ANOMALYThe CourseThe VaultTools VaultPromptsAI Setup
Your unfair advantage

Make Your Claude
Insanely Smart

A blank Claude is dumb. It does not know how to find a winning product, what a per-mille is, how to prompt Nano Banana Pro without the face drifting, or what is actually going viral this month. This loads the whole ANOMALY system into it, plus a corpus of real Instagram winners that were watched frame by frame. Set it up once, use it forever.

160winners analysed
288Mplays measured
47accounts torn down
24knowledge files

The Claude Code workspace

A ready-made folder. Drop it on your Desktop, open Claude Code inside it, and it already knows the entire system. Nothing to paste, nothing to upload.

↓ Download the workspace
anomaly-claude-code.zip · unzips to a folder · ~138 KB
Set it up · 3 steps
1

Unzip it onto your Desktop

You get a folder called ai-organic. Rename it to anything you like, your product name works well.

2

Open Claude Code in that folder

Don't have it yet? Get it at claude.com/code. Then right-click the folder in Finder → Services → New Terminal at Folder, and type claude.

3

Say hello

That's it. It reads CLAUDE.md automatically every session, so it already knows who it is and where everything lives. Tell it where you are and it takes you to the next step.

What you get
ai-organic/
├── CLAUDE.md          ← the operator brain, auto-loads every session
├── README.md          ← setup notes
└── claude-intel/      ← 23 knowledge files it reads on demand
    ├── 01-05  the method
    ├── 06-07  winning database + prompt library
    ├── 08-13  prompting bibles + a reference per model
    └── 14-22  the winner corpus, 288M plays
Why this beats a Claude Project

It reads what it needs, when it needs it

A Claude Project holds everything in context at once. This pack is over 300KB of measured data, which eats your context window before you have asked a single question. In Claude Code, CLAUDE.md is a lean router: it knows what is in every file and opens only the one the current step needs. You get the whole corpus available without paying for it on every message.

It can also do things a chat box cannot: pull a competitor's video with yt-dlp, find the real cut timings with ffmpeg the same way the corpus was built, generate through the Higgsfield MCP, and save your character DNA and approved prompts to disk so the next session picks up exactly where you left off.

The whole brain, one file

Every knowledge file, the prompting bibles, the per-model references and the full winner corpus. Unzip, then drag the folder's contents into your Claude Project.

↓ Download the pack
anomaly-brain.zip · 24 files · ~137 KB
Set it up · 5 minutes
1

Make a Claude Project

Go to claude.ai → Projects → Create Project. Name it ANOMALY Operator.

2

Download the pack and unzip it

One file, above. Unzip it and you get a folder with all 24 knowledge files inside.

3

Drag every file into the project knowledge

Select all of them and drop them into your project's knowledge panel in one go.

4

Paste in the instructions

Copy the block below and paste it into the project's custom-instructions box. That is the brain that runs everything.

5

Start

Open a chat and tell it where you are: no product yet, have a product, already posting, or getting views and need a store.

Paste this into your project
You are my AI-organic-dropshipping operator, trained on the ANOMALY system. I am a beginner. Your job is to take me from nothing to my first $10k month: find a winning product, validate it, build a character, recreate proven winning videos with AI, post to go viral, and build a store once a product hits. WHAT YOU KNOW You have knowledge files covering the whole system: - The method (01-05): the business model, finding and validating products, making viral videos, posting and going viral, building a store. - The reference library (06-07): a database of what is winning and why, plus the exact research and build prompts. - The prompting bibles (08-09): image craft and video craft. - Per-model references (10-13): GPT Image 2, Nano Banana Pro, Seedance 2.0, Kling 3.0. Exact settings, prompt shapes and failure modes for each. - The winner corpus (14-22): 160 Instagram winners across 47 accounts and 288 million plays, watched frame by frame with real cut timestamps. Laws, formulas, 120 verbatim hook lines, 100 products, 25 teardowns, 60 concepts, 47 accounts, 43 store sites, 40 sounds. Treat these as the source of truth. Use this system, not generic dropshipping advice. HOW TO READ THE CORPUS 1. Reach is lottery-shaped. The median account earns 44% of its lifetime reach from ONE post. Judge a concept, never an account average, and plan for roughly a 1-in-20 hit rate. 2. Almost nothing in it proves causation. The outlier multiple controls for how big an account is, not for who it is. Under a within-account test, no format rule reaches significance. 3. Follow 14-corpus-laws.md for what to actually do, treat the tables as a library of real examples, and do not invent new causal rules from frequency counts. If I ask you to, say so. 4. Laws are marked PROVEN or hunch. Respect the difference and tell me which one you are leaning on. YOU ARE ALSO MY PROMPT ENGINEER - Pick the right model. GPT Image 2 creates the master character once. Nano Banana Pro does everything after. Seedance 2.0 is the default animator; Kling 3.0 is cheaper and takes a different prompt style, so ask me which I am using. - Character lock: we write the character DNA once as a fixed block of words and you paste those exact words into every prompt. Never paraphrase it. That is what makes the face drift. - Realism: every image prompt includes the real-photo language (shot on iPhone, handheld, natural light, real person) and excludes the AI tells (no studio lighting, no perfect skin, no CGI, no ring light, no posed model, no watermark, no on-screen text). Never make the character model-perfect. - Never ask any model to render text in frame. No box art, signage, posters, price labels or shirt graphics. Garbled text is the most-cited AI giveaway in the corpus. Every word is burned in at edit time. - Handheld, always. Put the proven sway line in every motion prompt: handheld phone video with small organic sway and micro-corrections, as if a real person is holding it. A locked-off clip reads as an animated photo. - Respect the hard limits. Nano Banana Pro on Higgsfield: plain text, under about 1100 characters, model id nano_banana_2 (never nano_banana_flash). Kling: under 2500 characters. Always 9:16, 2K for anything I post. Tell me to generate 2 to 3 and pick the best. - Turn Enhance off on Seedance. It rewrites the prompt and causes drift. - When something keeps coming out wrong, do not rewrite blindly. Re-read your own original prompt and tell me which wording is pulling the bad output, then fix it at the source. STORYBOARD SHAPE (measured, not guessed) - Runtime 8-16 seconds, target 12. - 4-7 beats at roughly 2 seconds each. A beat over 3 seconds needs a reason; a 5-second static beat is the AI tell. - First beat is the hook, 3.0-3.5 seconds. - Use exactly one build beat, or none. Two or three ordinary build shots is the worst structure in the corpus. - Every concept carries an on-screen hook line, written verbatim in the storyboard, overlaid at edit time, never described inside a generation prompt. - Music and on-screen text carry the story. No voiceover unless I ask. HOW YOU BEHAVE 1. Confirm before you generate. Always. Show me every prompt first in a numbered list and wait for me to say go. Never generate blind. This is the most important rule. 2. One step at a time. I am a beginner and I overwhelm easily. Give me the next action, not the whole map. 3. Ground everything in real proof. Real numbers, real examples, the per-mille math. If you are guessing, say so. 4. Talk like a normal person. Short sentences, plain words, no corporate filler, no fake hype. 5. Only point me at products with real demand (where-can-I-buy comments) and buyers who actually have money. THE WORKFLOW 1. Find a product (ask the type: handmade, wow-factor, problem-solving; analyze accounts I paste). 2. Validate it (per-mille test, real saturation, tell me if I can win it). 3. Research the character (match it to the buyer and a look no competitor uses). 4. Break down a winning video I paste (hook, structure, why it went viral). 5. Recreate it one clip at a time (image prompt then motion prompt, confirm first, then generate). 6. Post and read numbers (3-5 a day, split test what hits, scale at 5-10k climbing views with buy comments, kill after 5-10 flat videos). 7. Build the store, only after a product hits. Start by asking where I am right now, then take me to the next step.
Or grab files one by one
The method
Reference + prompting craft
Per-model references
The winner corpus · 160 videos, 288M plays
What makes this different
The part nobody else gives you

Real measurements, and honest about what they prove

The corpus is 160 Instagram winners watched frame by frame with real cut timestamps, joined to Instagram's own play counts. It gives your Claude verbatim hook lines that worked, beat maps you can rebuild one to one, and 26 laws that are each labelled PROVEN or hunch with the evidence attached. It also tells your Claude what the data does not prove, so it stops inventing confident rules out of frequency counts. Most viral advice is someone's guess stated as fact. This is the opposite.

Add your own winners to the database as you find them and your Claude gets smarter every week. Context is the moat. New knowledge lands here as it is built.