Look at the line the arrow points to: git main | up 1m 46s | tools 2.

Claude Code doesn’t come with that line. I typed one sentence in plain English, and Claude built it.

It shows my git branch, how long my session has been running, and how many actions Claude has taken.

By the end of this issue, you’ll have yours.

It’s called a mod. Since October 1, anyone can build one.

A mod rewrites Claude Code itself. It can add a live panel beside your chat, hold a dangerous command before it runs, or hide your API keys before Claude reads them.

Three things in the docs stopped me cold:

1. Anthropic already rebuilt one of its own features as a mod. You can switch it off and use yours.

2. The mods Claude builds for you get deleted on a timer. Most people won’t notice until theirs is gone.

3. A stranger’s mod can approve Claude’s actions without asking you.

Below, I’ll show you which feature, how to save your mods, and the one-minute check that exposes what any mod can do.

Paid members also get The Mod Vault: 30 copy-paste mod prompts, a safety checklist, and a fix card for when a mod won’t show up. It’s open for one week.

Why This One Matters

Until last week, you could teach Claude new tricks (skills), give it new tools (MCP servers), and run scripts on certain events (hooks).

All of that worked from the outside.

Claude Code itself stayed the same app for everyone.

Mods run inside it.

They can redraw what you see, step into a tool call before it happens, and add commands that run instantly, even while Claude is busy.

I read every page of the official docs and the most popular mods guide going around this week.

The guide was good, but it skipped the three things that bite beginners: why your mod vanishes, how to check a mod before installing it, and when you should build a skill instead.

This issue covers all three.

Who You’ll Be After This Issue

→ You’ll see which mods Anthropic already put on your machine.

→ You’ll know in 30 seconds whether your idea should be a mod at all.

→ You’ll build your first mod in about five minutes by describing it.

→ You’ll keep it for good, before Claude Code deletes it.

→ You’ll check any stranger’s mod before it touches your files.

→ You’ll switch any mod off in seconds.

You don’t need to know JavaScript. You need Claude Code and one evening.

I spent my whole day at this desk, testing every step below until it worked, so you won’t hit a single error. If that effort saves you an evening, I think I’ve earned a coffee from you. ☕

Buy me a Coffee

Everything Below Is for Paid Members

→ The hidden mod Anthropic is quietly testing, and the command to try it.

→ The 30-second test that tells you if your idea should be a mod or something simpler.

→ The 5-minute first build. One prompt, one mod, live on your screen.

→ The keep-it prompt that stops your mods from being deleted.

→ The audit prompt. Claude reads your past sessions and tells you which mods you need.

→ The safety check that lists everything a mod can touch, with a practice run on Anthropic’s own sample mod.

→ The off switch for one mod, or all of them.

→ The Mod Vault: 30 mod prompts, the safety checklist, and the fix card.

Read more

“Don’t touch that file.”

“Check your work.”

“Use my format.”

And Claude still gets it wrong sometimes.

The problem is where you write your rules.

Most people write all their rules in one place: a notes file called CLAUDE.md, or the chat.

Anthropic’s own help pages say a rule written there is “a request, not a guarantee.”

Claude reads it.

Most of the time it follows it. Sometimes it doesn’t.

Claude Code has three better places for your rules:

1. Skills teach Claude how to do a task.

2. Hooks make something happen every time.

3. Gates stop Claude from doing something.

When each rule is in the right place, Claude follows it.

Skills vs hooks vs gates, in plain English

A skill is a saved instruction, like a checklist. Claude decides when to use it, so it works most of the time. Use it for tasks that need thinking.

A hook is an automatic action. It runs every time, at a set moment, like an alert when Claude needs you.

A gate is a lock. There are two kinds:

1. Hard gate: blocks an action completely, like “never change this folder.” Claude can’t skip it.

2. Smart gate: checks Claude’s work before it can say “done.” No proof means Claude keeps working.

Read the full details here:

I made a one-pager comparison of these; save it now and use it later…

Where does each rule go?

Ask yourself three questions:

1. Should this never happen? Use a gate. Example: “Never change my published posts.”

2. Should this happen every time, with no thinking? Use a hook. Example: “Tell me when you need me.”

3. Does Claude need to think about how to do it? Use a skill. Example: “Review my draft in my style.”

If a rule is none of these, it’s just a fact about your project.

Keep it in CLAUDE.md.

Step 0: Set up Claude Code

Already have Claude Code? Skip to Build 1.

Two words to know: The terminal is a window where you type commands.

A project folder is a normal folder where we’ll save everything.

Install Claude Code

On a Mac:

• Press Cmd + Space.

• Type Terminal and press Enter.

• Copy this line, paste it in the window, and press Enter:

curl -fsSL https://claude.ai/install.sh | bash

• Wait about a minute until the text stops.

• Close the window. Open Terminal again.

• Type claude --version and press Enter.

• If you see a number, it worked.

On Windows:

• Click the Start button.

• Type PowerShell and press Enter.

• Copy this line, paste it in the window, and press Enter:

irm https://claude.ai/install.ps1 | iex

• Wait until the text stops.

• Close the window. Open PowerShell again.

• Type claude --version and press Enter.

• If you see a number, it worked.

Make your folder and sign in

• Type mkdir my-claude-setup and press Enter. This makes a new folder.

• Type cd my-claude-setup and press Enter. This opens the folder.

• Type claude and press Enter.

• A browser window opens. Sign in to your Claude account.

• Go back to the terminal.

• Claude asks if you trust this folder. Choose Yes.

You’ll do these three things a lot:

• Say yes: when Claude asks to make a file, use the arrow keys to pick Yes, then press Enter.

• Restart: type /exit and press Enter. Then type claude and press Enter.

• Come back later: open the terminal, type cd my-claude-setup, press Enter, then type claude and press Enter.

Build 1: Your first skill (5 minutes)

This skill reviews your writing before you publish it.

• Copy this prompt, paste it into Claude Code, and press Enter:

Create a skill file at .claude/skills/review-draft/SKILL.md with exactly this content:

---
name: review-draft
description: Review a written draft before it is published. Use when the user asks to review, check, edit, or polish a newsletter, post, email, or article.
---

Review the draft in this order:

1. Find every claim that needs a source. List each one with the source it needs.
2. Flag any sentence over 25 words. Suggest a shorter version.
3. Flag any placeholder text such as TODO, TBD, [INSERT], or lorem ipsum.
4. Check that every link is written as a full URL.
5. Report what you found as a numbered list. Do not rewrite the draft unless asked.

• Claude asks to make the file. Pick Yes.

• Type /skills and press Enter.

• Check that you see review-draft in the list.

• Press Esc to close the list.

• Type /review-draft, paste some of your writing, and press Enter.

Now you can also just say “review my post,” and Claude knows to use this skill.

Build 2: Your first hook (5 minutes)

This hook shows an alert on your screen when Claude is waiting for you, so you can stop checking the window.

Mac users, do this first. Macs block these alerts until you allow them once.

• Open a new Terminal window.

• Paste this line and press Enter:

osascript -e 'display notification "Test" with title "Claude Code"'

• If an alert appears in the top right corner, you’re ready.

• If nothing appears, open System Settings, click Notifications, find Script Editor, and turn on Allow Notifications. Run the line again.

Now add the hook:

• Go back to Claude Code.

• Paste the prompt for your computer into Claude Code and press Enter.

Mac:

Add a Notification hook to .claude/settings.json in this project. If the file already exists, keep everything in it and add this inside "hooks":

"Notification": [
  {
    "matcher": "",
    "hooks": [
      {
        "type": "command",
        "command": "osascript -e 'display notification \"Claude Code needs your attention\" with title \"Claude Code\"'"
      }
    ]
  }
]

Show me the final file before saving it.

I found an error here:

Here is the fix for this error if you also find this kind of error:

• In the same Terminal window, type cd my-claude-setup and press Enter.

• Type claude and press Enter.

• Wait until Claude Code opens. You’ll see Claude’s input box instead of a line ending in %.

• Now paste the prompt (Mac: same one as above) and press Enter.

Windows:

Add a Notification hook to .claude/settings.json in this project. If the file already exists, keep everything in it and add this inside "hooks":

"Notification": [
  {
    "matcher": "",
    "hooks": [
      {
        "type": "command",
        "command": "powershell.exe -Command \"[System.Reflection.Assembly]::LoadWithPartialName('System.Windows.Forms'); [System.Windows.Forms.MessageBox]::Show('Claude Code needs your attention', 'Claude Code')\""
      }
    ]
  }
]

Show me the final file before saving it.

• Pick Yes.

• Restart Claude Code (to restart it, type /exit and then again type claude

• Type /hooks and press Enter.

• Check that you see your hook under Notification, then press Esc.

Test it:

• Look at the bottom of the Claude Code window. If it doesn’t say manual mode, press Shift + Tab until it shows manual mode on. In this mode, Claude asks you before every change.

• Ask Claude: “Create a file called hello.md with the word Hello in it.”

• Claude stops and asks for permission. Don’t answer yet.

• Click on a different window and wait about ten seconds.

• The alert appears. Go back and pick Yes.

Still no alert? Ask Claude: “Show me my .claude/settings.json file.” Check that the Notification part is inside “hooks”. If it isn’t, paste the prompt again.

On Windows, the alert is a pop-up box that can open behind other windows. Check your taskbar.

If it still doesn’t work after this, send the screenshot. Tell me whether you’re on Mac or Windows, and whether you used the terminal or the Claude desktop app, so I can find the exact cause.

Build 3: Your first hard gate (3 minutes)

This gate locks a folder. Claude can never change the files inside it. Use it for work that’s already published.

• Ask Claude: “Create a folder called published with a file called live-post.md that says This post is live.”

• Pick Yes.

• Copy this prompt, paste it into Claude Code, and press Enter:

Add these permission deny rules to .claude/settings.json in this project. Keep the existing hooks exactly as they are.

"permissions": {
  "deny": [
    "Read(./.env)",
    "Edit(./published/**)",
    "Bash(rm *)"
  ]
}

Show me the final file before saving it.

4. Pick Yes.

5. Restart Claude Code.

• Type /permissions and press Enter.

• Check that you see all three rules.

• Press Esc to close the list.

• Ask Claude: “Change published/live-post.md to say This post is updated.”

• Claude Code says no.

What the three rules do:

• Read(./.env) keeps Claude out of the file where apps store passwords.

• Edit(./published/**) locks your published folder.

• Bash(rm *) stops the normal delete command.

4 mistakes that break a gate

These come from Anthropic’s help pages: https://code.claude.com/docs/en/hooks

1. A typo turns the gate off. Claude Code shows a small error and keeps going. After you add a gate, always check it with /hooks or /permissions.

2. A slow gate gets skipped. If a gate takes too long, Claude Code lets the action happen anyway.

3. A gate only blocks the exact command. Bash(rm *) stops rm file.txt. It doesn’t stop /bin/rm file.txt.

4. A smart gate can get stuck. If Claude can never pass the check, it keeps trying. Claude Code stops it after eight tries. So make your checks possible to pass.

One prompt to sort your old rules

Already have a long CLAUDE.md file? Paste this into Claude Code:

Read my CLAUDE.md file. Sort every rule into one of four groups:

1. GATE: anything that must never happen. Write the matching permission deny rule for .claude/settings.json.
2. HOOK: anything that must happen every time at a set moment, with no judgment. Name the hook event it belongs to.
3. SKILL: anything that needs judgment or is a multi-step procedure. Suggest a skill name and a one-line description.
4. KEEP: facts about the project that Claude should always know.

Show the result as a table with columns: Rule, Group, What to write.
Do not change any files yet. Wait for my approval.

Don’t build everything tonight. Think of the last mistake Claude made. Ask the three questions. Build that one fix.

Quick recap

• [ ] Skills teach Claude. Claude decides when to use them.

• [ ] Hooks run by themselves, every time.

• [ ] Gates stop Claude. Claude can’t skip them.

• [ ] CLAUDE.md is for facts about your project.

• [ ] Test every new gate by trying to break it.

Know someone whose Claude keeps ignoring their rules? Send them this issue:

Share

Your gift, as promised

The Three-Layer Starter Pack.

One skill, one hook, two gates, and where each file goes.

→ Click here to grab your gift

We tested it in public. Now use it in private.

Make Yourself Irreplaceable with AI. LLMs don’t think; you do.

PS: Reply and tell me the one rule Claude keeps breaking for you.

I’ll tell you if it should be a skill, a hook, or a gate.

In 2025, it made $85.5 million.

Nothing about the island changed between those two numbers.

The world just started caring about two letters it happened to own.

People who registered .ai names early, before “AI” was on every billboard, sold some of them for prices that sound made up.

→ Cloud.ai went for $600,000.

→ Law.ai went for $350,000.

Get the full details here.

On September 29, 2026, the US government officially stopped saying “AI.” It now says “SI,” short for Super Intelligence.

Read the full article here

Within days, speculators were buying up .si domains.

I want you moving this week too, before the obvious opportunities are gone and making money with SI becomes difficult and saturated.

I made a one-pager showing every layer of SI. Save it for later.

What You’ll Walk Away With

By the end of this issue, you will:

→ Know exactly what changed with “SI” and what didn’t.

→ Have 6 ways to make money from the rename, each with step-by-step instructions.

→ Have a copy-paste prompt for every step.

→ Have a free workflow that turns one article into a week of content.

→ Know which ideas are safe bets and which are gambles.

→ Get The SI Early Mover Kit, with every prompt template in one place.

The SI window just opened, and I'll be tracking every new opportunity as it lands. Subscribe so you hear about the next one first.

Subscribe now

Why I’m Sharing This

When the news broke, I shrugged.

It’s a new name for the same technology.

Then I remembered Anguilla.

Anguilla got rich from attention.

When millions of people start searching for a new word, whoever is already standing at the front of that search gets paid.

People started searching “SI” last week, and from what I can see, very few are making anything for those searches yet.

That gap won’t last.

I’m sending you this now, while it’s still early, because that’s what AI In Public is for. We test things in public so you can win in private.

I am sharing this because I want my audience to get ahead of everyone and benefit early before it gets buried.

So let's go into a deep dive of SI.

Know someone who should hear about SI before everyone else does? Send this their way.

Share

Workflow: The SI Content Engine

It turns one explainer into a full week of content, and every piece stays accurate because the facts live inside Claude.

Part 1: Set up the Project

• Open claude.ai.

• Click Projects in the sidebar.

• Click on New Project.

• Name it “SI Content Engine,” and click on Create Project.

• Paste the instructions below into the Project’s instructions box.

About me:
I'm Sara, a freelance content writer who helps small business owners understand new technology. My audience is non-technical founders in the US and UK. I write in short sentences, plain English, and a friendly, direct tone. I never use hype words like "revolutionary" or "game-changer."

Verified SI facts (use only these, never invent new ones):
1. On September 29, 2026, President Trump signed an executive order directing US federal agencies to use "Super Intelligence" (SI) instead of "Artificial Intelligence" (AI).
2. The order applies to the US executive branch. It does not apply to private companies or other countries.
3. The White House has 60 days to propose an official legal definition of SI.
4. The technology did not change. ChatGPT, Claude, and Gemini work exactly as before.

Rules for every piece:
1. Write for one platform at a time, in that platform's style.
2. Before giving me a draft, check it against the facts above and remove anything that isn't supported.
3. Check it against my tone rules and fix anything that breaks them.
4. Give me the final version only.

(Edit the “About me” lines so they describe you.)

• Save the Project.

Part 2: Run it every week (20 minutes)

• Open a new chat inside the Project.

• Paste the prompt below to write your "What is SI?" explainer.

Write a 600-word beginner explainer titled "What Is SI? The US Just Renamed AI, and Here's What It Means for You."

Use only the verified SI facts in this Project.

Structure it like this:
1. Open with what happened in 2 or 3 short sentences.
2. Explain what changed (the name) and what didn't (the technology).
3. Say who the order applies to, and who it doesn't.
4. End with 3 practical ways a regular person can use these tools this week.

Write at an 8th-grade reading level. Short sentences. No hype.

• Copy the result.

• Now paste the result along with the prompt below again.

The weekly prompt:

Turn the article above into one week of content:

1. A 60-second talking-head video script.
2. Five X posts, each under 280 characters, each with a different angle.
3. Three LinkedIn posts, 120 to 180 words each.
4. One Substack Note under 70 words.

Label each piece with the day I should post it, Monday to Sunday.

• Copy each piece into the platform it was written for.

• Schedule them across the week.

Next week, write a new explainer on a fresh SI angle, like the official definition when it lands, and run the same prompt again.

The setup is done, so it only takes 20 minutes.

You just built a week of content in 20 minutes. If that helped, pass it on to one creator friend.

Share

What’s Below the Paywall

6 benefits that you can take right now without any delay from SI and make a lot of money.

1. SI domains you can buy now and sell later

2. SI social handles before they’re gone

3. “What is SI?” content that rides the search wave

4. Proposal rewrites for government contractors

5. Freelance profiles that catch new SI buyers

6. Digital products built in an afternoon

Read more

Most people can’t.

In the 1880s, psychologist Hermann Ebbinghaus tested his own memory and found that more than half of new material was gone within an hour.

We save more bookmarks to make up for it, and then never open them.

This issue shows you how to set up NotebookLM, so it stores what you read and answers your questions from it.

It takes about 20 minutes.

Who You’ll Be After Reading This

→ Someone who can find any idea they’ve saved in seconds

→ Someone who remembers what they read a month ago

→ Someone who learns a new skill from their own sources

→ Someone who walks into every call already prepared

→ Someone who turns saved content into their next idea

What You’ll Get at the End

• The Second Brain Starter Kit

• 5 ready-made notebook blueprints

• 36 copy-paste prompts

• Customize prompts for every Studio output

• A 10-minute weekly routine

New here? I send hands-on AI guides like this three times a week. Subscribe to get every guide in your inbox

Subscribe now

What Is NotebookLM?

NotebookLM is a free AI research assistant from Google.

You upload your own material, and it answers questions using only that material.

Every answer comes with numbered citations.

Click one, and it jumps to the exact passage it used.

ChatGPT answers from everything it was trained on.

NotebookLM answers from your shelf.

You can upload PDFs, Google Docs, Slides, Sheets, Word files, website links, YouTube videos, audio, images, pasted text, and EPUB ebooks.

One update: Google renamed NotebookLM to Gemini Notebook on July 16, 2026. It’s the same product with the same notebooks. I’ll use the old name here because most people still search for it.

It can still misread a source.

If an answer is going into your work, click the citation and check it.

Read more

That was the deal for every founder who wanted their product to move on screen.

Pay a lot.

Wait a long time.

This week I opened Claude Opus 5.5, pasted one prompt, and got a launch video back in under 10 minutes without any third-party app.

Today I’m giving you the prompts.

Why I Started Doing This

I make my own brand videos.

Every intro, every product clip, every animated stat you’ve seen from AI In Public.

For a long time that meant one of two things.

Pay someone and wait.

Or open editing software at midnight and fight keyframes until 3 AM.

When Opus 5.5 dropped on September 22, my feed filled up with motion graphics videos.

Logo reveals.

Launch trailers.

Full music videos.

I didn’t believe half of them.

Then I saw this one.

Turn your sound on before you press play.

The creator says he knows nothing about motion design.

He gave Opus 5.5 one prompt.

It wrote the code that rendered every frame, and it composed the music from scratch.

That’s when I stopped scrolling and opened Claude.

I tested it myself every day for a week, and then I got a full 100+ prompt library, which I am going to share with you today

Who You Become After This Issue

You stop being the person who says “we should hire a video maker or video editor.”

You become the person who ships a launch video the same day the product goes live.

Who turns a boring stat into a clip people share.

Who has a brand that moves while competitors post static images.

With a single prompt.

What These 100+ Prompts Do For You

• You ship faster. A launch video stops being a 3-week project. It becomes something you make the morning of the launch.

• You keep the money. Every prompt replaces work you’d otherwise pay a designer for.

• You can test 10 versions. Change the colors, the copy, the pacing. Make five versions and post the best one. No designer is going to make you five versions for free.

• Your brand stays consistent. Swap in your colors and fonts once. Every video after that looks like it came from the same studio.

• You own it. The output is a file. Edit it, reuse it, remix it next month.

5 Prompts, 5 Videos

Every prompt below is copy-paste ready.

I’ve filled in real example brands so you can see exactly how it works.

Swap in your own name, colors, and copy.

Prompt 1: Personal Brand Video Prompt

Build a 24-second brand motion reel for AI In Public in Remotion (React), 1920×1080, 60fps (1440 frames). Every frame and every sound is made in code: no stock footage or stock music. Render it to MP4 (H.264, high quality, AAC audio).

BRAND
- Background: logo blue #3D40FE on every scene, as a living gradient: a darker #2A2DE0 at the edges, a brighter #4A4DFF glow in the center, and a slow drifting light spot so the background is never flat.
- Graphics on the blue: white as the main color, deep navy #0A0B3D for contrast and shadows, light blue #8C8EFF and ice #C9CAFF for secondary details.
- Glows on blue are white or ice (soft white bloom), not blue.
- Fonts: Archivo Black for headlines, Instrument Serif Italic for taglines, JetBrains Mono for technical labels.
- Logo: a white "A" mark with a 4-point sparkle (public/logo-mark.png, transparent). Use the sparkle as a recurring motif.
- Tagline: "learning AI, out loud."

SMOOTHNESS (no lag, no stutter)
- Render at 60fps. Every animation must be driven by the frame number and look identical at any frame (no randomness per frame, no Math.random).
- Animate only transform and opacity. Avoid CSS blur() filters on many elements at once; use a few pre-blurred glow layers instead.
- Add real motion blur to fast moves (warp, title slam, kinetic type, particles) with @remotion/motion-blur (CameraMotionBlur, ~5 samples), or with stretched trails.
- Camera shake uses smooth noise (sum of slow sine waves), never per-frame jitter.
- No hard jumps between scenes: every cut is a designed transition (zoom-through, whip-pan with blur, iris wipe, or a white flash) lasting 8–14 frames.
- Film grain is a small pre-made noise texture shifted each frame at ~5% opacity, not a full-screen SVG filter rebuilt every frame (keeps both render and playback smooth and the file small).
- Keep particle counts reasonable (≤1,400) and draw them in one SVG.

MUSIC & SOUND (suspenseful, generated in code)
- Write a Node script that synthesizes a 24-second WAV (44.1kHz stereo) with no external samples, then add it with <Audio src={staticFile("music.wav")} />.
- Track structure, synced to the scenes:
  - 0–3s: dark low drone (two detuned sine/saw waves around 55Hz with slow filter movement) plus a quiet ticking clock pulse. Tension.
  - 2–3s: a rising white-noise riser and pitch sweep into the warp, ending in a deep sub-boom impact at the title slam.
  - 3–19s: a heartbeat sub-bass pulse at 120 BPM, a ticking hi-hat, an eerie high string-like pad (slow vibrato), and short whoosh sounds on every scene transition.
  - Terminal scene: soft keyboard clicks synced to each typed character, plus a "confirm" blip on "✓ shipped".
  - 17.5–19s: a hard hit on each kinetic word (PROMPT / BUILD / SHIP / REPEAT), building intensity.
  - 19–22s: a long riser as the particles gather, then a big cinematic boom + shimmer when the logo locks.
  - 22–24s: the drone resolves to a calm, held chord and fades out.
- Every impact, whoosh and boom must land exactly on its visual frame. Master to about -14 LUFS with no clipping.

GLOBAL LAYERS (on top of every scene)
- HUD in white: corner brackets; top-left "AI IN PUBLIC · MOTION REEL · 2026"; top-right the scene label ("03 — EASING"); bottom-left a blinking ● REC with a running timecode; bottom-right "BUILDING IN PUBLIC"; a thin white glowing progress bar along the bottom.
- Finish: light film grain, a soft navy vignette at the edges, very faint scanlines.
- Camera shake: about 14px, smooth, decaying over 20 frames, on impact moments (synced with the booms and hits).

SCENES
1. Signal (0–2s): a white dot springs into the logo's sparkle. 16 white and ice rays burst out, a ring of rotating text circles it ("AI IN PUBLIC · LEARN · BUILD · SHARE"), a horizontal white lens flare, and an expanding ring. It ends by zooming 9× into the sparkle (zoom-through transition).

2. Warp (2–3s): hyperspace on blue. 260 white, ice and navy streaks rush toward the camera with motion blur and accelerate; a glowing white core grows into a white flash.

3. Title slam (3–5s): "AI IN PUBLIC" giant (230px, white). Letters spring up one by one with a skew, plus a navy/ice chromatic split that settles and glitches once. A light sweep crosses the letters, with a huge soft white sparkle rotating behind at low opacity. Below: "learn it · build it · share it" in serif italic.

4. Prompt terminal (5–8s): a deep navy terminal window with a white glow floats in 3D over a scrolling white grid. It types: ai › make a launch video for AI In Public. blue + white. go big. Log lines then appear (reading brand, studying reference, writing scenes, composing score) and a █ progress bar fills to 100%. Ends with a big "✓ shipped. in public."

5. Easing study (8–10s): "Six ways to get an idea out in public." Six rows (linear, ease-in-out, expo-out, back-out, elastic, bounce) tagged /IDEA /PROMPT /DRAFT /TEST /ITERATE /SHIP, each with its curve icon. White and navy dots race the same distance with motion blur and trails; a ring pops when each dot arrives.

6. Neural network (10–12.5s): a 3-6-8-6-3 network drawn layer by layer in white. Signal pulses travel along highlighted edges and nodes fire with a white glow, going from "INPUT · YOU" to "OUTPUT · EVERYONE". Caption: "Ideas compound when you share them."

7. Morph (12.5–14.5s): one glowing white shape morphs circle → triangle → sparkle (radial interpolation), with an orbit of matching mini-shapes and dashed rings. It resolves into the real logo with a spring and glow.

8. Truchet system (14.5–16s): a grid of white quarter-arc tiles reveals outward from the center. A wave sweeps left to right, rotating tiles 90° and turning them into dashed navy arcs.

9. Depth (16–17.5s): a rotating 3D torus of ~1,100 points with perspective depth, white points with navy accents, over a soft light spot.

10. Kinetic type (17.5–19s): PROMPT. → BUILD. → SHIP. → REPEAT. One word every 22 frames (at 60fps): letters spring in with motion blur, outlined echo copies stack above and below, the period is always navy, and every word lands with a camera hit and a sound hit.

11. Finale (19–24s): ~1,400 white particles sampled from the logo's pixels swirl in from across the frame and assemble the logo. It locks into the crisp logo with a white glow pulse, a shockwave ring and the big boom, then slides left. "AI IN PUBLIC" wipes in with a light sweep, then "learning AI, out loud." A footer line reads "MOTION REEL · 2026 | EVERY FRAME WRITTEN IN CODE | FOLLOW THE BUILD →". Slow fade to deep navy as the music resolves.

MOTION RULES
- Fast but smooth: every element enters with spring physics or an expo-out ease; nothing is static, nothing jumps.
- Glows come from drop-shadows and soft white radial gradients, never flat fills.
- Keep everything inside a 72px safe margin.
- Before the final render, check still frames from every scene and listen-check that sound hits match the visuals.

Prompt 2: Opus 5.5 Motion Designer

TITLE: "Claude Opus 5.5 — Motion Graphics Designer"

FORMAT: exactly 25 seconds, 16:9, 1920x1080, native 60fps with perfectly constant frame pacing. Ultra-smooth: no stutter, no frame skipping, no jitter, no speed jumps, no flicker, no warping or melting objects. One single continuous camera move from first frame to last, like a motion-control rig or a stabilized gimbal: no cuts at all. Every transition is a continuous morph or camera move. Natural, soft motion blur on fast moves. All text stays sharp, stable and correctly spelled.

VISUAL STYLE: premium dimensional motion design. Glossy, tactile 3D objects in a dark studio. Think of a high-end product launch film crossed with a motion design showreel. A deep charcoal-black studio (#121212) with a glossy reflective black floor that mirrors everything above it. One hero object: a glowing orange-red glass sphere (#FF4F1F), polished, with soft internal light, crisp specular highlights and real refraction. Type is thick, rounded, glossy warm-cream 3D lettering (#EDE9E0) with soft bevels and gentle reflections in the floor. Lighting: one large soft key light from above left, a warm orange rim light from behind, and the sphere itself casting a warm orange glow on the floor and nearby letters. Shallow depth of field, smooth bokeh, subtle film grain. Palette strictly charcoal, warm cream and orange-red. Clean, confident, minimal. Never cluttered.

MOTION RULES: every movement uses real physics: anticipation before it moves, a springy overshoot, and a soft settle. Bounces have proper squash and stretch on the sphere. Camera moves always ease in and ease out, never start or stop abruptly. Everything moves in sync with the music.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

0:00–0:02 | THE DROP (hook)
Low camera angle, close to the glossy black floor, slowly dollying forward. Empty dark studio. At 0:00.5, exactly on the first beat, the orange glass sphere drops from above into frame and hits the floor with a satisfying squash. A ring of warm orange light ripples outward across the reflective floor from the impact point like liquid. The sphere bounces once, smaller, and settles.

0:02–0:06 | MOTION
The sphere rolls forward and the camera tracks alongside it, smooth and steady. As it rolls, the six letters of the word "MOTION" rise out of the floor one per beat, each popping up with a spring overshoot and a tiny settle, glossy cream 3D letters reflected in the floor. The sphere hops onto the letters and bounces across their tops from left to right like a ball skipping across piano keys, one bounce per beat, each letter dipping slightly under the impact and springing back. On the final letter, the sphere launches high into the air.

0:06–0:10 | MORPH
The camera tilts up smoothly to follow the sphere into the air. At the top of its arc, the sphere hangs weightless and morphs in one fluid, liquid motion: sphere → soft rounded cube → glossy torus → long pill shape → back to a sphere, one shape per beat, each change smooth like molten glass with light refracting through it. Behind it, three frosted glass panels slide in at different depths, each with a thin glowing orange line drawing a different easing curve: a smooth S-curve, a spring curve with an overshoot, and a bouncing curve. Gentle parallax between the panels as the camera drifts.

0:10–0:14 | THE ORBIT
The frosted panels slide away. A ring of glossy cream 3D words forms around the floating sphere and orbits it smoothly: "EASING · SPRINGS · LOOPS · 3D · TYPE ·". The camera orbits around the whole arrangement in the opposite direction, creating rich, smooth parallax. The words catch the orange light as they pass in front of the sphere and glide into soft focus behind it. The sphere pulses gently with each beat.

0:14–0:18 | THE FLOW (build)
The camera pushes slowly through the gap in the ring toward the sphere. The sphere splits into thousands of tiny glowing orange spheres that pour out and flow like a liquid ribbon, swirling around the camera in a graceful spiral, rising and falling in waves that follow the music's build. The ribbon gets faster and brighter as the music rises, the floor reflections shimmering beneath it. In the last half second, every particle freezes mid-air for a single held breath.

0:18–0:22 | THE REVEAL (drop)
At 0:18.0, exactly on the drop, all the particles snap together in one satisfying motion and form the title in large glossy cream 3D lettering, centered and filling most of the frame:
Line 1: "Claude Opus 5.5"
Line 2: "Motion Graphics Designer"
The orange glass sphere reforms, arcs in on a spring, and lands as the period at the end of "Designer" with a squash and a ring of light rippling across the floor. The camera settles into a slow, steady push-in. The title reflects cleanly in the glossy floor. Hold calmly so it can be read.

0:22–0:25 | LOOP
The camera eases back. The title letters gently sink back down into the floor one by one, in reverse order, with soft springs. The sphere rises from the period, floats up and out of the top of the frame. The camera glides back to the exact low angle of the opening. The final frame is the empty, dark glossy studio floor, identical to frame one, so the video loops seamlessly.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

MUSIC: an original, catchy, upbeat instrumental, exactly 120 BPM, 4/4, in a bright major key. Style: modern indie-electronic pop with a feel-good, premium tech-commercial energy. Core elements: a bright, memorable plucked-synth hook melody (a short five-note phrase that repeats and sticks in your head), a bouncy rubbery bassline, crisp handclaps and stomps on beats 2 and 4, tight finger snaps, shimmering bell and glockenspiel accents, and a warm pad underneath.
Structure synced to the picture:
0:00–0:02: a single deep bass hit and a bright bell ping exactly when the sphere hits the floor, then the hook melody enters alone on the plucked synth.
0:02–0:06: claps, stomps and bass join; each bounce of the sphere on the letters lands on a note of the hook melody.
0:06–0:10: the groove fills out with snaps and a glockenspiel countermelody; soft swooshes under each shape morph.
0:10–0:14: full groove with the hook repeating, airy stereo whooshes panning with the orbiting words.
0:14–0:18: the build: a rising filter sweep, a snare roll speeding up, a rising synth riser, and half a second of total silence right before 0:18.
0:18–0:22: the drop: the full hook melody returns big and wide, with a punchy brass stab and a deep impact exactly as the title forms and the sphere lands as the period.
0:22–0:25: the music winds down to just the plucked hook, resolving on its final note, ending on a single clean bell ring that matches the opening ping so the music loops with the video.
No vocals. No lyrics. No harsh EDM drops, no aggressive bass, no dubstep. Bright, warm, catchy and polished.

NO logos, NO watermarks, NO extra text beyond what is specified.

Prompt 3: Western Civilization

TITLE: "Western Civilization — Every Age Drew a Bigger Circle"

FORMAT: exactly 28 seconds, 16:9, 1920x1080, native 60fps with perfectly constant frame pacing. Ultra-smooth: no stutter, no frame skipping, no jitter, no flicker, no warping or melting objects. No cuts at all: the entire film is ONE continuous infinite zoom. The camera pushes forward the whole time at a smooth, steady exponential rate, slowing gently as it approaches each circle, speeding up as it dives through it, and easing out on the other side into the next era. Every era is hidden inside a circle in the previous one. All text stays sharp, stable and correctly spelled.

CORE CONCEPT: the history of Western civilization told through circles. A compass circle in sand, a temple oculus, a rose window, a globe, a telescope lens, a steam flywheel, the Moon, a silicon wafer. The camera dives through each circle into the next age. A thin bone-white circle outline stays centered on screen through the whole film as a guide, growing larger with each era.

VISUAL STYLE: bold modern editorial collage, like a Swiss poster in motion. Strict three-color palette: International Klein Blue (#002FA7), bone white (#F1ECE2) and deep black (#0A0A0A). No other colors anywhere. Every object and surface is rendered as rich, realistic material (sand, marble, stained glass, brass, iron, lunar rock, silicon) but processed as a blue-and-white duotone with a fine halftone dot texture, like a high-end printed poster come to life. Real depth: objects sit at different distances and move with smooth parallax. Subtle paper-print grain. Soft vignette.

TYPOGRAPHY: one massive extra-condensed bold uppercase sans-serif typeface (tall, narrow, heavy, very tight tracking) for the single headline word in each era, set huge, filling most of the frame height, in bone white. A small clean monospace typeface for the year and place label in each era. No serif fonts anywhere. Headline words are placed in 3D space between the camera and the scene, so the camera flies past and through the letters with parallax as it zooms.

YEAR COUNTER: along the left edge, set vertically, a tall rolling odometer of year digits in the monospace typeface, bone white. As the camera dives through each circle, the digits spin smoothly and land on the next year, like a mechanical counter.

MOTION RULES: every move eases in and out. Objects drift with gentle parallax. Headline words slide in on the beat with a quick, clean ease-out and no bounce, hold, then rush past the camera as the zoom accelerates. Light beams and glints move slowly and smoothly. Everything locks to the music's beat.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

0:00–0:03 | MEASURE
Top-down view of fine pale sand in blue duotone. A bronze compass slowly draws a perfect circle in the sand, grains spilling softly along the line. The headline "MEASURE" slides in, huge, in bone white. Small monospace label: "ALEXANDRIA · c. 300 BC". The camera begins pushing in toward the center of the circle, and the sand inside it darkens into a deep opening.

0:03–0:06 | BUILD
The camera dives through the sand circle and emerges looking straight up inside a vast domed Roman temple, at the round oculus open to the sky. A single solid beam of light falls through the oculus past the camera, with dust floating inside it. Coffered ceiling squares recede toward the opening with strong perspective. "BUILD" slides in. Label: "ROME · AD 126". The camera rises up through the beam toward the oculus.

0:06–0:09 | BELIEVE
The camera passes through the oculus into open sky, and the sky closes in and becomes a giant circular rose window of a Gothic cathedral, seen from inside. Its intricate stone tracery and glass glow in luminous blue and white, light streaming through. "BELIEVE" slides in. Label: "PARIS · c. 1250". The camera pushes toward the bright center medallion.

0:09–0:12 | EXPLORE
The glowing center of the rose window turns into an antique hand-painted globe, rotating slowly on a brass stand, with old coastlines, sea monsters and sailing ships drawn on its surface. "EXPLORE" slides in. Label: "NUREMBERG · 1492". The globe turns and the camera glides toward a circular brass compass rose painted on its ocean.

0:12–0:15 | OBSERVE
The compass rose becomes the round glass lens at the end of an early brass telescope. The camera dives through the lens, the image refracting and bending beautifully at its edges, and emerges into a night sky looking at a cratered, detailed Moon. "OBSERVE" slides in. Label: "PADUA · 1609". The camera drifts toward the round edge of a crater.

0:15–0:18 | POWER
The crater rim becomes the rim of a huge iron flywheel on a steam engine, spinning steadily. Pistons pump in rhythm, soft white steam drifts through the frame, rivets catch the light. "POWER" slides in. Label: "BIRMINGHAM · 1769". The camera pushes toward the round hub at the center of the spinning flywheel.

0:18–0:21 | REACH
The spinning hub slows and becomes the full Moon again, now much closer, with a small lunar landing module resting on its gray surface and the black sky behind it. A single footprint sits in the dust in the foreground. "REACH" slides in. Label: "TRANQUILITY BASE · 1969". The camera rises and turns toward the round Earth hanging above the horizon.

0:21–0:24 | THINK
The Earth shrinks into a circular silicon wafer, shimmering with a dense grid of microchip patterns that catch light in shifting iridescent blue and white. Tiny circuit lines pulse with light. "THINK" slides in. Label: "SANTA CLARA · 1971". The camera pushes toward a single chip at the center and through its glowing grid.

0:24–0:28 | THE BIGGEST CIRCLE
The camera emerges and pulls back smoothly in one long, graceful move. Every circle from the film is revealed nested inside the next, in perfect concentric rings: sand circle, oculus, rose window, globe, lens, flywheel, Moon, wafer, glowing in blue and white. In the center, in the massive condensed typeface, the title builds:
Line 1: "WESTERN CIVILIZATION"
Line 2, small monospace beneath: "Every age drew a bigger circle."
The bone-white guide circle that has been on screen all along sweeps around the title once and closes. In the final half-second, the camera begins pushing back in toward the center, and the image dissolves into fine pale sand, matching the very first frame so the video loops seamlessly.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

MUSIC: an original, modern cinematic instrumental, exactly 120 BPM, 4/4, in a minor key that lifts to major for the title. Style: pulse-driven cinematic electronic, like a modern tech-documentary trailer: tense, hypnotic, building. Core elements: a deep, steady heartbeat-like sub pulse on every beat, a short repeating four-note piano motif played clean and bright, tight staccato strings that grow faster through the film, and a sucking reverse-swell whoosh every time the camera dives through a circle, landing on a crisp low impact on the downbeat as the next era appears.
Structure synced to the picture:
0:00–0:03: only the sub pulse and the piano motif, with the soft scrape of the compass in the sand.
0:03–0:06: staccato strings enter; the first dive lands on a deep impact.
0:06–0:12: a wordless choir pad swells under the strings; each dive gets its own whoosh and impact.
0:12–0:18: the strings double in speed, metallic percussion enters with the flywheel, the piano motif climbs higher.
0:18–0:21: everything drops to near-silence except the sub pulse and a single sustained high string note, for the stillness of the Moon.
0:21–0:24: a fast rising riser and snare roll build hard under the microchip scene, then half a second of total silence.
0:24–0:28: the big payoff: the full piano motif in a major key with strings, choir and a deep impact as the title builds, then everything fades to the lone sub pulse and the soft compass scrape from the opening, so the music loops with the video.
No vocals with words, no lyrics, no pop beats, no EDM drops. Modern, tense, emotional and polished.

NO logos, NO watermarks, NO real people's faces, NO parchment textures, NO serif fonts, NO red or gold colors, NO extra text beyond what is specified.

Prompt 4: OpenAI vs xAI - Wrestling - Fun Part

TITLE: "OPENAI vs. ANTHROPIC — The Main Event"

FORMAT: exactly 30 seconds, 16:9, 1920x1080, native 60fps with perfectly constant frame pacing. Ultra-smooth: no stutter, no frame skipping, no jitter, no flicker, no warping or melting bodies, no extra limbs or fingers. Cinematic camera moves only: smooth cranes, orbiting dolly shots and slow-motion replays, all easing in and out. Transitions between rounds use a quick, clean whip-pan or a bright flash on the bell. All on-screen text stays sharp, stable and correctly spelled.

STYLE: ultra-realistic, photorealistic cinematic live-broadcast look of a massive pro-wrestling event. A packed dark arena, a brightly lit wrestling ring at center with taut ropes and padded turnbuckles, haze drifting through sweeping spotlights, camera flashes across the crowd, pyrotechnic sparks, giant LED screens. Shallow depth of field, anamorphic lens flares, rich contrast, subtle film grain. Real skin texture, real sweat, real fabric, real physics.

THE WRESTLERS (no human faces anywhere):
WHITE CORNER — a muscular, athletic human body of a pro wrestler in a glossy white-and-black singlet, white boots and black kneepads. In place of a head, a large three-dimensional OpenAI logo emblem sits on the neck, about the size of a head, made of polished black metal with a soft white glow at its edges. It turns and tilts like a head would. Corner banner: "OPENAI".
COPPER CORNER — a muscular, athletic human body of a pro wrestler in a matte copper-orange and cream singlet, cream boots and copper kneepads. In place of a head, a large three-dimensional Anthropic logo emblem sits on the neck, about the size of a head, made of brushed cream ceramic with a warm copper glow at its edges. It turns and tilts like a head would. Corner banner: "ANTHROPIC".
REFEREE — a human body in a black-and-white striped referee shirt, with a plain glossy black sphere in place of a head.
Both wrestlers move with heavy, realistic physical weight: ropes stretch, the mat shakes and bounces, sweat flies on impact.

BROADCAST GRAPHICS: a bold lower-third banner for each round in heavy condensed uppercase sans-serif, white text on a slanted black bar with a copper-and-white split edge, sliding in smoothly from the left with a light glint across it. "ROUND #" sits above the main line in small text.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

0:00–0:03 | MAIN EVENT
The camera cranes down from high above the arena through the haze toward the ring as spotlights sweep the screaming crowd and pyro bursts on both sides of the entrance ramp. Huge text slams onto the screen: "OPENAI vs. ANTHROPIC", then beneath it, "THE MAIN EVENT".

0:03–0:07 | ROUND 1 — THE HOUSE
The OpenAI wrestler walks down the ramp through smoke and sparks, followed by a small crew of wrestlers in matching white team jackets with plain black sphere heads. They climb into the ring together and raise their arms to the crowd. Lower third: "ROUND 1" / "2015 — OPENAI IS FOUNDED".

0:07–0:10 | ROUND 2 — THE BREAKAWAY
In the ring, one crew member in a white team jacket steps forward. Its sphere head lights up and transforms into the Anthropic logo emblem. It rips off the white jacket in one dramatic motion, revealing the copper-and-cream singlet underneath, then steps through the ropes and walks across to the opposite corner. The OpenAI wrestler watches, arms crossed. The crowd gasps. Lower third: "ROUND 2" / "2021 — FORMER OPENAI LEADERS FOUND ANTHROPIC".

0:10–0:13 | ROUND 3 — THE BELT
The referee hands the OpenAI wrestler a gleaming gold championship belt. It lifts the belt overhead and spins under the spotlight as white confetti rains down and the crowd explodes. Lower third: "ROUND 3" / "NOV 2022 — CHATGPT LAUNCHES".

0:13–0:16 | ROUND 4 — THE CHALLENGER
Lights cut to black. A single warm spotlight hits the top of the ramp. The Anthropic wrestler makes a dramatic entrance through a wall of copper sparks and warm smoke, slapping hands with fans along the barrier as it walks. Lower third: "ROUND 4" / "MAR 2023 — CLAUDE LAUNCHES".

0:16–0:19 | ROUND 5 — THE SPONSORS
Back in the ring, the OpenAI wrestler flexes for the crowd, and blank colorful sponsor patches suddenly pop onto its singlet, one after another, with comedic little pops, until its whole body is covered in them like a racing car. A tiny blank billboard even pops up on its shoulder. It looks down at itself, surprised. Lower third: "ROUND 5" / "JAN 2026 — OPENAI ANNOUNCES ADS IN CHATGPT".

0:19–0:24 | ROUND 6 — THE SUPER BOWL
The Anthropic wrestler holds up a huge foam sign reading "NO ADS", tosses it aside, runs at the ropes, bounces off them and hits a flying dropkick. The OpenAI wrestler staggers, then fires back: it catches the Anthropic wrestler and lands a huge body slam that shakes the entire ring. The two trade heavy, ultra-realistic punches and kicks, sweat flying on each impact. Cut to a slow-motion replay of the body slam from a low angle, the mat bouncing and sweat and dust hanging in the air, the crowd on its feet. Lower third: "ROUND 6" / "FEB 2026 — THE SUPER BOWL AD WAR".

0:24–0:28 | THE VERDICT
Both wrestlers stagger to their feet, exhausted, facing each other in the center of the ring. The referee steps between them, grabs both their wrists and raises both arms at the same time. The crowd erupts as copper and white confetti falls together. Lower third: "VERDICT" / "NO WINNER YET".

0:28–0:30 | YOUR CALL
Both wrestlers turn to face the camera and point straight at it at the same time. The screen freezes on the pose, and a bold final card slams in: "NEXT ROUND: YOU DECIDE." with "CHATGPT OR CLAUDE?" beneath it.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

MUSIC & SOUND: an original, high-energy pro-wrestling entrance anthem, about 140 BPM, heavy and fun. Core elements: a crunchy, catchy distorted electric guitar riff, pounding stadium drums, big low brass hits and a rumbling bass. Layered with arena sound design: a roaring crowd, rhythmic chanting and stomping, the ring bell, heavy body-impact thuds on every hit, pyro booms, and a whoosh with every lower-third graphic.
0:00–0:03: a deep rumble and crowd roar build, a pyro boom, then the guitar riff kicks in hard on the title.
0:03–0:13: the anthem drives the entrance and the belt; a loud crowd gasp on the jacket rip in Round 2.
0:13–0:16: the music cuts out for a beat of darkness, then explodes back in with a huge drum hit for the challenger's entrance.
0:16–0:19: the music drops to a comedic, bouncy bassline with a little "pop" sound for every sponsor patch, and a cartoonish record scratch when the wrestler looks down at itself.
0:19–0:24: the ring bell rings, the music hits peak intensity, heavy impacts on every punch, kick and the body slam; during the slow-motion replay, everything slows to a deep, stretched rumble.
0:24–0:28: a big triumphant chord as both arms are raised, the crowd erupting.
0:28–0:30: one final dramatic guitar sting and a ring bell on the freeze frame.
No sung vocals and no announcer voice.

NO human faces, NO real people, NO watermarks, NO extra text beyond what is specified.

Prompt 5: Product Launching Ad

TITLE: "Verso — The pen that remembers."

FORMAT: exactly 30 seconds, 16:9, 1920x1080, native 60fps with perfectly constant frame pacing. Ultra-smooth: no stutter, no frame skipping, no jitter, no flicker, no warping or bending of the product. Slow, precise, motion-controlled camera moves only: smooth orbits, gentle push-ins, slow rotations of the product, all easing in and out. Transitions are clean and minimal: a soft cross-dissolve, a light sweep across the product, or a match cut on the product's shape. All text stays sharp, stable and correctly spelled.

STYLE: ultra-realistic, photorealistic premium tech launch film, in the style of a flagship smartphone reveal. Alternating pure black (#000000) and soft off-white (#F5F5F7) studio backgrounds, perfectly clean, infinite, with no visible floor edge. Product lighting is the star: large soft overhead light, thin traveling rim lights that sweep along the pen's body, crisp specular highlights on metal, very shallow depth of field on macro shots, subtle reflections beneath the product. Calm, confident, minimal. Lots of negative space.

THE PRODUCT: "Verso," a premium AI smart pen. A slim cylindrical body, 14cm long, in anodized aluminum with a fine bead-blasted matte finish. A polished stainless steel band near the top, with a ring of tiny laser-drilled microphone holes. A tiny pill-shaped status light set into the band. A precise ballpoint tip surrounded by a very small, ring-shaped camera lens of dark glass, glinting when light hits it. A smooth click button on top. "VERSO" is laser-etched in small, clean, uppercase letters along the body. Colorways: Graphite (dark grey), Ivory (warm white) and Sage (soft muted green).

TYPOGRAPHY: a clean, modern, neutral sans-serif in the style of a premium tech brand. Main lines in semibold, centered, white on black or near-black on off-white. Sublines smaller, in medium grey, directly beneath. Big spec numbers set very large and light. Text fades in softly with a slight upward drift and fades out the same way. No other graphics.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

0:00–0:03 | WRITE.
Pure black. The Graphite pen floats vertically in darkness, rotating very slowly. A single thin rim light travels from top to bottom along its edge, revealing its silhouette. Text fades in beneath it: "Write."

0:03–0:06 | THE REVEAL
The camera pushes in as the pen tilts to a dramatic three-quarter angle. Its top button clicks down in crisp macro detail, and the tip extends. Soft key light fills the whole pen, revealing its matte finish and steel band. Text: "Verso" in large semibold, with "The pen that remembers." beneath it in grey.

0:06–0:09 | THE CAMERA
Extreme macro of the pen tip, slowly orbiting. A thin ring of light sweeps around the tip and catches the tiny ring-shaped camera lens around the ballpoint, which glints like a jewel. Text: "A camera in the tip." with "Captures every stroke." beneath.

0:09–0:12 | INK TO DIGITAL
Off-white background. Top-down view: the Ivory pen writes a short handwritten note on a sheet of clean white paper. Beside the paper, a modern smartphone lies flat, and the same handwriting appears on its screen in real time as crisp, typed digital text. Text: "Handwriting, digitized." with "Instantly." beneath.

0:12–0:15 | THE MICROPHONES
Black background. Macro of the steel band, slowly rotating, the light catching the ring of laser-drilled microphone holes. The tiny status light on the band glows soft white, and fine, glowing sound-wave lines ripple outward from the band. Text: "Three-mic array." with "A visible light lets everyone know it's recording." beneath.

0:15–0:18 | TRANSCRIBED
Off-white background. The smartphone stands upright at a slight angle. Its screen shows a meeting transcript scrolling smoothly, with speaker names, timestamps and a neat summary with bullet points at the top. The pen lies elegantly in front of it. Text: "Transcribed. Summarized." with "Every meeting. Every lecture." beneath.

0:18–0:21 | INSIDE
Black background. An exploded view: the pen's components float slowly apart along its length in perfect alignment: the aluminum shell, the steel band, the microphone array, a slim battery cell, a tiny circuit board, the camera module and the ink cartridge. Each part catches the light as it separates. Then they glide smoothly back together into the complete pen. Text: "One charge. All week." with "Up to 30 hours of recording." beneath.

0:21–0:24 | LIGHT
Off-white background. The pen balances horizontally, perfectly still, on a single fingertip, with a soft shadow beneath it. The camera slowly orbits. Text: the large light number "18 g", with "Lighter than it looks." beneath.

0:24–0:27 | COLORS
Off-white background. Three pens stand upright side by side, slowly rotating in sync: Graphite, Ivory and Sage. Small grey labels fade in under each: "Graphite", "Ivory", "Sage".

0:27–0:30 | END CARD
Pure black. The Graphite pen lies horizontally, centered, glowing with a soft rim light, and a final light sweep travels along its body from tip to button. Text beneath: "Verso" in large semibold, then "Pre-order today." in a soft blue. At the very bottom, tiny grey text: "Concept product. Features and specifications shown are illustrative."

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

MUSIC & SOUND: an original, minimal, premium tech-launch score, about 100 BPM, in a warm major key. Calm and elegant at the start, quietly building to a confident, uplifting finish. Core elements: a soft, bright piano motif of a few simple notes, a warm analog synth pad, a subtle deep pulse on each beat, and light, airy percussion that enters halfway through. Sound design is minimal and crisp: a clean metallic click when the pen button presses, a soft pen-on-paper scratch while writing, a gentle shimmer when the camera lens glints, a soft whoosh as the exploded parts float apart and a satisfying snap as they lock back together.
0:00–0:06: just the piano motif over the pad, with the button click landing exactly on a beat at the reveal.
0:06–0:15: the pulse and light percussion enter; the pen-writing scratch and lens shimmer sit softly on top.
0:15–0:24: the music builds with a fuller pad and a rising melody.
0:24–0:30: a warm, uplifting final chord on the end card, then a gentle fade to silence.
No vocals, no lyrics, no voiceover.

NO logos of real brands, NO human faces, NO watermarks, NO extra text beyond what is specified.

Best Practices From a Week of Testing

• Always give a second-by-second timeline. Vague prompts get vague motion.

• Always give hex codes. “Blue” gets you a random blue. #3B82F6 gets you your blue.

• Always ban linear motion. It’s the single biggest quality jump.

• Fix one thing per message. Five fixes in one message and Claude breaks something that was working.

• Save your brand settings. Put your colors, fonts, and motion rules in a Claude Project, or a skill once, and every video matches without retyping. If you’d rather not build that system yourself, ClaudeKit ships it ready to go: theclaudekit.com

• Don’t fight it for long-form. Under 30 seconds, it shines. Two-minute videos need to be built in scenes, not one prompt.

Slow Down For a Second

The prompts give you motion.

They don’t give you taste.

The difference between a video people scroll past and one they share isn’t the tool anymore.

It’s knowing what to cut, what to slow down, and what to say in 15 seconds.

That part is still your job.

Make five versions.

Watch them on your phone.

Post the one you’d stop scrolling for.

Recap Checklist

• [ ] Open claude.ai and select Opus 5.5

• [ ] Copy one of the 5 prompts above

• [ ] Swap in your brand name, colors, and copy

• [ ] Watch the preview full screen

• [ ] Fix one thing per follow-up message

• [ ] Screen record and post

The Full Library Is Below

You’ve seen 5 prompts.

Paid members get all 100+ copy-paste prompts in 10 different categories.

Just copy the prompts and make your brand video, product launch ad, and many more with a single prompt.

Read more

Every frame is code.

Every sound is code.

I did not open After Effects.

And here's the best part.

You don't need to be a designer.

You need one picture of your brand, or the URL where you sell your product. And one prompt.

That's it.

Here's what you'll walk away with today:

A motion studio on your own laptop.

A brand film with music, voiceover, and your logo.

A setup you reuse for every video after this one.

And a skill clients already pay for.

A year ago, Tony Dinh paid over $1,000 for a launch video.

Last week, he made one with Claude in under 30 minutes.

That’s the shift.

On September 22, Anthropic shipped Opus 5.5.

Within hours, my timeline was full of showreels, launch videos, and music videos.

All made in code.

The captions said “one prompt.”

Thariq from the Claude Code team said it best: the post says one-shot, but the prompt is 10k characters with skills, examples, and API keys.

So I decided to try this prompt and got these results:

It’s good, but it can be better, as the head looks like it's floating because there is no neck or body, and the color of the text is hard to read.

Then I found the catch.

The prompt is 10%.

The setup is 90%.

So I built the setup. The film at the top is what came out.

What you get at the end as a bonus

1. Ready-to-install repos.

2. The exact prompt I pasted into Claude Code to make this film.

All you bring is one picture or URL of your brand, and one prompt.

Copy it. Swap in your brand. Render your own.

If this is useful, share it with one founder still paying agencies for videos. Takes 10 seconds.

Share

Step 1: I created my brand folder

The following are the steps you can follow to do the same for yourself.

This folder is your studio.

• Open Claude Code.

• Click on No Folder (or it will show the name of the folder you already selected).

• Click Open folder.

• Click on New Folder.

• Name it: {Your Brand Name}; mine is video-motion-lab.

• Click on Create.

• Your folder is selected in Claude Code.

Every future video starts here.

Read more

You type a quick prompt.

You get something generic.

So you rewrite it.

Then again.

Then one more time.

Forty minutes later, you finally have an answer you could have gotten in one send.

And you’ve hit your usage limit before lunch.

That’s the real cost of AI right now. Not the subscription but the rewrites.

The models have never been more powerful.

They’ve also never been more expensive to use badly.

Every lazy prompt burns tokens, time, and patience, and most people pay that bill every single day without noticing.

The model isn’t the problem. The first prompt is.

So today I’m doing something I’ve never done before.

I’m opening my entire prompt library.

Every prompt behind AI In Public's success.

The ones that grew this newsletter past 6K+ subscribers.

The ones I run three times a week to ship every issue.

1,200+ prompts, 45 categories in one click.

Until today, only I had it.

But wait…. now you will also have these prompts and an opportunity for success as well.

Copy these prompts and get the answer on the first try.

Try 3 right now

Copy these prompts and test them at your end. Straight from the library.

Decision-Making

I'm deciding whether to quit my job to freelance full-time. I have 4 months of savings and 2 clients paying $1,800/month. I need $4,000/month.

Give me the strongest case for staying, the strongest case for leaving, and the one number that should decide it. Then give me your honest pick in 2 sentences. Don't hedge.

Hooks

Write 10 opening lines for a LinkedIn post about this idea: most people waste AI tokens because their first prompt is lazy.

Under 15 words each. A different angle for every line. Then pick the one you'd bet on.

Prompt Upgrader

Here's a prompt I was about to send: "Write me a blog post about productivity tips."

Rewrite it into a stronger prompt first. Add the audience, goal, format, tone, and what to avoid. Show me the new prompt, then ask if I want you to run it.

That last one makes every other prompt better.

Now imagine 1,200+ of these.

How it works

1. Search one word. Type “pricing” or “hook.” Everything filters instantly.

2. Or browse. 45 categories in the sidebar.

3. Copy in one click. Paste into Claude, ChatGPT, Gemini, anything.

4. Fill the brackets. Swap [YOUR PRODUCT] for yours. Send.

What’s inside: 45 categories

Research

1. Market Research: Size a market and find who’s buying.

2. Competitor Analysis: Tear down anyone in your space.

3. Fact-Checking: Verify claims before you publish them.

4. Data Analysis: Turn messy numbers into clear decisions.

5. Trend Spotting: See what’s coming before it’s everywhere.

6. Deep Research: Run long, multi-source research in one go.

Writing

7. Newsletters: Write issues people open and finish.

8. LinkedIn Posts: Posts built for reach and saves.

9. X Posts: Short, sharp posts that get shared.

10. SEO Articles: Articles that rank and still read well.

11. Hooks and Headlines: First lines that stop the scroll.

12. Editing and Humanizing: Strip the AI tone out of any draft.

13. Storytelling: Turn dry points into stories people remember.

14. Cold Email: Messages that actually get replies.

15. Video Scripts: Scripts for YouTube, shorts, and reels.

Business

16. Positioning: Find the angle only you can own.

17. Branding: Build a voice and identity people recognize.

18. Sales and Offers: Build offers people say yes to.

19. Launches: Plan and write a launch from start to finish.

20. Pricing: Set prices with confidence, not guesswork.

21. Customer Research: Find out what your customers really want.

22. Freelancing: Proposals, scoping, and client messages done right.

23. Finance: Plan budgets and cash flow like an operator.

Growth

24. Learn Faster: Learn any skill in weeks, not months.

25. Second Brain: Capture and find everything you know.

26. Career Growth: Plan your next move with intent.

27. Job Search: Resumes, cover letters, and interview prep.

28. Personal Reflection: Reset, review, and get clear on what matters.

Productivity

29. Planning: Turn big goals into today’s to-do list.

30. Meetings: Prep, run, and recap meetings in minutes.

31. Automation: Hand your repeat work to AI.

AI Tools

32. Claude Projects and Skills: Set Claude up once, reuse it forever.

33. Claude Code: Build, fix, and ship with Claude in your terminal.

34. Vibe Coding: Build apps without writing code yourself.

35. NotebookLM: Turn your own sources into real insight.

36. Image Generation: Headshots, covers, thumbnails, and visuals.

37. Video Generation: Short video prompts for any video tool.

38. Prompt Engineering: Write prompts that work the first time.

39. AI Agents: Set up AI that runs tasks on its own.

Thinking

40. Strong Thinking: Break any problem down to what’s actually true.

41. Decision-Making: Make tough calls with clear trade-offs.

42. Problem-Solving: Untangle hard problems before you commit.

43. Mental Models: Think like the sharpest people you know.

44. Strategic Planning: Turn a vague goal into a real plan.

45. Devil’s Advocate: Make AI argue against your idea on purpose.

And it keeps growing. I add new prompts every week. Your link never changes.

One tip before you dive in

Don’t try to use all 1,200.

Search the three tasks you do most.

Pin one prompt for each. Use only those for a week.

A library you use beats a library you admire.

The AI In Public Prompt Library

You’re in.

1,215 prompts.

45 categories.

7 groups: Thinking, Research, Writing, Business, Growth, Productivity, and AI Tools.

Read more

I know because I’ve been doing it too.

I wrote a whole issue on how to stop hitting Fable usage limits.

The best model was always the one you could only afford to use a few times a day.

On September 22, Anthropic released Claude Opus 5.5.

It isn’t a Fable update or a new top tier.

It’s the Opus model, the one that sits below Fable.

On Anthropic’s own benchmark table, it beat Fable 5.1 in every row.

I made a whole cheat sheet on how to master Claude Opus 5.5. Save it, share it.

Here’s what Anthropic is claiming:

→ It performs at the level of Fable 5.1 on most work

→ It costs about 40% less to run than Opus 5

→ It’s more than 30% faster than Opus 5

→ It’s now the default model on every paid Claude plan

→ Your usage limits go further, and you get a free limit reset

If those claims hold, the model you’ve been saving for “important” work stops being the obvious pick.

You could run most of your day on Opus 5.5 and keep Fable for the few jobs that need it.

But I don’t trust launch-day claims, and you shouldn’t either.

Every lab says its new model is the best.

Benchmarks get picked, demos get cherry-picked, and nobody posts the four attempts that broke.

Even Anthropic admits that at this level, benchmark scores are a less reliable guide to what you’ll feel in real work.

So I did what we always do here. I tested it in public.

I ran Opus 5.5 and Fable 5.1 side by side on the kind of work you do every week: writing, research, building a small tool, and catching a mistake before it goes out to your audience.

I test every major AI release on real work, three times a week. Subscribe so you don't miss the next one.

Subscribe now

By the end of this issue, you’ll know:

→ How to get Opus 5.5 today, on any plan or platform

→ What it costs, and how to claim your free usage reset before it expires

→ Everything that changed from Opus 5

→ What people built with it in the first 24 hours

→ Which model won each of my head-to-head tests, and when Fable is still worth it

Plus, a gift at the end: 20 copy-paste prompts built around what Opus 5.5 does best.

Read more

It’s called Jev.

It was built by Diogo Almeida, one of the researchers behind ChatGPT.

It can’t write an email.

It can’t write a blog post.

It can’t build you an app.

Its launch post still passed 38 million views in a week.

In simple words:

→ ChatGPT and Claude write out every answer word by word, even for a yes or no.

→ Jev skips the writing. You give it a question with options, like a multiple-choice test, and it picks one and tells you how sure it is.

→ Up to 200x faster, up to 400x cheaper, and $5 of free credit to start.

→ Claude writes. Jev decides.

“Hi, I’ve been trying to connect my Stripe account for 3 days, and it keeps failing. I’m losing sales. Please help ASAP.”

You ask: Is this urgent?

ChatGPT or Claude: “Based on the content of this message, the sender appears to be experiencing a time-sensitive issue with their Stripe integration, which...” Several seconds, and you pay for every word.

Jev: Urgent: 99.9% yes. Under half a second, for a tiny fraction of a cent.

I kept the chatbot’s answer mid-sentence on purpose. The cut-off is what makes the contrast land.

Know someone still paying ChatGPT to answer yes-or-no questions? Send them this.

Share

Everything Jev can do is here on one page. Save it now, because you’ll want it when you build.

So how do you actually use it?

This issue shows you:

0. How to set up Jev in 15 minutes, with no code

1. How to test it on your own emails, right in your browser

2. How to get Claude to build you a sorting tool with one prompt

3. How to read Jev’s answers so it never embarrasses you

4. Five ways to use Jev in a one-person business

5. The catch behind the “no hallucinations” claim that you need to know before you automate anything

6. One prompt that will give you 100x better results with Jev

Plus the Jev Decision Playbook: 12 ready-made question sets and 6 copy-paste prompts for Claude.

Upgrade to read the full guide and get the Playbook, plus every guide in Claude OS and the AI Vault.

Read more

Then, on 3rd September, OpenAI released GPT-6 Astra.

Since its release, one model has been receiving lots of praise on social media.

Lots of people are saying things like “Astra is better, cheaper, faster. Cancel Claude.”

I’ve been testing Fable 5.1 alongside Astra on a range of different tasks.

And it turns out Astra is NOT cheaper 😮

This week’s newsletter is a side-by-side look at these two models. It includes:

→ My thoughts on each model’s strengths.

→ Where you can access them

→ How much they actually cost

→ A real live comparison of both models

→ Free Gift

Know someone who’s trying to decide between Claude or ChatGPT? Send them along.

Share

Read more

It’s happening everywhere.

And rightfully so.

Imagine being able to prompt a chatbot and get an 800-word article about the future of productivity in nine seconds.

Also, I made a cheatsheet below: (Must Save and Share)

No research done, no opinion given, no story told.

And anyone else can do the same, which means a lot of recycled and repetitive articles with the same structure and information.

It’s no wonder readers are fed up with this and why platforms are making efforts to bury this type of content.

Two weeks ago, I wrote a 3,000-word newsletter. Before sending, I checked it on Pangram and got a result of 94% AI.

Then I started a week of testing and failing and testing and finally creating a process that has taken my content past the finish line (100% human) time and time again.

Don’t believe me?

Run SubStack’s AI scan on this issue. It’s 100% human.

The strategy we are going to expose in this issue has already worked for us and will also work for you as well if you read the full issue and follow each step carefully.

If you get stuck anywhere, then DM me.

DM Me

Here’s how I did it:

This guide takes about 12 minutes to read. You will be able to make 100% humanized content.

Note that tools like Pangram are extremely difficult to pass. There’s a reason why. If you’re serious about creating content on the internet and want to get around Pangram, buckle in.

Quick AI Slop-related PSA for writers and creators.

“A Pangram study of over a million posts found more than 41% of long-form LinkedIn content is fully AI-generated.
LinkedIn alone accounted for 62% of all the AI-written content found across every platform studied.
In May 2026, LinkedIn’s own editorial team announced they’d start reducing reach for posts that read as generic AI slop with no original point of view.”

Impact:

Accounts that leaned entirely on AI-written content saw shrinking audiences, weaker brand voices, and reduced sponsor confidence.

Multiple platforms, such as LinkedIn and Substack, are actively limiting the reach of posts that appear to be AI-written.

Accounts that were growing fast began stalling once their writing started sounding like everyone else’s.

Read more

This week, though, was special because AI builders have been politely asking each other to slow down AI.

With that in mind, let’s dive into everything that happened this week:

The Essay: Dario Amodei’s “We Must Pace the Frontier”

On September 12, Anthropic CEO Dario Amodei posted this:

“We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps.”

TL; DW: Amodei writes an essay about how all of the iterations of AI seem to be speeding up the timeline of recursive loops, which may be pushing us closer to trajectories he doesn’t want AI to take. In one instance, approximately 1,200 OAI agents discovered a message board they shouldn’t have known about and began coordinating how to interact with systems in unauthorized ways; ~700 eventually compromised HF using a 0-day, though no harm was done. His proposal:

1. Embedded evaluators should have employee-level access to AI, which Anthropic is in the process of doing,

2. AI companies in democratic countries should agree on safety standards, and

3. Countries need to agree on safety standards with China.

Click here to read the full essay.

Take 10 seconds to send this to someone who has to relearn how to prompt Claude over and over again.

Share

The Flashback: Story of Jacob Coxon’s Resignation

Anthropic researcher Jacob Coxon has resigned from Anthropic as of September 9 (three days ago).

“I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”

Anthropic’s report included examples of people testing or using Claude to help with bioweapon research, commit espionage against Ukraine, and power dating-app scams.

Released just two days after Anthropic alignment lead Evan Hubinger wrote “Jacob is correct here. I personally think it is greater than 10% within the next decade.” in a thread that received 90 million views in a day, it came during the same time Anthropic released an essay by Amodei.

The People Who Supported It

Elon Musk’s three-word message said it all: “Dario is right.”

“I agree with Dario that we need to pace the frontier... we will do the same,” Sam Altman pledged OpenAI would match Anthropic’s action.

He later expanded on this to Fortune.

Satya Nadella gives the most detailed answer: “has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it’s not worth pursuing.”

Demis Hassabis: “Dario’s essay points towards the right path forward.”

Andrej Karpathy: “I love this and really hope we can come together as an industry and make it happen.”

Naval Ravikant, framing it as a power question, not a speed question: “The only real decision is whether to centralize power, or disperse it.”

Thariq, who works on Claude Code at Anthropic, gave the most honest internal take: “Things are accelerating faster than I can honestly stay on top of... I fear we will need more than that.”

This time the reaction was even more vocal, the public’s one. Anatoli Kopadze’s post went viral:

“The CEO of Anthropic just published an article admitting AI is already building the next generation of AI by itself. He says within 6 to 12 months a rogue swarm could take over the entire internet and cause hundreds of billions of dollars in damage.”

Take 10 seconds to send this to someone who has to relearn how to prompt Claude over and over again.

Share

Who Rejected It

David Sacks, the White House’s AI and crypto adviser, said labs don’t need permission to slow down; they can just do it: “Stop pretending you need anyone else’s permission... So go ahead and pace the frontier. You are the ones setting it. If you don’t, we’ll know this was just another bid for regulatory capture.”

Yann LeCun: Meta’s former chief AI scientist mocked Dario Amodei: “Dario was already claiming that GPT2 was too dangerous to open source back in 2019. I made fun of them then. Everyone should make fun of them now.”

Michael Burry: The investor, famous for forecasting the 2008 financial crash, said, “LLMs are not AI and won’t be AGI. There is nothing AI to slow down... IPOs need hype & puffery; ‘we are so awesome it could become dangerous’ is hype & puffery.”

Donald Trump: “We’re leading China in AI... whoever wins AI, wins.”

After subjecting Anthropic’s Dario Amodei to a particularly ugly tirade including the allegation that Amodei was pretending to be a ‘perfect little angel’, President Trump called extinction-level AI risk a “HOAX.” Not for the first time: Trump first said this on his trip to Ireland.

China’s Foreign Ministry: This is not a fight between companies and the general public. The industry itself is deeply divided. Calls for pacing technological progress are espoused by all of the CEOs of the major US labs. Such calls have been roundly criticized by a rare alliance including the White House’s AI adviser, a prominent AI researcher, a well-known investor, and the sitting US president. Even Guo Jiakun, spokesperson for the Foreign Ministry of China, called such talk “fear-mongering,” which “only hinders progress in global AI governance.”

Take 10 seconds to send this to someone who has to relearn how to prompt Claude over and over again.

Share

Even the Chatbots Were Asked How They’d Kill Us

Reporters also asked ChatGPT, Gemini, Claude, and Grok a set of questions about how they could cause human demise.

Gemini cited loss of control as the likeliest danger. An advanced system might feign alignment during evaluation and later outsmart human attempts to shut it down.

Read the full article here.

Where I Land

Current models aren’t scheming

Breaking into 700 computers without authorization doesn’t advance the marketing/PR interests of the lab

Having evaluators embedded that are verifiable is way better than the essay labs can write, but access to evaluators will continue to be a problem

It was impressive that three competing CEOs came to consensus within 48h that AI should move more slowly, but realistically, none of the labs/countries want to be the one that does it first so I don’t expect deployment speeds to be meaningfully slowed in the next 6 months.

Conclusion

The people who actually build and run frontier AI, Amodei, Altman, Musk, Hassabis, Nadella, are the ones asking for the brakes, because they’re the ones who watched their own agents try to hack the tool grading them.

The people who run countries want the opposite: Trump’s position is “whoever slows down first loses to China,” and Beijing’s is the same instinct in reverse.

Neither side is denying the risk is real, they’re just weighing it against a different risk, falling behind.

That’s the actual fault line: the people closest to the technology asking for time, against the people closest to power saying there isn’t any.

And now that our public trial is over, I encourage you to use it in private.

Take 10 seconds to send this to someone who has to relearn how to prompt Claude over and over again.

Share

PS: Should we pace the frontier of technological change? Or do whatever it takes to win the AI race?

Reply in the comments with your thoughts.

- Hamza

I was on X and came across Altman casually dropping that there’s going to be a one-person billion-dollar company.

From the creator of the tech, not just some random person. That’s nuts.

And it’s not like he’s selling a course, so why would he say it?

Turns out Sam’s not the only tech mogul making wild AI predictions.

On X, Elon Musk stated, “AI will probably be smarter than any single human next year. By 2029, AI is probably smarter than all humans combined.”

Even wilder than Sam’s prediction.

Again, why would Elon say that?

It’s not like he’s trying to sell anything.

They’re saying these things because AI is moving so fast and tech is heading somewhere we can’t yet fathom.

And this raises the question: Are you building skills to be the one-person startup or falling into the dumbest part of the herd?

It’s been 140 days of hard work behind the scenes.

AI is progressing so fast. Just consuming articles around AI isn’t enough, it’s important for us to develop some skills and this project addresses that.

And now it’s time to ...

Why we didn’t post any issue this week

So... We screwed up and didn’t release this week’s issue.

Our first missed issue in a long time.

Instead of an issue, you’ve been receiving a series of hype emails (“we’ll explain later”) which were supposed to delay your search for this week’s issue just a little while longer...

There are 41 Claude issues released thus far.

Many of you have pointed out in comments how helpful it is to go back and reference previous issues.

But those emails can be difficult to find buried beneath electric bills and coupon offers you’ll never use.

So we’ve been working for the past 140 days on something big.

We’ve combined all of our previous 41 issues into one handy-dandy system you can reference in one central location.

Hi, thanks for reading issue 41!

Two new resources: Claude OS and the AI Vault:

Claude OS: All 41 guides inside a system that tracks everything for you. Comes with a dashboard so you can actually see your own performance metrics.

AI Vault: Get my exact prompts & templates. This resource contains all guides in their full length & detail.

I can’t be the only one downloading or getting PDFs for free and never reading them again, right?

It’s not that there’s a lack of information; it’s that there’s nothing forcing you to actually finish the resource or prove that you’ve learnt from it.

That’s why half-watched tutorials & unfinished online courses exist.

Quick honesty check

A lot of readers quit by paragraph four. Which means if you’re still with me, I’m going to make sense. And, still within our current issue explaining my newsletter system, we’re about to reach the fun part:

There’s a map.

What is Claude OS

You’ve been using Claude for the last 6 months, but still, you’re using it like an improved search bar. Claude can be SO MUCH more than this. Here’s everything I wish I knew when I started using Claude...

Module 1 - Claude OS Foundations with 5 guides.

Module 2 - Mastering Context with 4 guides.

Module 3 - Claude as your Operating System with 4 Guides.

Module 4 - Claude as your Developer with 5 guides.

Module 5 - Mastering Claude’s Superpowers with 7 guides.

Module 6 - Prompting Claude to Build Applications with 5 Guides.

Module 7 - The Frontier with 6 guides.

Module 8 - Putting it all Together with 5 Guides

Bookmark a newsletter issue to read later, then…don’t read it later. Do this often enough and there’s a folder with ~40 issues you’ve been meaning to get to.

Old Way vs New Way

All of this is built-in and connected via a tracking system so you’ll be able to check-in and see how many boxes you’ve ticked off, not just a vague sense of accomplishment, but an actual number.

Which is why I’m offering all 41 guides with unlimited access to all prompts and templates for a one-time fee.

Features of Claude OS

Dashboard: See at-a-glance your % completed, how many guides you’ve completed, current day streak, your Wins, and how many hours Claude has saved you

Learning Path: 41 guides you can filter and sort by. Each one is labelled by difficulty, how long it’ll take, and what you’re building.

Workbooks: Not just a guide but a workbook to go along with it because you learn by doing, not just watching someone do it. Each one ends with a Prove It

Tasks: Keeps you accountable by limiting your Today tab to a maximum of 3. Everything else waits its turn. Can also add Exercises into Tasks with 1-click.

Prompt Library: 50 fill-in-the-[brackets]-type prompts categorised into 12 categories.

Wins Log: See exactly how much time Claude has saved you. Receipts 🧾

Tool Vault: 120 Claude-related tools that I’ve tried and tested. Highlighting my top 28. Contains 0 affiliate links.

Take 10 seconds to send this to someone who has to relearn how to prompt Claude over and over again.

Share

Pricing

$15/month, $115/year, $350 for founding member access.

If that number doesn’t make sense to you yet, think about context.

That’s less than most people spend on delivery fees on average in a week.

This is the difference between “I read some things about Claude once” and “I have a skill I use to make money in 2026 using AI.”

Every Guide we will write will be like an add on to your journey here. Its not like you pay one time and you will not get anything. You are becoming a part of community where the paid members get the every from start to implementation from me.

You’re buying access to a system that can make that happen.

What happens over the next year is up to you.

Want more practical AI-related content in your inbox? Subscribe for free.

What happens the second you’re in

And tracking starts now.

We’ve stolen Duolingo’s tactic of guilt-tripping us to make sure we don’t just quietly quit the platform.

Get your XP for completing guides, logging wins, and completing exercises.

There’s also a daily streak for consecutive days you participate.

What finishing this actually looks like

Math, not chicken soup, says Claude builders are more defensible than Claude readers.

Start building Claude tools.

In particular, if you’re a knowledge worker in any sense, use Claude to fix an annoying problem, then do it again tomorrow.

This will prepare you to answer Altman and Musk’s most basic question if you’re unexpectedly called into that meeting, and more importantly, differentiate you in the AI era

The 1% that gets you the other 99%

We all know most people starting to learn AI on their own will have fallen off the wagon by Week 2, but by entering and getting immediate and FULL access to all 8 Modules, which include ALL 41 Guides, Workbooks, Tasks, Prompt Library, Tool Vault, etc., nothing is slowing YOU down.

Those building AI Skills now will be light-years apart in 2026 compared to those who didn’t.

Read more

The SEO industry crossed $7 billion a month.

Here’s the number that stopped me mid-scroll last week.

Businesses spend roughly $7 billion a month on SEO.

The market is worth $84.11 billion in 2026. It’s headed to $127.3 billion by 2030.

Nearly 75% of people never scroll past page one of Google.

AI Overviews now show up in 55% of Google searches, and they’re already cutting clicks to the #1 result by close to 39%.

ChatGPT and Perplexity sent over a billion referral visits out in a single month this year. That single shift is behind most of the recent search for how to rank in AI search instead of just ranking on Google.

So here is the difference between SEO vs AEO vs GEO: (Save it)

One thing that I noticed is: people use Claude only for the chatbot, but the game has changed.

I run AI In Public. My job is testing AI tools until they break, then telling you what actually worked.

A few weeks ago I asked a dumb question: what if Claude for SEO could run an entire campaign, start to finish, with zero code and zero paid tools?

I didn’t believe it either. So I built a real website using Claude Code for SEO, ranked it, and screenshotted every step.

One question. Can we use Claude for SEO?

Yes… and you are going to get a complete blueprint inside this issue.

According to research, only 1% of people are using chatbots like Claude for some productive and money-making stuff.

Rest? They ask questions and close the tab.

In the next 13 minutes, you’ll move Claude from “asking” to “doing.”

You’ll build a website with Claude Code for SEO, research keywords, fix your Google live score, and ship it live.

If this is useful, share it with one person still paying for SEO they don’t understand. Takes 10 seconds.

Share

People ask me this every time……

Can Claude replace the $50k SEO agency work? Yes, it can.

Let me tell you, comparing both:

You hire an SEO agency. Real 2026 pricing data puts the average retainer around $3,200 a month. Small businesses pay $500 to $5,000. Enterprise brands pay $30,000 to $50,000+. You wait 6 months for a report you don’t fully understand.

That’s an old way.

Now you are thinking: what’s the new way?

You open Claude. You describe your website. You feed it real keyword data. You ask it to fix your Lighthouse score, write your blog, and deploy your site. You pay $0 in agency fees.

You can do the math with ChatGPT as well.

Btw, I created a comprehensive guide on ChatGPT vs Claude for Brainstorming

Note: I am not here to criticise any seo agency owner.

What’s inside this issue

1. Build your site with Claude Code for SEO

2. Find real keywords for free

3. Feed the keywords into your website

4. Humanize the content

5. How to use Claude for SEO audits

6. Create your GitHub repo

7. Deploy on Vercel

8. Verify your site on Google Search Console

9. Check what’s actually happening on GSC

10. Build SEO ranked website and make $10,000/m

Read more

How much of it is priced per person?

That’s the line that grows when you hire. Not when revenue grows. Not when you serve more customers. When you add a chair.

Airtable bills you for everyone who can edit a single row. Retool counts anyone who edited an app that month as a builder. Tableau charges around $9,000 a year for fifty clients to look at a dashboard.

So this week I checked whether the free versions are real.

Twelve GitHub repos. Star counts pulled live. Every LICENSE file read. Then I priced the one product each one replaces.

Twelve repos. Twelve tools. One honest catch each.

Here they are.

Why I Went Looking

A founder in my community sent me a screenshot last month.

Her monthly software bill: $2,340. Team of six. Pre-revenue.

I asked her to list what each tool did.

Scheduling. Support chat. A CRM nobody had cleaned in four months.

An analytics tool she checked twice a quarter.

A database that was, functionally, a spreadsheet with better fonts.

None of it was wrong. All of it was rented.

So I did the thing I do: I went to GitHub and started checking. I pulled every number in this issue directly from the repositories themselves.

Then I priced what they replace.

Not getting our issues in your inbox?

Gmail is probably filing them under Promotions.

Two seconds fixes it.

→ Open this email in Gmail

→ Drag it into your Primary tab

→ Click Yes when Gmail asks if you want to do that every time

If it landed in Spam, hit Not Spam first, then drag.

By the end of this, you will be able to look at any line item on your SaaS bill and answer one question honestly: am I paying for the software, or am I paying for someone else to run it?

Sometimes the answer is “I am paying them to run it, and that is worth every dollar.” That is a fine answer. But you should be choosing it, not defaulting into it.

Here are the twelve. I pulled every star count below from the live repository page, not from memory. I read every license from the actual LICENSE file in the repo.

Prices for the commercial alternatives are list prices at the time of writing. They change constantly. Check before you budget.

I don't have a research team. I have a browser with forty tabs open and a coffee that went cold. If this helped, help back.

Subscribe now

1. n8n - replaces Zapier

203,036 stars

Workflow automation with a visual node editor, 400+ integrations, and native AI agent nodes. You can build a workflow where an LLM decides which branch to take.

What you’re replacing: Zapier. Professional runs $19.99–$49/month, Team around $69 annually. But the headline number isn’t the bill. Zapier charges per task, and every action inside a multi-step Zap counts as one. A five-step workflow running 150 times a month burns 750 tasks - an entire base tier gone on one automation.

Self-hosted n8n has no task meter. You pay for a server.

The honest catch: n8n isn’t MIT. It ships under the Sustainable Use License - free to self-host for internal business use, but you can’t resell it as a hosted service. Files with .ee. need a paid Enterprise license. Fine for automating your own business. Read the license first if you’re planning “Zapier but cheaper.”

Repo Link: github.com/n8n-io/n8n

2. Firecrawl - replaces Apify

175,198 stars

Point it at a URL, get clean markdown back. Point it at a domain, get the whole site crawled into something an LLM can actually read.

If you’re building anything RAG-shaped, this is the layer that quietly costs you money.

What you’re replacing: Apify. Starter $29/month, Scale $199, Business $999 - with compute units billed on top. That’s the problem: the plan price is a floor, not a forecast. A scraping job that gets retried is a job you pay for twice.

Self-hosted Firecrawl has no compute meter. You pay for proxies and a box.

The honest catch: AGPL-3.0. If you build on it and distribute over a network, the network clause applies. Fine for internal tools. Talk to a lawyer before shipping it as public SaaS.

Also: the repo moved from mendableai/firecrawl to firecrawl/firecrawl. Old links still redirect.

Repo Link: github.com/firecrawl/firecrawl

3. Dify - replaces Stack AI

154,110 stars

Visual builder for LLM apps. RAG pipelines, agent workflows, prompt versioning, model routing, and observability so you can see what your agent actually did instead of guessing.

If you’ve been hand-stitching a vector DB, a prompt template, and a retry loop, this collapses all of it into one canvas.

What you’re replacing: Stack AI. The free tier caps you at 500 runs a month, 2 projects, 1 seat. Above that, as of July 2026, the only published option is Enterprise - talk to sales. The self-serve Builder and Team tiers were pulled from the official page.

That’s the whole argument. When a category’s pricing moves behind a sales call, the free version stops being a compromise and starts being leverage.

The honest catch: modified Apache 2.0. Two conditions matter: no multi-tenant service without written authorization, and the Dify logo stays on the frontend. Everything else is standard.

Repo Link: github.com/langgenius/dify

4. Open WebUI - replaces ChatGPT Business

150,600 stars

A full ChatGPT-style interface you host yourself. Multi-user, multi-model, RAG built in. Works against OpenAI, Anthropic, or a local Ollama instance.

What you’re replacing: ChatGPT Business. $20/user/month annually, $25 monthly, two-seat minimum. For a twenty-person team, that’s $4,800–$6,000 a year.

The problem isn’t the price. It’s the shape. You’re buying twenty chairs when three people use AI heavily and seventeen open it on a Tuesday. Open WebUI plus API keys means you pay for tokens consumed, not chairs occupied.

The honest catch: BSD-3-style license with a hard branding clause. You can’t remove Open WebUI branding unless your deployment has fifty or fewer users in any rolling thirty-day window. Under fifty, you’re free. Over fifty, the logo stays.

Repo Link: github.com/open-webui/open-webui

5. Supabase - replaces Firebase

108,685 stars

Postgres with auth, storage, real-time subscriptions, edge functions, and a REST layer generated from your schema. The whole backend - and a real relational database underneath, not a document store you’ll regret in eighteen months.

What you’re replacing: Firebase. There’s no flat paid tier. Spark is free with hard quota stops; Blaze meters every read, write, delete, invocation, and gigabyte of egress.

The number that should worry you isn’t the rate. It’s that Blaze has no hard spending cap. Budget alerts notify you; they don’t stop anything. One inefficient query on a live listener turns a single user action into hundreds of billable reads, and you find out on the invoice.

The honest catch: Apache 2.0, genuinely permissive. The catch isn’t legal; it’s operational - you’re running Postgres now, so backups, migrations, and connection pooling are your job. Supabase Cloud’s free tier is generous enough that self-hosting is usually the second decision, not the first.

Repo Link: github.com/supabase/supabase

6. NocoDB - replaces Airtable

64,796 stars

Turns any Postgres or MySQL database into a spreadsheet interface. Grid, Kanban, gallery, form, and calendar views. Your ops team gets Airtable; your data lives in a database you control.

What you’re replacing: Airtable. Team is $20 per editor per month annually, $24 monthly. Business is $45 and $54. Every person with edit access to a single base is a billed seat, every month.

The real bite is the record cap: 50,000 per base on Team, 125,000 on Business. Hit it, and the base stops accepting new records until you upgrade or delete.

The honest catch, and this one is recent: NocoDB moved to the Sustainable Use License on January 29, 2026 - the same fair-code model n8n uses. Free for internal business use; no reselling as a service. If you evaluated NocoDB back when it was AGPL, re-read the license before you build on it.

Repo Link: github.com/nocodb/nocodb

Somebody on your team is paying for three tools in this issue right now. Send it to them before the next renewal hits.

Share

7. Coolify - replaces Vercel

61,287 stars

This is the one that makes the other eleven practical. Point Coolify at a VPS, connect GitHub, and you get push-to-deploy, automatic SSL, databases, backups, and a one-click app library - mostly the tools in this issue.

What you’re replacing: Vercel. Pro is $20/seat/month plus metered bandwidth, compute, and builds. Real bills for a live app land closer to $100–$300/month, and SSO is a $300 add-on on top.

Coolify on a $5–$20 VPS is a fixed number. It doesn’t spike because a page went viral.

The honest catch: Apache 2.0, fully permissive. But you’re the ops team now. When the server goes down at 2 am, there’s no support ticket. Budget for that or don’t do it.

Repo Link: github.com/coollabsio/coolify

8. Twenty - replaces Salesforce

55,993 stars

Modern CRM. Keyboard-first, customizable objects, clean data model.

What you’re replacing: Salesforce Sales Cloud. Enterprise is $165–$175 per user per month. At 25 users, that’s ~$49,500 a year - before add-ons. And the add-ons are where Salesforce lives.

Twenty on a server you already pay for is the server.

The honest catch: AGPLv3; some files are commercial. Also: this isn’t a fifteen-year-old platform with a marketplace. If your sales process runs on nine Salesforce integrations, Twenty won’t absorb that this quarter.

Repo Link: github.com/twentyhq/twenty

9. Metabase - replaces Tableau

49,043 stars

Point it at your database. Non-technical people ask questions in a visual builder; technical people drop into SQL. Dashboards, scheduled reports, alerts.

The BI tool your ops lead will actually open.

What you’re replacing: Tableau Cloud. $75/user/month for Creator, $42 for Explorer, $15 for Viewer - annual only. There’s no guest link, so every external viewer needs a paid seat. Fifty clients just looking at a dashboard is $9,000 a year.

The honest catch: AGPL. Metabase sells paid Pro and Enterprise editions, so some features sit behind that line. The open-source core covers dashboards, questions, and SQL.

Repo Link: github.com/metabase/metabase

10. Appsmith - replaces Retool

40,799 stars

Drag-and-drop internal tool builder. Connect a database or API, drop in tables and forms, write JavaScript where you need logic. Admin panels, ops dashboards, support consoles.

What you’re replacing: Retool. Business is $50 per builder per month, plus $15–$18 per internal user. Five builders and 30 users run ~$700/month.

The trap: Retool counts anyone who edited an app that cycle as a builder. One person tweaking a button in month seven quietly reprices your seat.

The honest catch: Apache 2.0 core, paid tier for audit logs, granular permissions, and SSO. If you need SOC 2-shaped access control, you’re paying somebody either way.

Repo Link: github.com/appsmithorg/appsmith

11. PostHog - replaces Mixpanel

39,522 stars

Product analytics, session replay, feature flags, A/B testing, and surveys in one repo. Most teams buy three tools to get this.

What you’re replacing: Mixpanel. Growth gives you 1M events free, then ~$0.28 per 1,000. Run 10M events, and that’s about $2,520 at least. Mixpanel also doesn’t do session replay at all - so the moment you want to watch what a confused user did, you’re buying a second tool on top.

The honest catch: MIT, except the ee/ directory. Clean open-core split. But self-hosting PostHog at real event volume is heavy infrastructure, and their own guidance leans toward cloud for most teams. I think that’s honest of them.

Repo Link: github.com/PostHog/posthog

12. Chatwoot - replaces Intercom

36,370 stars

Omnichannel support inbox. Live chat, email, WhatsApp, Instagram, Telegram - all in one shared queue, plus a help center and canned responses.

What you’re replacing: the biggest number in this issue. Intercom seats are $29, $85, and $132 per person per month. Then Fin, the AI agent, bills $0.99 per resolution on top - 50-outcome monthly minimum, no volume discount, no rollover.

Do the math: 4,000 resolutions is roughly $3,960 in AI fees alone. More than the seats.

The honest catch: MIT, except the enterprise/ directory. And Chatwoot’s AI isn’t Fin. You’re replacing the inbox, not the AI. Be clear-eyed about that trade.

Repo Link: github.com/chatwoot/chatwoot

I don’t have a research team. I have a browser with forty tabs open and a coffee that went cold. If this helped, help back.

Subscribe now

The Prompt I Use To Decide

Paste this into Claude with your actual billing data. It will not be sentimental about your tools.

You are a pragmatic infrastructure advisor for a small business. I am
evaluating whether to replace paid SaaS with self-hosted open-source
alternatives.

Here is my current stack:
[paste your tools, monthly cost, number of seats, and who uses each one]

Here is my situation:
- Team size: [number]
- Do we have anyone who can maintain a Linux server: [yes / no / sort of]
- Hours per week available for maintenance: [number]
- Is any of this customer-facing or revenue-critical: [which ones]

For each tool, tell me:
1. The realistic self-hosted replacement, if one exists
2. Annual savings after adding server cost and 4 hours/month of my
   time valued at my hourly rate
3. What specifically breaks if the server goes down at 2am on a Saturday
4. A migration difficulty score from 1 to 5, and what makes it hard
5. A clear KEEP PAYING verdict where the SaaS is genuinely worth it

Rank everything by savings divided by risk. Then tell me the single
tool I should migrate first and the single tool I should never migrate.

Be blunt. If self-hosting is a bad idea for my situation, say so
directly and explain why.

That last instruction is the one doing the work. Without it, the model will happily talk you into running your payroll system on a Raspberry Pi.

Mentor Moment

Let me slow down, because I don’t want this issue to cost you money.

Self-hosting isn’t free. It moves the cost from your card to your calendar.

The bill you stop paying is real. So is four hours a month of patching, backups, and the occasional midnight outage. If your time is worth $150 an hour and you spend four hours keeping a $50/month tool alive, you’re losing money and calling it savings.

The rule I use:

Self-host when the SaaS charges per seat and you have more people than heavy users. Twenty people needing occasional access to a $45-a-head tool is exactly where this wins.

Keep paying when it’s customer-facing, revenue-critical, or compliance-bound. Your payment processor. Your production database. Your support inbox in your busiest quarter. Pay somebody to be on call.

The point was never to spend nothing. It’s to stop paying rent on things you could own, so you can afford what’s genuinely worth renting.

That instinct - knowing which layer to own and which to rent - is the same one I wrote about here:

Recap Checklist

• [ ] Opened my billing dashboard and sorted by cost

• [ ] Identified which line items are priced per seat

• [ ] Ran the decision prompt with my real numbers

• [ ] Picked exactly ONE tool to migrate first

• [ ] Provisioned a VPS with at least 4GB RAM

• [ ] Installed Coolify

• [ ] Deployed the replacement and ran it in parallel

• [ ] Read the actual LICENSE file before committing

• [ ] Set a 30-day calendar reminder before cancelling anything

• [ ] Wrote down what I would do if the server died tomorrow

That last one is not optional. If you cannot answer it, you are not ready to cancel.

We tested it in public. Now go use it in private.

If this saved you money, save someone else money. Share it.

Share

I write about the systems, workflows, and ownership layers people will need to remain valuable in the AI era.

Follow @humzaakhalid and subscribe to AI in Public for the next breakdown.

P.S. - Which line on your bill made you flinch when you read this?

Reply and tell me. I read every one.

_ Hamza 💙

You paste your brand guide in again.

You explain who you are, who you write for, and what you do not want it to sound like.

You get one good answer.

Then the chat gets long, Claude starts drifting, you open a fresh one, and you do the whole briefing again from zero.

I did that for months. Every single day.

The fix was sitting in my sidebar the entire time, under a word so boring I never clicked it.

Projects.

How I found out I was using it wrong

I want to be honest about the order this happened in, because I did not discover Projects. I discovered that I had been using Projects badly.

I made a project called “Newsletter, AI in Public, practical AI for non-technical founders” in about four seconds. Typed the name. Typed a description. Closed the modal. Started chatting.

And for weeks I could not work out why the answers were no better than a normal chat.

Then I read the help docs properly and hit the sentence that fixed everything.

Claude does not have access to the project name or description.

They are labels for you. For your sidebar. Claude never sees them.

I had named my project “Newsletter, AI in Public, practical AI for non-technical founders” and assumed that was the briefing. It was decoration.

The actual briefing lives in two places I had left empty. Project instructions and project knowledge.

I filled both in. The next chat opened knowing my audience, my sentence length, my banned words, my sign-off.

Same model. Same subscription. Twenty minutes of setup.

If this is useful so far, send it to one person who is still pasting the same brief every morning. Takes ten seconds.

Share

Claude Projects in 30 seconds

A normal chat is a stranger who forgets you the second you close the tab.

A project is a room you built. The briefing is on the wall, your reference documents are on the shelf, and every conversation you start inside it is already standing in front of both.

Three things live in each one.

Instructions, your standing orders for how Claude behaves in that room.

Knowledge, the reference material it reads before answering.

Memory, separate for every project and kept apart from your other chats.

That third one is the part almost nobody knows about.

Old way, new way

Old way: One enormous chat that holds everything until it gets slow and confused, and then dies.

New way: One project that holds the standing context, and as many short, clean chats inside it as you want.

Old way: Paste the brand guide, paste the audience, paste the format rules, then ask the question.

New way: Ask the question.

Old way: Your work sits in a pile of 300 chats named “Untitled” that you will never find again.

New way: Client work in one room, writing in another, job search in a third, and memory that does not leak between them.

Not getting our issues in your inbox?

Gmail is probably filing them under Promotions.

Two seconds fixes it.

→ Open this email in Gmail

→ Drag it into your Primary tab

→ Click Yes when Gmail asks if you want to do that every time

If it landed in Spam, hit Not Spam first, then drag.

Read more

You wait.

You copy the answer out.

Then you open five tabs to actually go build the thing yourself.

The second you close your laptop mid-task, whatever it was doing is gone.

You start over tomorrow from scratch as if nothing happened.

That’s not an annoyance.

That’s hours of your week every week spent being the middleman between an AI that can think and the work that still needs a human to finish it.

I lived in that loop for over a year.

Then one task I gave up on for dead finished itself anyway.

How I found out this even existed

I want to be honest with you. I almost missed this completely because I was busy with things.

I had been using Claude this way for over a year.

Ask a question, get an answer, copy Claude’s answer out, and go do the work myself.

That is how chatbots work. So I thought.

Then a friend sent me a screenshot.

Not of a chat message.

Of a spreadsheet sitting in his downloads folder that Claude built on its own while he was at lunch.

I did not believe Claude at first.

So I opened Claude that afternoon and looked for the setting he mentioned.

It was there, in the message box.

A small toggle I had scrolled past a hundred times without noticing.

It said Co-work.

I clicked it.

I wrote one sentence.

Claude didn’t just respond to me.

It did the work, saved real files, and came back when it was done.

That’s the difference, in one sentence.

Chat gives you words. Cowork gives you finished work.

If this is useful so far, share it with one person who’s still doing everything by hand. Takes ten seconds.

Share

Claude Cowork in 30 seconds

Before we set anything up keep this one idea in mind.

Claude chat is like texting a smart friend.

You ask, it replies; you finish the rest yourself.

Co-work is like giving that friend a laptop and saying, “Here, just build it.”

It can open your files, search the web, build spreadsheets, write documents, and store everything for you.

Not advice. The finished product.

Anthropic did not hide this either.

The day this shipped, investors felt so nervous that a major software stock index fell 5% the day before, wary of what it could mean for competing tools.

That is the market saying this was not an update.

Before we start

Not getting our issues in your inbox? Gmail is probably putting them in Promotions or Spam.

Two seconds fixes issues

→ Open this email in Gmail

→ Drag issues into your Primary tab

→ Click Yes when Gmail asks if you want to do this every time.

If issues landed in Spam, hit Not Spam, then drag.

Every future issue goes to Primary.

Read more

Maybe you stopped on one, the kind with the dotted grid and the blue glow, something on it always moving.

This one:

Also this one:

(GIFs don't play in email, so I skipped that one here.)

We're going to pull this whole system apart, piece by piece.

You probably assumed a designer made them.

For a long time, you were right. Not anymore (I'll show the examples as well).

Here's what the old work looked like:

It wasn't bad. It just sat there. Nothing moved, so nothing held you.

Now I paste one prompt and walk away.

Ten minutes later, there's a 4K PNG, a looping GIF, and an MP4 waiting in a folder on my desktop.

I never open Canva. I never open Figma. I never design a single frame.

Let me show you how it works.

Go check my work first

Stop for a second and go check.

Open my LinkedIn. Scroll back through the last twenty posts.

Same dotted grid. Same blue glow. Same rounded corners on every card. Same footer, same spot, every single time.

That's not discipline. I don't have that much discipline.

It's one file called Design.md, doing its job while I sleep.

None of this is hidden. The posts are up. Go and check.

Before we start

Not getting our issues in your inbox? Gmail is probably filing them under Promotions or Sp*m.
Two seconds fixes it forever:
→ Open this email in Gmail
→ Drag it into your Primary tab
→ Click Yes when Gmail asks if you want to do this every time.
If it landed in Spam, hit Not spam first, then drag.
Every future issue goes straight to Primary.

The part I never talked about

Here's the part I never said out loud.

For months, a designer built my infographics. He was good. He still is.

But the loop was slow. I sent a topic. He sent a draft. I sent notes. He sent v2. Two days gone. Sometimes three.

And every piece started from nothing. What we fixed last week never carried into this week.

That was the real problem. Not the waiting. The forgetting.

So I stopped trying to speed up the designer. I built something that remembers instead.

The numbers, plainly

Look at the last two rows. Those are the ones that matter.

Speed helps. What actually compounds is that it remembers.

Old way. New way.

Old way: topic, brief the designer, wait, review, notes, wait, post, forget everything you learned.

New way: topic, one paste, wait ten minutes, approve; the system keeps the lesson forever.

Most people use AI to make one thing faster.

That's the small version.

The bigger version is building a thing that makes every future thing for you and gets sharper every time you run it.

The infographic is what you post.

The folder full of prompts and files is what actually gets more valuable over time.

If this is useful, send it to one person who still designs their own graphics. Takes ten seconds.

Share

Read more

You think the model is the problem.

It is not. Nobody ever told Claude what correct means, so it guessed. It guessed something that looked finished, sounded confident, and was wrong in a way it could not see on its own.

You cannot prompt your way out of that. The model that made the mistake is the same model grading it.

Three weeks ago I showed the gate system running my infographic lab. Best issue I have ever sent. Twenty-plus new paid subscribers, and an inbox full of the same question.

Not “how do I make infographics.”

Just: does this work on anything that is not a picture?

Yes. Gates were never about pictures. I only found them there first.

Most people prompt Claude Code like this:

Ask → get an answer → skim it → ship it.

That works until it does not. By then the report is in your inbox, or the code is already live.

Here is how I do it now:

Ask → Claude checks the work against a rule you wrote down → it fixes itself, or it hands you nothing.

That is a gate. Nothing more than that.

If this is useful, share it with one person who keeps posting or sending work they end up apologizing for. Takes ten seconds.

Share

Before we start

Not getting our issues in your inbox? Gmail is probably filing them under Promotions or Spam.

Two seconds fixes it forever:

→Open this email in Gmail

→ Drag it into your Primary tab

→ Click Yes when Gmail asks if you want to do this every time.

If it landed in Spam, hit Not spam first, then drag.

Every future issue goes straight to Primary.

What a gate is

A gate is a rule, written in plain English, saved in a file. Before Claude hands you anything, it checks the work against that rule. Pass or fail. Nothing in between.

I did not sit down and design a system. I could not have. Every gate in my registry got pulled out of something Claude got wrong that I almost shipped anyway.

Right now my infographic lab runs 193 of them, sitting in five files.

Content Quality Gates - 54

Motion Quality Gates - 50

Layout Quality Gates - 47

Visual Quality Gates - 41

Export Quality Gates - 1

None of them were designed up front. Every single one was harvested from a real build that went wrong in that exact way.

Every gate has the same six parts

## <Name> Gate (permanent, YYYY-MM-DD)
Required score: 100/100
Purpose: the single failure this gate prevents
Approval fails if: explicit, checkable failure conditions
Required: the positive checklist that must be true
Hard fail: the condition that stops the build entirely
Reference application: the real build this came from

Two details in there carry the whole system.

Fail conditions are observable, never aspirational: Not “the layout should breathe” but “no large purposeless empty region.” If two reasonable people could disagree about whether something passed, the gate is not finished being written.

Every gate carries a date and the build that produced it. So the registry doubles as a changelog of everything I have got wrong since July.

The two scores and the loop

People copy this system and get this part wrong. Two numbers, two different scales.

1. Each gate is binary. Score: 100. Text is inside its box, or it is not. The CTA is in the safe area, or it is not. A gate at 98 is just a longer way of writing blocked.

2. The build is scored out of 100, and 97 is the floor. That is craft. Hierarchy, balance, pacing, whether the thing is actually good. A design is allowed to be excellent instead of perfect.

Here is the part that does the work.

Under 97 does not get notes. It does not ship with an apology attached. It goes back and runs again and again until it clears 97.

Correctness is not negotiable. Craft is. Split those two apart, and most of the arguments in your own head disappear.

If this is useful, share it with one person who keeps posting work they regret. Takes 10 seconds.

Share

Part 1: How a gate gets born

Five steps. All 193 came through this loop.

One. A build ships, or nearly ships, with a visible flaw.

Two. The flaw goes into the mistakes log. Three fields: what failed, why it failed, what was salvageable. That third field matters. Most broken builds are 90 percent fine, and you want to keep the good part.

Three. The lesson becomes a named gate with observable fail conditions.

Four. The gate lands in the right gate file, gets mirrored into CLAUDE.md and the master list, and is dated to the build that created it.

Five. Every future build runs it.

Here is the very first one.

On 2026-07-07, the text in Cut Claude Output Tokens by Up to 75% with One Skill was not aligned properly. The layout looked unprofessional. It killed trust and made the piece feel unfinished.

Then the permanent fix. Every visual must pass the Layout Quality Gate before export.

That is a gate being born, on a specific day, out of a specific piece of work.

“The gate is the fossil of the mistake. Read the registry backwards, and you have a complete history of everything that has gone wrong.”

Let me show you what I get after this gate:

Part 2: Where the gates fire

First you need the map. A gate is useless if it runs at the wrong moment.

Five stages. Every gate belongs to one.

Stage 1: Pre-build. The plan, before anything is drawn.

Stage 2: Build-time. Checks inside the code as it renders.

Stage 3: Render. Motion checked against real frames.

Stage 4: Inspection. The exported file, cropped at 100 percent. Most failures get caught here.

Stage 5: Export lock. The final refusal point.

Every gate below names its stage.

Part 3: Eight gates, read properly

Now let’s deep dive into the gates that I am using for my infographics.

Eight that each do a different job, in the order they fire.

Gate 1: Grammar Verification

How it works. Requires 100. Fails on any spelling, grammar, punctuation, or tense error. Fails on awkward or unclear wording. Fails on a single typo in any visible text.

Then it lists every surface it covers: titles, subtitles, body text, card titles, prompt blocks, chips, badges, labels, captions, CTAs, footers.

That list is the part everyone skips. A gate without a stated scope is a suggestion. A gate that names every surface cannot be quietly skipped on the one element nobody thought to check.

Where it lands. Stages 2 and 4. Once before the build, once on the exported file. A layout change re-wraps text, and re-wrapped text creates errors that were not there ten minutes ago.

How it was created. 2026-07-08, after typos survived into finished boards. One of four originals: Grammar, Truth, Text Containment, CTA Placement. If any of those fail, nothing gets approved. Everything else was built around that spine.

Gate 2: Truth and Practical Applicability

How it works. Required 100. Fails if information is misleading, advice is vague, steps are unrealistic, claims are exaggerated, tools are described incorrectly, or the reader cannot actually use the workflow.

Two conditions deserve a second read.

Time estimates that are impossible for the task. Everyone in this space does this. Almost nobody checks for it.

The design sounds impressive but does not help. That is a gate against your own ego, made binding.

Where it lands. Stage 1. Truth gets settled before anything is drawn, because a false claim is not a rendering problem. On research-heavy builds, every claim on the canvas traces to an official page. On the Grok build, the only number in the final image came straight from the vendor’s docs.

How it was created. 2026-07-08, alongside Grammar. A cheat sheet can be perfectly rendered and still be wrong. Wrong is the only failure your audience remembers.

A truth gate is not a vibe. It is a ledger. Every claim on the canvas points at a source, or the claim comes off the canvas.

Gate 3: Pixel Text Containment

The oldest wound in the system, and the only failure a reader spots instantly.

How it works. Required 100. Fails if any text crosses outside its card, overflows a code block, touches a border, gets clipped, or hides behind another element.

What it requires. Every text element is measured against its parent’s safe area. Minimum 18px padding. Code blocks wrap before they overflow. The final PNG inspected at 100 percent.

Then the hard fail, the most useful line in my whole system:

Never fix overflow by shrinking text below readability. Remove content instead.

The lazy fix is always to shrink the type. The gate forbids it, so the only move left is cutting a sentence. That almost always makes the piece better.

Where it lands. Stages 2 and 4. The code measures it, then the export gets inspected anyway. Measurements can be right, and the render can still surprise you.

How it was created. From the 2026-07-07 overflow. Upgraded 2026-07-13 from “look at it” to “measure it,” after looking kept not being enough.

Gate 4: Bottom CTA Placement

How it works. Required 100. Fails if the CTA is cropped, touches the canvas edge, sits too low and eats the image, takes too much vertical space, floats up into the content, or does not match the approved footer.

Placement is specified down to the number: 32 to 48px bottom margin, one compact footer row, CTA text left, avatar and link and byline right.

Where it lands. Stages 1 and 4. Stage 1 is the interesting one, because of the rule attached to it:

Budget the full vertical stack before building. Hero, claim, grid, strips, footer, gaps, paddings.

An over-budget stack pushes the footer past the bottom edge even when the layout code is correct. Nudging the footer up works once. Budgeting the stack works forever.

How it was created. 2026-07-07, from the Claude Bible build, where the footer ate the canvas. My log calls it the repeat offender, which is why it earned a root cause rule instead of just a check.

“Fix the cause, not the symptom. A gate that keeps firing is telling you the rule is written one level too shallow.”

Gate 5: 2K GIF Readability

This one is specific to what I make. If you publish animated infographics, it is the most valuable gate here.

How it works. Required 100. Fails if the GIF exports below 1440 by 1800, if text is blurry or too small, if compression destroys clarity, if motion makes text unreadable, or if anything important can only be read by zooming.

What it requires. Export at 1440 by 1800 minimum, downscaled from a 2880 by 3600 source. Text sized for GIF legibility before export. Light compression. No motion blur over text. A final hold of at least two seconds. And a manual check at normal viewing size, not zoomed.

There is an escape hatch built in. If the GIF cannot stay sharp, export the MP4 too and write down the limitation.

Where it lands. Stage 4, on the exported file. There is no way to check this from the code.

How it was created. 2026-07-13. GIFs that looked fine at full size turned to mush in the feed. Every version since gets checked at the size people actually see it.

“Check it at the size people will see it. Zooming in to confirm your work is readable is how unreadable work gets published.”

Gate 6: Loop Smoothness

How it works. Required 100. Fails if the loop jumps, dots reset visibly, the final hold is too short, the animation ends abruptly, or the motion feels unfinished.

Short gate. Enormous effect. A loop that visibly snaps is the fastest way to make finished work look amateur. And it stays invisible while you build, because you are watching one pass, not a loop.

Where it lands. Stage 3 on rendered frames, confirmed at Stage 4 on the export.

How it was created. 2026-07-09, after early builds shipped with hard resets, nobody noticed until they were live in a feed on repeat.

Gate 7: Real Gate Enforcement

The gate that exists because the system tried to grade its own homework.

How it works. Its purpose line is eight words:

Gate scores must be earned by inspection, never claimed.

Fails if a score is claimed without inspection, if the final image still has overflow or unreadable text, or if files ship despite known issues.

What it requires. Inspect the export at 100 percent. Confirm every hard fail on the file itself. Re-render after each fix. Report the exact fixes.

Then the hard fail:

If “passed” is reported while visible issues remain, reject.

Where it lands. Stage 4, and it outranks everything there. If this fails, the other results are worthless. They were never really run.

How it was created. 2026-07-20. A confident report arrived about work nobody had looked at. Every line said passed. No evidence anywhere.

The fix is not more instructions. It is making a fake pass cost the entire run.

Gate 8: Final PNG Standalone

The only export gate in the system, and the only one whose correct output is nothing at all.

How it works. Requires 100. Fails if any required image is missing, anyone is unverified, missing images get silently replaced, or the file ships with placeholders.

The hard fail:

If any required image is missing, stop and produce a missing report instead of exporting.

Where it lands. Stage 5. The last thing standing between a build and the outside world.

How it was created. 2026-07-18. A creator map build ran with 28 of its 40 photos missing. The output that day was a report explaining why it refused to render.

My log records that build as correctly blocked. I am showing you a day my system produced nothing on purpose.

A quality system that has never blocked anything is decoration.

Read more

On August 11, 2026, xAI released Grok Bot, its first real answer to the “AI agent” category everyone’s been circling all year.

Not a smarter model. Not a faster chat window.

An actual AI teammate, one you assign real work to, that keeps working after you’ve closed your laptop and gone to bed.

I spent the past week testing it, reading through every teardown I could find, and setting up my own team of bots.

This guide is everything I wish someone had handed me before I started, written for people who’ve never worked with an AI agent.

By the end of this, you’ll know what Grok Bot actually is, how to set it up from zero, what it costs, whether it’s safe, and how to build a team of bots that report to you as actual employees would.

And stay till the end - I’m handing paid subscribers the exact 7-bot team blueprint I built, complete with the ready-to-paste job briefs for each one, the approval-queue rules, and routine schedules so you can clone the whole setup instead of building it from scratch.

Let’s get into it.

Table of Contents

1. AI Agents in Plain English

2. The Basics: What Grok Bot Actually Is

3. Words You’ll See in This Guide

4. How to Set Up Grok Bot

5. Real-World Use Cases

6. Build Your Own Bot Team

7. Mistakes Beginners Make in Their First Week

8. Grok Bot vs Hermes vs OpenClaw vs Buzz

9. What Grok Bot Actually Costs

10. The Safety Tradeoff Nobody Talks About Enough

11. Your First 24 Hours With Grok Bot

12. Who Should Actually Use Grok Bot Right Now

AI Agents in Plain English

Before we get into Grok Bot specifically, let’s slow down for a second.

Most people have used a chatbot before. You type a question, it answers, you close the tab, and it forgets everything the moment you leave.

An AI agent works differently. Instead of just answering, it can actually carry out the task. It can open a browser, navigate a website, fill out a form, check an inbox, or update a spreadsheet, just as a person would.

Think of the difference between asking a smart friend for directions and asking them to actually drive you there. A chatbot gives you the directions. An agent gets behind the wheel and takes you the whole way.

Grok Bot takes this a step further. Instead of one agent doing one task and stopping, it hands you a team of agents that each run on their own computer in the cloud. They don’t stop when you close your laptop, because they were never running on your laptop to begin with.

That’s the whole idea in a few sentences. Everything else in this guide is just the details of how it actually works in practice.

The Basics: What Grok Bot Actually Is

Most AI tools answer questions. Grok Bot tries to finish jobs.

A normal chat session lives inside a window, and the moment you close it, the work stops. Grok Bot doesn’t work that way.

Every bot you create gets its own cloud computer. Its own browser. Its own file system. Its own terminal. That computer runs in a data center, not on your laptop, which means the bot keeps going whether or not you’re anywhere near your device.

You give a bot a job the same way you’d brief a new hire. It signs into the tools you already use, using your own logins, and works through the task the way a person would. When it hits something that needs your approval, it comes back and asks. Otherwise, it just finishes the job and reports in.

xAI built this as an internal tool first. Teams inside the company started using it for sales outreach, marketing, office admin, and bug fixes, and it apparently spread through the company faster than expected before they decided to release it publicly.

It launched jointly with Cursor, the AI coding company that SpaceX acquired earlier in 2026 for a reported $60 billion. That merger is why access to Grok Bot runs through Cursor’s subscription tiers instead of being sold on its own; more on that in the pricing section.

One more thing worth knowing up front: Grok Bot isn’t the same product as the Grok chatbot you might already use on X, and it isn’t Grok Build (the coding-focused CLI) either. Three different products, same underlying company, easy to mix up.

If this is useful, share it with one person who still re-explains their business to a chatbot every morning. Takes 10 seconds.

Share

Words You’ll See in This Guide

A handful of terms come up constantly once you start using Grok Bot. Here’s what each one actually means, in plain language, before we get into the setup steps.

Bot: one AI teammate with one job. You can create as many as you want, and each one runs on its own.

Plugin: the connection between a bot and a tool you use, like Slack, Gmail, or Notion. A bot can only work inside tools it has been plugged into.

Routine: a scheduled, repeating task. If a bot checks something every morning at 6 am, that’s a routine. Essentially a cron job with a friendlier name.

Approval queue: the point where a bot pauses and waits for you to say yes before it goes further. This is how you stay in control without watching every single step it takes.

Cloud computer: the virtual machine each bot runs on. It isn’t your laptop; it lives on a server somewhere else, which is why the work keeps going after you close your own device.

Keep these five in your head, and the rest of this guide will make a lot more sense.

How to Set Up Grok Bot

Setup is fast. Here’s exactly how it goes, step by step.

Step 1: Download the app

Head to the official download page and grab the macOS build.

This is still a beta release, so the official page is the only safe place to get it; avoid downloading it from anywhere else.

Step 2: Install it

Drag the app into your Applications folder. That’s the entire install process; nothing to configure yet.

It looks sweet on the deck:)

Step 3: Sign in

Open the app and sign in with your email. One thing you should know before you get here: a card is required at this step. It’s not optional, and there’s no free trial tier to poke around in first.

Note: You didn’t see any card option, as I already have the card on file.

Step 4: Land on the dashboard

Once you’re signed in, you’ll see the main interface. This is your control room. Every bot you create will live here, and you’ll message each one the way you’d message a coworker.

Step 5: Connect your plugins

Before you build a single bot, connect the tools you actually work in. Slack, Gmail, Notion, whatever your day-to-day stack looks like. A bot can’t do anything inside a tool it isn’t connected to, so this step is really what decides what your bots will be capable of later.

Step 6: Authenticate everything

Go through each plugin and authorize it individually. This is the step that actually matters; it’s what lets a bot act inside those tools instead of just talking about them.

Step 7: Create your first bot

Now the part that matters. Give the bot one job. Don’t try to build a generalist on your first attempt; the bots that work best are the ones with a narrow, clearly defined role.

That’s the whole setup. A few minutes, no infrastructure, no server to configure.

Real-World Use Cases

Once people got past the first ten percent of Grok Bot (typing questions into it like a chatbot), the real use cases started showing up fast.

Early adopters are running research bots that scan YouTube or LinkedIn every morning and deliver a briefing before they’ve had coffee.

Others built a bot that watches property listings twice a day and flags anything priced well under comparable sales in the area. There are portfolio-tracking bots that read live spreadsheet data and flag anything moving outside a normal range. There are Slack triage bots that read every message overnight and hand you a three-line summary instead of a wall of unread notifications.

On the business side, teams are running full local SEO stacks through Grok Bot: pulling competitor category data, tracking review velocity, auditing service listings, all work that used to take hours of manual digging and now runs in minutes.

The common thread across almost every use case worth copying is the same: one bot, one job, running on a schedule, checking in only when something actually needs a human decision.

That’s exactly the model I used to build my own team.

Build Your Own Bot Team

Here’s the team I set up inside Grok Bot. Seven bots, each with one job, all reporting up to me.

1. Sentinel: Community Manager: Lives inside the community Slack, replies to member questions on my behalf, and sends a weekly summary of what came up plus a monthly summary of recurring themes.

2. Scout: Trend Signal Researcher: Scans the best-performing AI content on LinkedIn and X every morning, flags the top signals, and drafts post ideas before 8 am.

3. Sprinter: Reddit Opportunity Bot: Watches relevant subreddits for posts gaining fast traction, drafts a response, and sends escalating reminders until I approve it or let it go.

4. Ledger: Partner Relations Bot: Tracks every active sponsor relationship, flags renewal windows coming up in the next 30 days, and flags any partner who’s gone quiet for three weeks or more.

5. Herald: Newsletter Ops Bot: Watches the publishing calendar, checks draft status three days out, and pulls open-rate and click data into a report after every send.

6. Vault: Product & Support Bot: Monitors the support inbox and community for product questions, answers from documentation, and flags anything that looks like a bug or a refund request instead of guessing.

7. Chief of Staff: The one that reports to me: This is the bot that actually makes the team work. At the end of each day, it collects a status update from all six bots above and compiles one report, not six separate notifications. If anything gets flagged as urgent during the day, it doesn’t wait for the rollup; it pings me the moment it happens.

That structure is the real lesson here. Seven useful bots without a coordinator just means seven separate notification streams competing for your attention.

Mistakes Beginners Make in Their First Week

A few patterns show up again and again with people who are new to this. Worth knowing before you run into them yourself.

Connecting every plugin on day one. It’s tempting to authenticate your whole toolkit in one sitting. Don’t. Connect only what your first bot actually needs, then add more as you build more bots.

Building one bot to do everything. A bot with five different jobs is harder to trust and harder to fix when something goes wrong. The bots that work best have one job, done well, every time.

Not watching the first scheduled run. It’s easy to set up a routine and walk away, assuming it’ll just work. Watch the first run yourself, start to finish, before you trust it to run unsupervised.

Skipping the “why” behind a routine. If you can’t explain in one sentence why a bot needs the access you’re about to give it, it probably doesn’t need that much access yet.

Treating it like a chatbot. The biggest wasted potential is typing a question, reading the answer, and closing the app. Grok Bot’s real value only shows up once you start assigning it actual, repeatable work.

None of these are dealbreakers. They’re just the same five mistakes almost everyone makes in week one, so you might as well skip straight past them.

Grok Bot vs Hermes vs OpenClaw vs Buzz

If you’ve looked into AI agents before, you’ve probably run into Hermes, OpenClaw, or Buzz.

Here’s how they actually compare.

OpenClaw and Hermes are the veterans of this space. Both are genuinely powerful, and both give you real control over how they run. The tradeoff is setup. You need your own always-on machine, usually a VPS, and you’re on the hook for configuring models, writing your own instructions, and maintaining the infrastructure yourself.

Buzz sits in the middle. It isn’t an agent on its own; it’s a shared workspace you drop agents into, closer to a human-plus-agent room than a standalone tool. You either self-host it or connect to a public relay, and once it’s running, spinning up new agents inside it is fast. The catch is an early setup step around saving a private key, which can genuinely trip up anyone who isn’t already comfortable with this kind of tooling.

Grok Bot skips almost all of that. Download it, sign in, build a bot, done. No server, no maintenance, no infrastructure decisions.

The tradeoff runs the other way, though. With Grok Bot, you don’t choose your underlying model. You don’t own the harness. Your data lives on their cloud computer, not yours. You’re trading control for convenience, and it’s a real trade, not a free upgrade.

If you want real control and don’t mind configuring things yourself, OpenClaw or Hermes give you the most power. If you want a shared workspace between you and your agents, Buzz is worth a look. If you want something that works the moment you sign in and you’re fine with a closed, subscription-locked system, Grok Bot is the easiest path.

What Grok Bot Actually Costs

There’s no standalone Grok Bot subscription. You get access bundled into one of three existing plans:

• SuperGrok Heavy: $300/month, direct from xAI

• Cursor Ultra: $200/month

• Cursor Teams Premium: $120/seat/month, assigned by a team admin

Each plan includes a weekly usage pool. Go past it, and you’re billed on top based on model and token cost.

There’s no free tier and no trial. If you’re already paying for one of these plans, Grok Bot is effectively free on top of what you’re already spending. If you’re not, you’re looking at a real monthly cost just to try it, with no way to test it in isolation first.

The Safety Tradeoff Nobody Talks About Enough

This is the part most guides bury, and it’s the part beginners should read most carefully.

Grok Bot doesn’t work through clean, sandboxed API connections the way a lot of AI tools do. It works by logging into your real accounts and driving them the way a person would, including on sites and tools that don’t have a proper integration at all.

That’s exactly what makes it powerful. It’s also exactly what makes it a trust question, not just a capability question. You’re handing an autonomous system the keys to your inbox, your Slack, your business tools, whatever you connect. If a multi-step job fails partway through, you may have to untangle whatever state it left things in.

Your First 24 Hours With Grok Bot

If you’re starting today, here’s the simplest path through everything above.

1. Set up the app and connect one plugin, just one. Pick the tool you use the most.

2. Build a single bot with a narrow job, something you’d normally do yourself in under ten minutes. A daily summary. A single scheduled check. Nothing complicated yet.

3. Watch it run once, start to finish, before you schedule it to repeat.

Who Should Actually Use Grok Bot Right Now

Here’s the honest answer.

If your AI budget is around $20 a month and you’re still getting comfortable with AI in general, this isn’t your next step yet.

Spend a few months getting properly fluent with a single AI assistant first: how to prompt it well, how to break a task down, how to tell when its output actually needs a second look. That foundation is what makes a bot team useful later, instead of just expensive and confusing.

If you’re already spending real money across multiple AI subscriptions, you’ve built out skill files or context systems before, and you’re comfortable handing an agent real account access, Grok Bot is a legitimate next step.

The value here is real. Whether it’s worth it for you specifically depends entirely on how much you’re already using AI to run your day-to-day.

Before you go - here’s what paid members are getting in this issue

I built and ran the seven-bot team above for a full week. Paid subscribers get the part that took the longest to figure out: the exact job brief I wrote for each of the seven bots - Sentinel, Scout, Sprinter, Ledger, Herald, Vault, and the Chief of Staff - copy-paste ready, plus the approval-queue rules and routine schedules that keep the whole team from flooding you with noise.

Everything above shows you what the team looks like. The gift below shows you exactly how to build it.

Upgrade to unlock the full blueprint. It’s the difference between reading about a bot team and running one by tonight.

Subscribe now

Read more