do NOT need to understand every line. Copy it, give Claude your video, and watch what it does. If Claude asks to add a workflow for this, say yes.

Here’s the next level prompt:

Turn a talking-head explainer into a designed motion-graphics edit. Transcribe for word-level timing, then clean the take first: cut false starts, stutters, repeats, and filler, and trim pauses over 0.3s, all transcript-located, seamless, audio + video together. Split the cleaned script into concept beats; for each concept invent the strongest visual metaphor (object, mechanism, diagram, comparison) that makes it instantly legible. No bullet lists or icon rows; vary the forms beat to beat.

One visual system throughout:

• Palette (light, blue): soft white radial-gradient bg (#ffffff→#f4f6fb→#e8edf7) with faint blue blobs; accent #2563eb, secondary #38bdf8, ink #0f172a, muted #64748b, cards #fff.

• Type: bold grotesque (Archivo/Sora) 800-900 for headlines/numbers, clean sans (Inter/Outfit) for body; only the emphasized keyword is blue.

• Motion: white cards (~22px radius, soft shadow) that rise + fade in and settle, no jitter. One paused, seek-safe GSAP timeline; 1920×1080@30.

3 graphic roles: (a) full-frame opaque cards replacing the face, showing metaphors, quotes, comparisons, steps, titles; (b) transparent overlays over the face, like a blue “pill” keyword pop, lower-thirds, captions; (c) one recurring spine motif for the video’s throughline that returns at each section marker and pays off at the climax.

Sync: align each graphic to its spoken line, cut to a card as the concept lands, return to the presenter between sections. Hold settled cards for their beat; keep pops brief. Output one MP4 at the source aspect ratio.

Now hold up, let’s be realistic…

This is NOT a magical slot machine where 1 prompt gives you a perfect video every time.

The AI gets you 80% there. You still preview it, spot awkward things to be fixed, and ask AI to iterate. That last 20% is you.

The good news: the more you invest in teaching HyperFrames your visual preferences and editing style, the more consistent the results will be!

Highly suggest spending a couple hours playing around with it :)

6. Take Control of the Whole Video

Most importantly:

HyperFrames doesn’t just hand you a finished video and hide how it got there.

It shows its work.

And you can EDIT any part.

The AI always runs through 7 steps.

Each step leaves you a text file you can open and tweak:

• Capture: it grabs your website’s look (colors, fonts, images).

• Design: it writes a DESIGN.md file… your brand rules for the video.

• Script: it writes the narration into a SCRIPT.md file.

• Storyboard: it plans each scene in a STORYBOARD.md file.

• Voiceover: it makes the narration audio and lines up the timing.

• Build: it writes the actual video scenes.

• Validate: it checks each scene looks right before finishing.

You don’t have to touch any of these… but when 1 scene feels off, you don’t re-roll the whole video.

Just open STORYBOARD.md, change that ONE scene (make it slower, swap the mood, change the words), and tell Claude to rebuild just that scene.

The magic isn’t “1 perfect prompt.”

It’s that every piece is a plain file you can open and fix.

That’s the difference between a black box and a tool you control.

Try this. Edit 1 scene by hand.

Make a short video, open STORYBOARD.md, change 1 scene’s words or mood, and re-create just that scene.

7. Import Extra Workflows

What you installed is the general-purpose HyperFrames. It can make any video.

But it also has OPTIONAL “workflow” skills. Each one is a recipe for a specific kind of video.

Think of them like add-ons. You import the one that matches what you’re making, and it already knows the best steps for that job.

The ones worth grabbing:

• /product-launch-video: a product link or brief turns into a launch or promo video.

• /website-to-video: any website turns into a tour or a short social clip.

• /faceless-explainer: plain text with no link turns into a faceless explainer, narration included.

• /embedded-captions: takes a talking-head clip you already have and adds captions.

• /talking-head-recut: takes your talking-head clip and dresses it up with title cards and name banners.

• /motion-graphics: a short, no-voice animation… a logo, moving text, a stat.

• /music-to-video: a music track turns into a beat-synced video.

• /slideshow: builds a slideshow-style video.

There are 2 more for developers (/pr-to-video for GitHub code changes and /remotion-to-hyperframes for porting remotion videos). Skip those if not relevant.

How to import them. Easiest way, same as before: just tell Claude.

Type `install all the hyperframes workflows`.

Claude grabs them all.

After that, just describe your video and Claude routes it to the most appropriate workflow.

8. Using HyperFrames in Codex or Antigravity

Not on Claude Code?

HyperFrames isn’t locked to it. The same tool works in other AI coding agents, including Codex and Google Antigravity.

Setup is identical: make a folder, open your agent in it, and install HyperFrames the same way. Tell your agent “install hyperframes”, or run npx skills add heygen-com/hyperframes yourself.

Only the trigger changes a little:

• Codex: start your message with /hyperframes, then describe the video. Like “/hyperframes make a 15-second product intro, bold and cinematic.”

• Google Antigravity: no command needed. Just describe the video in the agent sidebar, and Antigravity loads HyperFrames on its own. Name a workflow (like /faceless-explainer) when you want to be exact.

It also runs in Cursor, Gemini CLI, and GitHub Copilot CLI, same install.

Everything else in this guide works the same, the prompts, the preview, the save.

Totally fine to use the AI agent you already have.

RECAP

1. Set it up once. Make a folder, start Claude Code, type “install hyperframes”. Claude does the rest. It’s a 1-time thing.

2. Describe your video. Source plus length plus vibe. A link, “25 seconds”, “bold and cinematic”. The AI builds it, you preview it and get a real video.

3. Take control. Every scene is a plain file. Open the storyboard, fix 1 scene, rebuild. No black box.

Once you’ve set this up, you can make unlimited on-brand videos from a sentence, on your own computer, essentially for free.

No timeline.

No credits.

No accounts (except your normal Claude or Codex).

HyperFrames isn’t a video editor you learn.

It’s a video team you speak to.

FAQ

What is HeyGen HyperFrames in simple terms?

It’s a free, open-source tool that turns a plain-English prompt into a real video. You describe the video, an AI coding agent writes it as an animated website, and your computer turns it into a video file, right on your own machine.

Is HyperFrames free? Does it use HeyGen credits?

HyperFrames itself is free and open-source, and it runs on your own computer, so it uses no HeyGen credits. Credits only come into play if you connect it to HeyGen’s cloud avatar engine, which is a separate, advanced step.

Do I need a HeyGen account to use HyperFrames?

No. The tool runs locally on your machine. You only need a HeyGen account if you later want to add talking AI avatars from HeyGen’s cloud.

What do I need to install before I start?

Almost nothing by hand. You need Claude Code, the hands-on version of Claude. Make a new folder, start Claude in it, and type “install hyperframes”. Claude installs the rest and tells you if it needs a free helper like FFmpeg.

Do I need to know how to code?

No. You use Claude Code and type plain English, like “install hyperframes” and then a description of your video. Claude runs any commands and writes all the video code for you.

What does a good HyperFrames prompt look like?

Include 3 things: the source (a website link or topic), the length (like “25 seconds”), and the vibe (like “bold, cinematic, dark theme”). Example: “Create a 25-second product launch video from https://yoursite.com. Bold, cinematic, dark theme energy.”

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

• Prompt Engineering for Claude Fable

• 3 Simple Ways to Save Money (with AI)

• Women Build AI (free community)

TLDR… GLM-5.2 is a new open-weight AI model from Zhipu AI (Z.ai) in China: a 750-billion-parameter Mixture-of-Experts model with a 1-million-token memory and an MIT license, so anyone can download and run it for free. It matters because the best FREE, open AI is now only about ~6 months behind the best PAID, closed AI from the US labs, and on one Semgrep security test it out-scored Claude Code for about 1/6 the cost. Here’s what’s real, what’s hype, and how to try it on a Mac or Windows.

A model you’ve never heard of, from a company most people can’t pronounce, just out-scored Claude Code on a hard security test.

For ~1/6 the price.

It’s called GLM-5.2.

It’s open, it’s cheap, and AI researchers are raving!

So I read the actual benchmark from Semgrep and the fine print most people skipped past.

Here’s what’s REAL vs what’s HYPE…

1. So what IS GLM-5.2?

GLM-5.2 is the newest AI model from a Chinese company called Zhipu AI, which goes by Z.ai.

They dropped it on a Saturday, June 13th, and published the model itself 3 days later.

3 things make it different from Claude or ChatGPT.

• It’s open-weight. The model’s “brain” (the file that holds everything it learned) is published for anyone to download, run, and tweak. Claude and ChatGPT are the opposite: you rent access, you never get the model itself.

• It’s free to use commercially. The weights ship under an MIT license, which is about as permissive as it gets. You can build a business on it and owe Z.ai nothing.

• It’s big, but cheaper to run than it looks. It has around 750 billion parameters (the dials that store what it knows), but only about 40 billion switch on for each word. So it runs at a fraction of the cost you’d expect from its size.

It can hold ~1 million tokens in mind at once, roughly 750,000 words. That’s up from 200,000 in the last version.

Translation:

You can feed it an entire book series and it keeps track.

However, “open-weight” is NOT the same as “open-source.”

Open-weight means you get the finished brain, but not the recipe (the training data stays private).

2. Why the AI world lost its mind

Open models from China have been creeping up on the expensive American ones for a while.

GLM-5.2 is the moment a lot of researchers say they finally caught up on the thing that matters most right now:

AI agents that DO tasks, not just chat.

The receipts:

• It tops the open-model rankings on the coding and agent tests people actually track (Terminal-Bench, SWE-bench Pro, and friends), the best any open model has done.

• On several long, multi-step coding tests it even edges out GPT-5.5, for a fraction of the cost (per VentureBeat).

• On a public agent leaderboard, it’s the ONE open model mixing it up with the latest from OpenAI and Anthropic.

• Nathan Lambert, one of the most respected open-model researchers, called it “the step change for open agents” and compared the moment to DeepSeek R1, the last time an open model rattled everyone this hard.

Lambert points out the gap between the best closed model and the best open one is now about 6.8 months. Not years. Months!

The expensive, closed AI you pay for today…

is roughly what you’ll be able to download for free in ~6 months.

That’s the big deal!

3. About that “beat Claude” headline (read this part)

This is where most AI news got sloppy, so I’ll be careful.

A security company called Semgrep ran a bunch of AI models against the same test: find a common website security flaw called IDOR (the kind of bug where you change a number in a web address and suddenly see someone else’s data).

The result that made headlines:

GLM-5.2 scored 39% and beat Claude Code Opus (which scored about 32%) on the same bare setup, at about $0.17 per bug found. Roughly 1/6 the cost.

True. But here’s what the headline leaves out, straight from Semgrep’s own write-up:

• It was ONE task, one dataset, one run. Not a sweeping “open models have caught up.” In their words: one open model, on this task, under these conditions.

• The setup around the model mattered more than the model. Semgrep’s own purpose-built security tool still won by a mile (53 to 61%), whether it ran on GPT or Claude underneath.

• It was tested against Claude Code, not Claude’s top model Fable/Mythos. GLM-5.2 did NOT beat Fable 5. It beat the Opus-powered coding agent on a narrow security task.

So the honest version is less sexy than “China beats Claude,” but more useful:

A cheap, open model just matched an expensive, closed one on a real task.

Not everywhere. Not always.

But it happened, and that was UNTHINKABLE just 1 year ago.

4. Why “open” is the whole point

This is the part that actually matters…

Right now, the best AI on earth lives inside a few US companies.

You don’t own it. You rent it.

They set the price. They set the rules.

They decide what it’s allowed to say and do.

And they can change ANY of that overnight. A price hike. A model getting retired. A new rule about what you can ask.

You have no say.

An open-weight model is the ONLY real counterweight to that.

When the model’s brain is a file YOU can download, the power tips back to the PEOPLE:

• Price stops being their decision. Free, open models force the expensive labs to compete. Their prices drop whether you ever touch GLM-5.2 or not.

• Your data can stay home. You can run an open model fully inside your own computer or company, so nothing you type ever leaves the building. For anyone handling private or client work, that’s a big deal.

• Nobody can rug-pull you. Once those weights exist, they exist forever. No subscription to cancel on you, no model getting deprecated, no terms-of-service surprise.

• Anyone can build on it. The next great AI app doesn’t need one company’s blessing. That’s how progress stays fast and cheap.

Here’s why GLM-5.2 specifically is a turning point.

For YEARS, “open” meant… “fine for hobbyists, not for real work.”

That’s over.

The best open model is now about ~6 months behind the best closed one… and the gap is hopefully closing.

And look at WHO is carrying open AI right now:

CHINESE labs like Z.ai, while the big US labs become more closed every month.

If “open” loses this race, here’s the future we get: 1 or 2 companies own AI that’s 100X better than everything else, and we are stuck with their decisions.

THAT is why this matters.

Not the benchmark.

The balance of power.

5. How to try GLM-5.2 (the easy ways)

You do not need to install anything to test it.

• Easiest: open Z.ai’s chat at chat.z.ai and just talk to it like ChatGPT. You get a free tier, then it caps.

• Pay-as-you-go: through OpenRouter or the Z.ai API, it runs roughly $1 per million tokens of input and $3 to $4 per million of output, depending on the provider, and the price keeps dropping as hosts compete. For most people that’s a few dollars a month of real use.

• Flat monthly plan: Z.ai’s “GLM Coding Plan” starts at $18/month (Lite), $72/month (Pro), or $160/month (Max), and it plugs GLM-5.2 into coding tools like Cursor and Cline.

For 99% of you, the easy browser chat option #1 is all you need to form an opinion.

6. How to set it up on your computer (Mac vs Windows) (ADVANCED)

Skip this if you just want to use it. The browser chat above is the easy path. This section is for people who want it running on their own machine.

But first, let’s be real…

GLM-5.2 is a 750-BILLION-parameter model.

The full version needs roughly 250+ GB of memory to run.

A normal MacBook or Windows laptop CANNOT run the real thing.

There are only 2 reasons to bother putting it on your computer.

Reason 1: you want GLM-5.2 inside your OWN tools

Say you want it powering a code editor, an automation, or an app on your machine. You install one small program called Ollama, and then any tool on your computer can call GLM-5.2 the same way.

• On a Mac: go to ollama.com/download, download the Mac app, and open it. Then open the “Terminal” app and type ollama run glm-5.2:cloud. Sign in when it asks.

• On Windows: go to the same ollama.com/download page, run the Windows installer, then open “PowerShell” (search for it in the Start menu) and type ollama run glm-5.2:cloud. Sign in when prompted.

Now this trips people up: you’re still running GLM-5.2 on Ollama’s servers, NOT your machine. You get no privacy and no offline use out of this. The ONLY thing you’re buying is 1 local hookup that every app on your computer can point at. If you don’t need that, close this and just use the browser chat.

Reason 2: you want it 100% private and fully offline

THIS is the version that truly runs on your own hardware, where nothing you type ever leaves the room. It’s also the expensive one.

To run a shrunk-down (2-bit) version offline, you need about 240 GB of memory, which in practice means a maxed-out Mac Studio (256 GB) or a workstation with a serious graphics card plus 256 GB of system RAM. No normal computer comes close.

If that’s you, the simplest path is LM Studio (a free app for Mac and Windows): install it, search “GLM-5.2” in its model browser, pick a quantized version that fits your memory, and download it. Expect about 3-9 words per second on the best consumer machines. Slow AF, but it’s yours.

For everyone else, stick to the browser.

Recap

GLM-5.2 is not “the Claude killer” and anyone screaming that didn’t read the study.

What it IS: proof that the gap between the expensive AI you rent and the free AI you can own shrank from years to months.

That’s good for all of us, no matter which tools you use.

Cheaper AI. More choice. Less lock-in.

I still reach for Claude Code for my daily work.

But I’m watching the open models closely now, and you should too.

The big story isn’t “China beat Claude”

It’s that good AI is getting cheap and hard to lock up.

FAQ

Is GLM-5.2 free? The model weights are free under an MIT license if you download and run them yourself. Using it through a hosted service (the browser chat, an API, or a coding plan) costs money, though far less than the big closed models. There’s no permanent free unlimited API.

Can I run GLM-5.2 on my laptop? Not the full model. It needs around 250 GB of memory, which no normal laptop has. You can use it on any computer through Ollama’s glm-5.2:cloud option (which runs on Ollama’s cloud servers, not your machine) or the browser chat. True offline use needs a 256 GB Mac Studio or a heavy-duty workstation.

Is GLM-5.2 better than Claude or ChatGPT? On some coding and agent benchmarks it’s the best open model and competitive with the top closed ones. For everyday writing and content tasks, Claude, ChatGPT, and Gemini are still smoother for beginners. A benchmark win does not always match your day-to-day experience.

Did GLM-5.2 really beat Claude? On one security test from Semgrep, GLM-5.2 out-scored Claude Code running Opus 4.8, at about 1/6 the cost. But it was one task, one run, it did not beat Claude’s top Fable 5 model, and Semgrep’s own security tool still beat both. Read it as “a cheap open model matched an expensive one once,” not “Claude lost.”

What’s the cheapest way to use GLM-5.2? The browser chat at chat.z.ai to try it free, then pay-as-you-go through OpenRouter or Z.ai (a few dollars a month for most people) if you want it in your own tools.

Do I need to be technical to use it? No. The browser chat works like ChatGPT. You only need technical skills for the run-it-on-your-own-hardware path in Section 6, which most people should skip.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

• REMINDER: I will never ask you for money. Never pay for “courses”, trading platforms, or investment opportunities. Those are all fake.

• The Most Important Scientific Revolution Has Already Started

• How to Be So Productive with Claude it Feels ILLEGAL

• How I used ChatGPT to erase my digital footprint

TLDR… here are 12 ways to make money with Claude in 2026, ranked from your fastest first dollar to the most scalable income: 1:1 coaching, corporate training, paid courses, AI strategy consulting, brand sponsorships, a paid community, paid speaking, then 4 agency models (social media, AEO, vibe code, AI automation), and finally vibe coding your own SaaS product.

The job market is slightly better, but still TERRIBLE.

An income you control is the only safety net left.

So I mapped 12 ways to make money with Claude, ranked from the fastest first dollar to the most scalable income, all on ONE chart.

I built an AI company and sold it for millions.

Now I’m approaching 3 million followers and teach millions of people how to use AI - for free.

If you’re thinking, “I’m not technical enough to make money with AI”… trust me, you don’t need to be.”

You need to understand the tradeoffs.

The pros and cons of each path.

I know that many of you reading this feel OVERWHELMED with INFORMATION OVERLOAD. So this is my best attempt to explain different ways to make money with AI… so you can make the right choice for yourself.

None of these are “easy passive income”… expect months and years of hard work.

But I gatekeep nothing!

This is the only playbook you need.

Full livestream here:

Who this is for:

People who want to seriously monetize AI in 2026.

NOT play with prompts.

But confused where to start, drowning in information.

Who this is NOT for: people who already run a profitable AI business and know exactly which lever to pull next.

First, the ONE chart that decides everything

Before the list, you need the map.

Every path lives somewhere on 2 axes.

Leverage (vertical): does this scale WITHOUT your time? High leverage means I could sell the same course to 1,000 more people and do zero extra work. Low leverage means you’re selling your 24 hours for dollars.

Time to First Dollar (horizontal): how fast does money hit your account? Services and consulting pay FAST. Building a brand or a product pays slow, but scales way bigger.

Fast money lives bottom-left (trade your time).

Scalable money lives top-right (build an asset).

So when you feel stuck thinking “why is this taking so long to make money?”… it’s because you picked a path on the slow right side without knowing it.

Fast money = trading your time.

Scalable money = building assets.

Agencies sit in the middle. Cold-outreach cash AND a process you can productize.

CAVEAT: the examples below walk through ONE example target buyer. Because I want to illustrate each income path with a clear example. But there are many possible target buyers or offer variations you can pursue.

1. 1:1 Coaching For Busy Professionals

This is the fastest cash of all 12. Lowest leverage too.

I could open any of my inboxes right now and find people who want to pay me for coaching.

• Buyer: Busy professionals. Tech, middle management, anyone employed full-time who can easily afford a couple hundred dollars and wants to become “AI native” for a promotion.

• Outcome: Build them 5 reusable skills in Claude that save 10 hours per week. A “skill” is just a saved set of instructions Claude runs on command, like a mini app you build in plain English. For total beginners, start simpler: set up their Claude with their bio and automate 1 weekly task like email triage.

• Distribution: Instagram, Facebook, LinkedIn. IG and FB are easiest because you set up DM automation and those audiences are used to spending money. I treat Instagram as the new LinkedIn of 2026.

• The catch: It’s your time. You have 24 hours a day like everyone else, and you’re trading them for dollars.

Coaching is the fastest way to your first dollar.

It’s also the hardest one to scale, because the product is YOU.

2. Corporate Training For Small-to-Mid Enterprise

Same idea as coaching, but you train a whole TEAM at once. That’s where the leverage kicks up.

• Buyer: Small-to-medium businesses. They have employees, which means they have money.

• Outcome: Train the team to save 15+ hours per week with tailored Claude skills. Train 10 people, save each 5 hours, and you just handed the company back 50 hours a week.

• Distribution: LinkedIn (where companies hang out), or go LOCAL. The SBA, local Facebook groups, chambers of commerce, Entrepreneurs’ Organization. Go hyper local and you’ll have ZERO competition. I live near Salt Lake City. Nobody here cares about AI. The good news? Zero competition.

• Why it sells: AI is being shoved down every company’s throat. Small businesses are panicking about falling behind. My take: don’t lay people off, upskill them. They already have the domain expertise.

Your curriculum stays about 70% the same across clients, 30% tailored. That reusable 70% is why this beats coaching on leverage.

3. Paid Courses

Now we jump to high leverage.

Sell the same course to 1,000 people, do no extra work.

Here’s an EXAMPLE topic for the course, including how I’d position the offer. But your course can be anything related to learning AI:

• Buyer: Content creators trying to figure out how to use AI (one example, this works for any niche).

• Outcome: Help them grow automatically on 2 more platforms and save 10 hours per week. Take the tedious stuff off their plate: research, hooks, scripts, clipping long videos into shorts.

• Distribution: Youtube, Instagram, Facebook. IG and FB give you comment-to-DM automations (“comment CLAUDE and I’ll send you the link”). YouTube is the gold standard for education… everyone says “I’ll go to YouTube to learn this.”

• The #1 way people fail: They spend 99% of their time perfecting the course and 1% on marketing. Then they launch and nobody buys. I see this same mistake over and over.

Honest confession: I’m not a great content creator. I build content SYSTEMS that make a lot of content, and a small percent of it is good.

The course is the easy part.

The marketing is the whole game. Flip that ratio.

4. AI Strategy Consulting

This is strategy only.

You don’t build anything. You find the money.

Similar to 3, you can target many types of businesses with this model. Here, I am talking through ONE example of a target buyer.

• Buyer: Local businesses (works for any business, really).

• Outcome: Go into the business, understand everything, and identify 3 to 5 places AI can make more revenue or save hours per week. For a multi-team operation, finding 100 hours a week of waste is not far-fetched.

• Distribution: Local Facebook groups, chambers of commerce, the SBA. Become known as THE AI person in your town and referrals flow. Even better, run local Facebook ads to a free AI workshop, host it, and collect referrals.

• The universal frame: Business owners care about 2 things. Making more money and saving time. The closer your offer sits to those, the more you charge.

The 2 leaks almost every business has: inbound leads that never get a callback, and customer support that never replies. Start there.

Business owners want the outcome.

They don’t care if you use AI, an alien, or voodoo.

5. Brand Sponsorships For AI Companies

Get paid to film a video. That’s the whole job.

• Buyer: AI and SaaS companies, especially recently funded ones with budgets set aside for influencers. Real example: a startup that just launched an MCP server (a connector that plugs their tool into Claude) and wants tutorials showing it off.

• Outcome: Film a video demoing their product. That’s it.

• Distribution: Build on YouTube. Brands reach out, or you cold email: “I’m a YouTuber, here’s my follower count and average views, I love your app, I’d love to feature it.” Then send a rate card with 3 options.

• The money: Sub 10K-subscriber on Youtube charges a few thousand dollars per video. Bigger channels charge $5,000 to $10,000+. And 99% of the time payment is NOT tied to performance. You get paid even if the video flops… which is most of the time. Terrible for sponsors like me, but that’s the current market.

New YouTubers are blowing up right now simply by talking about Claude. The search volume is enormous. Copy a viral topic and thumbnail, add your spin, and you’ll hit 10K subscribers fast. When I say “long,” I mean 3 months, not 3 years.

Claude raised billions at nearly a trillion-dollar valuation.

Startups are throwing money at any AI creator who can hold a camera.

Even if your camera’s blurry af like mine… you can get paid:

6. Paid Community

The math on this one is so good it’s hard to ignore: paid communities to learn AI.

Here’s an EXAMPLE target buyer:

• Buyer: People who’ve heard of Claude and want to go deeper. Segment them by demographic, tools, or skill level. My example: women who want to learn AI.

• Outcome: Weekly new skills, weekly trainings, Q&A, accountability, and peers at the same level.

• Distribution: YouTube, Facebook, Instagram, plus Substack. You can even build a profitable community right inside Substack… the playbook is to start free and gatekeep your best stuff. Skip paid ads unless you know what you’re doing, or you’ll burn thousands fast.

• The math: $50/month X 200 members is $10,000/month. RECURRING. There are 1,000+ people watching this livestream. You’re telling me you can’t find 200 who want to learn Claude?

Honest caveat: recurring revenue means recurring work. People expect fresh content or they churn. And with no audience, nobody joins an empty room.

I run a free community and it will NEVER be paid.

BUT the math on paid is so good, it’s an amazing way to start.

7. Paid Speaking For Non-Technical Companies

Lower leverage (it has to be YOU on stage), but you get some credibility.

• Buyer: Non-technical companies and organizations that gather lots of businesses, so you teach many members at once.

• Outcome: Anywhere from a hands-on Claude workshop (“let’s set up 3 skills today”) to a high-level “what is AI and why it matters” keynote.

• Distribution: LinkedIn plus any social presence. Grow a brand and you’ll get inbound DMs asking you to speak.

• The money: Pay ranges from a couple thousand to well above $20,000. And people who know nothing about you suddenly trust you because “wow, you spoke at that event.”

The honest downside: it’s slow to break in. Organizers want a shiny resume, and when a slot opens they pick from several candidates. Prep is heavy too. You’re in front of hundreds of people, so you can’t have random bugs in your slides.

Coaching: if Claude goes down, you reschedule.

Speaking: you’re live in front of 100s of people…

8. Social Media Agency For Online Businesses

The next 4 are all agency models you turbocharge with AI. Proven, profitable businesses you can now scale without a giant team.

First up, the social media agency. Every business needs help here. Hint: every single one.

• Buyer: Online business owners running a coaching program or paid community whose social presence is weak.

• Outcome: Turn 1 business idea into 30 days of posts using Claude. One monthly meeting, AI records the call, turns notes into 30 posts, and hooks into Canva for carousels.

• Distribution: Instagram DMs (genuinely reach out to owners whose IG is struggling), cold email (works great once you have case studies, lead with results), and local businesses.

• Medium leverage, fast cash: It’s a repeatable process you install in any business, and every business on earth needs marketing.

The honest catch: clients need babysitting. Onboarding, expectations, recurring tweaks. That’s what makes $100K/month harder than it looks.

Saying social media agencies are dead is just false.

Every business needs help with their content. Every single one.

9. AEO Agency For SaaS And Local Businesses

AEO is Answer Engine Optimization. Getting businesses to show up in Google AND AI search results. I’m literally a buyer here… I recently hired an AEO agency for Blotato.

• Buyer: SaaS businesses and local businesses. Two valid buyers.

• Outcome: Show up in the top 5 Google and AI search results. For SaaS, rank for specific keywords. For local, show up when someone searches “plumbers in Salt Lake City.” Google is still king for local.

• Distribution: LinkedIn shines. It’s easy to make a LinkedIn post go viral by claiming you have the AEO secrets. Cold email and content work too.

• The money: Agencies charge thousands per month for this. If it delivers $10,000 in ROI, why wouldn’t a business pay $9,000 for it? The math works.

Honest caveat: it’s harder than the social media agency. Non-technical people get lost by step 2, and results take longer to show. My spicy take: a lot of “AEO” is just SEO with a fresh name, and that’s fine. It’s still important.

Build a base set of Claude SEO skills.

Fine-tune a version per client, plug into WordPress or Webflow, repeat.

10. Vibe Code Agency For Non-Technical Founders

There’s an army of non-technical founders stuck at the last mile. They have an 80% app and can’t finish it.

• Buyer: Non-technical founders who built an MVP with AI but can’t handle auth, payments, security, or deployment. Also companies that want a fast, reliable MVP to test an idea.

• Outcome: Finish, fix, and launch their half-built app. You’re selling speed, reliability, and security. Speed matters most, because most startups die before they ship.

• Distribution: Join AI and Skool communities, find people stuck, and be GENUINELY helpful. Write the best 3-paragraph answer to their question and they’ll DM you “can I pay you to finish this?” This happens constantly.

• Higher leverage: It doesn’t have to be only you. Your team can handle fulfillment.

The honest truth: if you have zero technical background, this one is HARD. You need to learn deployment, databases, and Stripe first, and that takes real time. You’ll probably freak out every time you see an error message.

A founder who sweated 8 hours to get to 80% is desperate.

They just want someone to take it and finish it.

11. AI Automation Agency For Service Businesses

Easiest agency to land a first client. Also my lowest-leverage technical model. Both are true.

Let’s walk through the AI Automation Agency for one sample target buyer:

• Buyer: Service businesses. Established, with revenue and existing processes for sales, support, and hiring, so there’s a lot to improve.

• Outcome: Audit what they do today, then automate intake, follow-ups, reporting, and admin work to save time and make money.

• Distribution: Upwork is underrated. I posted a test job there and rising to the top was not hard, because most applicants don’t even read the post.

• Why it’s easy to start: Owners on Upwork decide fast. “We have this need, you seem credible, you’re hired.”

The honest reality: every business is messy. Tools don’t talk to each other, data is trapped in silos, and owners aren’t systems people. I’ve never seen 1 person run a $100K/month automation agency without a team. You’re expected to perform miracles.

Automation adds huge leverage to YOUR work.

But as an AGENCY business model, it’s surprisingly low leverage. Know that going in.

Phrased differently: you couldn’t pay me to start an AI Automation Agency today, because I have other better options.

12. Vibe Code Your Own Product

This is the one I actually do. The ONLY one I do.

Highest leverage, but longest road to your first dollar.

• Buyer: Whoever your product serves. Mine is Blotato, built for content creators.

• Outcome: A recurring-subscription SaaS. Blotato’s promise: create 1 week of content in 10 minutes. Your north star question: what’s the ONE thing a user should walk away with in their first 90 seconds?

• Distribution: TikTok shines. It’s the most generous with views and the highest chance of going viral. When a product goes viral on TikTok, you could bring in a million dollars overnight. Youtube tutorials are also great, FB/IG also great because of DM automations.

• The highest leverage there is: Adding another Blotato user takes zero extra work from me. No onboarding, no sales call, nothing. And if your SaaS product is at $1M/year in revenue, a buyer paying $5M to $20M for the company is reasonable.

But be honest with yourself. This is the HARDEST path and the longest to your first dollar. The brutal part of SaaS is getting to your first $10,000 in recurring monthly revenue. I don’t recommend this if you need money right now. Every other path is faster.

This is the path I do.

It’s the hardest, the slowest to pay, and the one almost nobody will stick with.

But I’ve done it before. I’m introverted and love not having sales calls, consulting clients, or a big team.

RECAP

Here’s the crazy part I saved for the end…

During my livestream, I ran live polls:

“Type 1 if you want my community, 2 for my course, 3 for coaching with me.” Hundreds of you responded.

Then I told you… I don’t do ANY of paths 1 through 11.

I just proved the demand is real, even for the things I’d never sell. So if YOU offer these, the demand is sitting right there.

Quick map of where everything lands:

1/ Fastest cash: 1:1 coaching. People make impulsive decisions, money hits fast. Low leverage.

2/ Best-of-both: the 4 agency models. Cold-outreach cash plus a process you can productize.

3/ Highest leverage, slowest road: courses, a paid community, and your own product. Build the asset, get paid for years.

4/ The one I do: vibe coding my own product. Few of you will stick with it, and that’s exactly why it’s worth so much.

Can you do multiple? Yes, and I encourage a little experimentation at the start. But experimenting means 6 months each, not a half-hearted part-time 2 weeks. Eventually, pick ONE. It’s hard to scale past $100K/month when you’re split across 5 things.

And if you have NO IDEA where to start, I always say posting content is the safest bet because it preserves ALL your options.

And don’t agonize over today’s choice. You can always change.

The whole point is to acquire skills and build optionality.

Pick the path you genuinely enjoy.

Because if you enjoy it, you stay consistent, and consistency is how you build something real.

Remember, think in YEARS-long timelines.

Full livestream here:

FAQ

Do I need to be technical to make money with Claude? No. Paths 1 through 9 are about helping people USE AI, not building software. Only the vibe code agency and your own product lean technical, and you can start non-technical and partner with a builder.

What’s the fastest way to my first dollar? 1:1 coaching. You can find a paying client in your existing network or DMs this week. The tradeoff is low leverage, since you’re trading your time for money.

Is a paid community worth it with only a few hundred members? Yes. 200 members at $50/month is $10,000/month recurring. You only need 200 people out of millions of AI learners, but expect ongoing work to keep them from churning.

Which path is the hardest? Vibe coding your own SaaS product. It’s the highest leverage but the longest road to your first dollar, and the hardest part is getting to your first $1,000 in recurring payments.

Should I run several income streams at once? Not at the start. Experiment with a few for 3 to 6 months each, then commit to ONE. Spreading across too many is why most people stall before $100K/month.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

• You have full permission to repurpose all my education! Today’s newsletter contains thousands of dollars worth of training. It may take you hours to get through it, but it’s 1000% worth it. And it’s all yours for free.

• NEW Claude Skills: 100 Viral Hooks and Repurpose Engine when you start a free trial of Blotato

TLDR… loop engineering means designing the LOOP that runs an AI (do the work, check it, repeat) instead of typing each prompt yourself. In Claude Code, the /goal command gives the AI a goal it can check and stop on, and a routine makes it run on a schedule by itself. Put those 2 together and you’ve built your first autonomous agent.

You do NOT need to be a coder for any of this.

We’re connecting 3 ideas:

1. What “loop engineering” actually means

2. The Claude Code /goal command, with 1 example you can copy.

3. Build your first autonomous agent with a Claude Code routine.

Who this is for:

People who keep seeing “loop engineering” everywhere in 2026 and want to understand autonomous agents and build their first.

And you do NOT need a business or an audience for this. If you’ve got an inbox you’re drowning in or a report you dread every month, it’s for you.

Prerequisites

First, what’s Claude Code? It’s the hands-on version of Claude that works on your real files and tools, not just chats. It’s different from the Claude chat site you might already use.

Claude Code Setup Guide.

What you need before you start:

• A Claude account. The /goal command and routines are paid features, so a free account won’t show them.

• For Section 2 (/goal): the Claude Code app installed on your computer.

• For Section 3 (routine): just a browser, plus any tools you want it to use (like Gmail) connected first under Connectors.

Read it once.

Then follow the Cheatsheet below :)

If you prefer to follow along visually, here’s my raw livestream, it’s a bit messy.

1. What “Loop Engineering” Actually Means

Start with the word everyone got tired of: prompt engineering.

That’s writing 1 good message to an AI, one at a time.

You type. It answers. You read it. You type the next thing.

See the problem?

YOU are the loop!

You’re the one checking the work and deciding the next step, every single time.

Loop engineering is stepping out of that, and building a little system that does the prompting FOR you.

The clearest definition going around:

Loop engineering is building a system that prompts your AI on a schedule and against a goal, instead of you typing each prompt yourself.

Your job changes from doing every step to designing the machine that does the steps.

Where the word came from:

AI “loops” (do a step, check the result, do the next step) have been around for YEARS. I’d often append this to a complex prompt: “Don’t stop until…”

Peter Steinberger said the line that popularized it:

Addy Osmani gave it the name “loop engineering” in June 2026.

A loop has 6 parts:

1. Automations: a timer that starts the job on its own, so you don’t push a button.

2. Worktrees: separate work areas, so 2 AI helpers don’t trip over each other.

3. Skills: saved instructions that teach the AI how YOU do a task, so it stops guessing.

4. Connectors: plug-ins to the real tools you already use (Gmail, Slack, Blotato).

5. Sub-agents: 1 AI does the work, a DIFFERENT one checks it. No grading its own homework.

6. Memory: a notes file that lives outside the chat and remembers what’s done and what’s next.

Key Concepts

Every loop has a DOER (the AI that makes stuff) and a CHECKER (the part that decides if it’s good).

The AI was never the hard part.

The CHECKER is.

If you can’t say what “done” looks like, you don’t have a loop. You have a wish.

So the real skill is simple to say and hard to do: write the check.

Here’s the good part: all 6 are already built into Claude Code.

You don’t assemble them.

You just need to know how to switch them on.

For your FIRST loop, 2 do the heavy lifting: a goal the AI can check (/goal), and a timer to run it on its own (a routine).

So that’s where we start.

Cheatsheet 1

Write 1 sentence for a task you repeat every week.

Starts when: __________ (a time of day? an email arriving?)

Done looks like: __________ (email labeled or archived)

The check: __________ (how the AI PROVES it’s done)

Keep that sentence.

You’ll use it in steps 2 and 3.

2. The Claude Code /goal Command (ADVANCED)

Heads up: this one needs the Claude Code app on your computer, so it’s a notch more technical.

Claude Code Setup Guide.

New to all this? Skim it for the idea and jump to Section 3, the routine, which runs in just your browser.

/goal is the “goal it can check” piece.

You give Claude a finish line. It keeps working on its own until it crosses it.

After each round, a small fast AI checks: is the goal met, yes or no?

A “no” and Claude keeps going.

A “yes” and it stops on its own.

You need Claude Code version 2.1.139 or newer.

Now, open Claude Code in a terminal.

How you type it:

/goal <what “done” looks like>

Simple Example

Say your Downloads folder is a mess, hundreds of random files, and you want it sorted.

/goal sort every file in my Downloads folder into subfolders by type (Images, Documents, Videos, Other), keep going until no files are left, do not delete anything, and stop after 30 turns.

Claude moves one file, checks what’s left, moves the next, and repeats.

It stops on its OWN when the folder is sorted, not before.

3 things make this a GOOD finish line:

• A clear end state: “no loose left” (not “tidy it up”).

• A check it can run: count the files in the folder.

• A guardrail: “do not delete anything” and “stop after 30 turns.”

That last one matters.

Always cap the loop so it can’t run forever.

Intermediate Level Examples

1. Standardizing a messy folder of files until every name matches one format:

/goal rename every file in the “receipts” folder to

“receipt_NN.pdf” in date order, keep going until no file

still has its original name, do not delete or move

anything, and stop after 25 turns

End state it can verify: zero files left with old-style name.

2. Categorizing a spreadsheet until no row is missing a label:

/goal fill in the “Category” column for every row in

expenses.csv using Food, Travel, Bills, Shopping, or

Other, keep going until no row is blank, do not change

any other column, and stop after 30 turns

End state it can verify: zero blank cells in the Category column.

3. Working through a stack of documents until none are left (the “empty the queue” pattern):

/goal write a 5-line plain-English summary of every

PDF in the “reports” folder into summaries.md, keep

going until every PDF has a summary, do not edit the

PDFs, and stop after 40 turns

End state it can verify: every PDF has a matching summary.

4. Cleaning up a batch of captions until they all fit your rules (if you make content):

/goal rewrite every caption in captions.txt to under

150 characters with no hashtags, keep going until all of

them are done, do not touch any other file, and stop

after 30 turns

End state it can verify: zero captions over 150 characters or containing a hashtag.

5. Turning a list of rough ideas into hooks until the list is done:

/goal turn each of the 20 rough ideas in ideas.txt

into a hook under 10 words, keep going until all 20 are

done, do not change any other file, and stop after 30

turns

End state it can verify: all 20 ideas rewritten as hooks <10 words.

Cheatsheet 2

Turn your Cheatsheet 1 sentence into a real /goal finish line.

/goal __________ (what done looks like), checked by __________ (how it proves it), and stop after __________ turns.

Try it on something small first.

Watch it loop.

3. Build Your First Autonomous Agent With a Routine

/goal runs while YOU sit there.

A routine adds the timer, so the loop runs without you.

A timer plus a goal it can check equals a real loop.

Now you have both halves.

What a Claude Code Routine is:

A saved AI job that runs by itself on a schedule, in the cloud, even when your laptop is closed.

You set it up in your browser at claude.ai/code/routines.

No terminal needed.

Heads up: routines are a paid feature, so you need a paid Claude plan to see them.

Start here:

1. Go to claude.ai/code/routines

2. Click the “New routine” button.

3. Name it “Daily Email Cleanup”.

4. In the “Instructions” box, say plainly what it should do: “Every morning, read my unread email and post the 3 most important ones to my Slack DM Sabrina Ramonov, 1 line each. Do NOT reply to anything.”

5. “Connectors” lists the apps already linked to your Claude (like Gmail and Slack). Keep ONLY the ones it needs, and remove the rest. If the tool you want isn’t there, connect it first in Claude settings.

6. Under “Select a trigger”, choose “Daily”, then 9am.

7. Click the “Create” button.

That’s a real autonomous agent.

Every morning at 8am it runs your loop, leaving you a tidy summary.

Intermediate Level Examples

• Buried under 10,000 unread emails? “Sort my inbox, label what matters, and trash the junk.”

• Dreading that monthly report? “Every month, pull the numbers and write the first draft of my report.”

• Too much news to keep up with? “Every Monday, read the updates on [your topic] and hand me 1 short brief, so I stop scrolling.”

Pick the one that you resonate with.

Make your first routine.

Next Level

Here’s my real example of a customer support agent routine, “Daily Support Ticket Cleanup”:

• The Skill: in the “Instructions” box I write “Run the ‘cleanup-tickets’ skill on all open customer support tickets, using Intercom API.” That cleanup-tickets skill is my saved know-how, so it handles tickets MY way every time.

• The Repository: I have a github project called help.blotato.com which serves as the memory, or notepad.

• The Connectors: in the “Connectors” tab I turned on Intercom and Slack, so it reads real support tickets and posts updates to my real Slack channel.

• The Schedule: it runs every day at 8am on its own.

And here’s my CHECKER agent, which independently reviews the work of the primary agent closing the tickets.

If the checker decides an auto-closed support ticket should remain open for human escalation, then it will re-open it.

Next Level: Integrate Any Tool with API keys

Connectors only cover tools that already have a plug-in.

But a routine also lets you add your own secret keys as “environment variables.”

Translation: paste in the API key for ANY tool, and your AI agent routine can use that tool too.

For example, in the support cleanup routine above, I created an environment called “Support” and added my Intercom API token. This allows my agent to use the full Intercom API, not just the limited tool calls available in Intercom MCP.

A few things before you trust it:

• Start read-only. Have it SUMMARIZE for a few days before you let it change or send anything.

• Set the limits in plain words, like “do not reply” or “do not delete.”

• Routine runs use the same plan as your normal Claude, and there’s a daily limit in settings.

Cheatsheet 3

Ship 1 tiny routine this week.

Runs when: __________ (every morning? when an email arrives?)

What it does: __________

Tools it needs (connectors & API): __________

The 1 limit: __________ (“don’t reply”, “don’t delete”)

Start read-only.

Let it SUMMARIZE for a few days before you let it edit anything.

RECAP

All 3 pieces we’ve covered:

1. Loop engineering is the shift from typing prompts to designing the loop. Osmani’s definition has 6 parts: automations, worktrees, skills, connectors, sub-agents, and memory.

2. /goal is the “goal it can check” part. Claude works to a finish line and stops itself when it gets there.

3. A routine is the timer (plus the skill, the connectors, and your API keys). It runs the loop in the cloud without you. That’s your first autonomous agent.

The hard part is saying clearly what “done” looks like, and capping the loop so it can’t run away.

This is why my personal formula is always:

FAQ

What is loop engineering in simple terms?

It’s designing a small system that prompts an AI on a schedule and against a goal, instead of you typing each prompt. The loop is: do the work, check it, repeat until done. Your role moves from doing every step to designing the system.

Do I need to know how to code to do this?

No. The routine in step 3 is set up in your browser with plain instructions. Only the /goal command in step 2 needs the Claude Code app, so it’s marked ADVANCED.

What does the Claude Code /goal command do?

It gives the AI a finish line and keeps it working until that finish line is met. A small fast AI checks after each round whether it’s done. You need Claude Code version 2.1.139 or newer. Type /goal clear to stop early.

How do I build an autonomous agent in Claude Code?

Create a routine at claude.ai/code/routines. Write plain instructions, turn on the connectors it needs, pick a schedule, and click “Create”. It then runs on its own in the cloud.

Can a routine use tools that don’t have a connector?

Yes. Add that tool’s API key as an environment variable in the routine, and it can reach almost any online service. This part is more advanced.

Is it safe to let a routine run on its own?

Start read-only. Have it summarize for a few days, set plain limits like “do not reply” or “do not delete,” and watch the first few runs before you trust it with anything that sends or changes things.