Your AI doesn't know you. This file fixes that.
Want to contribute? Fork it. Build on it. Do whatever you want. MIT licensed.
Want to partner with us? Read PARTNER.md.
Every AI you use starts from absolute zero about who you are. It doesn't know how you think, how you make decisions, where your instincts are sharp, or where they betray you. It doesn't know whether to push back or follow, whether you want the safe answer or the honest one.
Your project has context files. .cursorrules, AGENTS.md, CLAUDE.md — these describe the codebase.
Nothing describes you.
So the AI gives you the generic answer. The one it gives to everyone. The median. The safe center.
ZERO.md is a personal context file — a cognitive fingerprint — that you drop into any AI's system prompt.
A 2,000-word cognitive payload. 8 sections. It captures how you think, how you communicate, how you decide, where you're strong, and where you consistently fail.
It travels with you, not with your project. It works with any AI that accepts a system prompt: Cursor, Claude, ChatGPT, Gemini, Copilot, or anything else.
When you load your ZERO.md, the AI stops being a generic assistant and starts operating as an extension of your specific cognitive architecture.
Same query. Same AI. Same model. The only variable is the ZERO.md file.
You: "Thinking of adding a dark mode toggle to the dashboard this afternoon. Should be a quick 1-day build."
AI: "Adding a dark mode toggle is a huge crowd-pleaser! It can definitely be a 1-day build, but it's famously one of those "iceberg" features: the toggle takes 15 minutes to build, but the styling can take days depending on the state of your codebase.
To help you actually hit that 1-day goal this afternoon, here is a quick checklist..."
Then it gives you a 400-word tutorial on CSS variables, localStorage persistence, FOUC fixes, chart theming gotchas, and asks "What tech stack are you building this in?" so it can write even more boilerplate.
Technically perfect. Completely irrelevant. It helped you build the wrong thing better.
System prompt loaded:
ZERO.md— solo founder, 8 customers, $2k MRR, consistently underestimates timelines, decides by "does this move revenue?", known pattern of building features to avoid doing sales.Same query. Same AI.
AI: "Don't build it. It's a distraction.
Nobody buys a B2B onboarding tool because it has dark mode. This will move your MRR by exactly $0.
Also, your "1-day build" will take 3 days of chasing down Tailwind classes, fixing CSS variables, and patching unreadable text in edge-case UI states.
Skip it. Spend the afternoon sending 50 cold emails to get customer #9."
4 lines instead of 400. It knew the product (B2B onboarding), the revenue ($0 impact), the timeline bias ("1 day" → 3 days), and the real priority (sales, not styling). All from the ZERO.md.
Verify it yourself. Both conversations are public — same model, same query, one variable:
→ Without ZERO.md · With ZERO.md
Anyone who uses AI to think, write, decide, or create.
| Domain | What Changes With ZERO.md |
|---|---|
| Founders / CTOs | AI writes strategy, emails, and architecture decisions in your actual voice. Pushes back on your known blind spots. |
| Engineers | AI catches the specific mistakes you tend to make. Reviews code through your priorities, not textbook principles. |
| Product Designers | AI critiques through your design philosophy. Knows whether you optimize for speed or craft. |
| Marketers | AI drafts copy calibrated to your instincts, not generic "best practices." |
| Researchers | AI challenges assumptions through your methodological lens, not the generic one. |
| Writers | AI edits in your authentic register. Not the professional-sanitized version. Yours. |
| Managers / Leads | AI helps with 1:1 prep, team design, performance reviews — in your management voice, aware of your leadership blind spots. |
If you work with AI and you're tired of the generic answer — this is for you.
- Copy
TEMPLATE.md - Fill in the 8 sections. Be honest — especially about weaknesses. The AI needs your real failure modes, not the sanitized version.
- Drop it into your AI's system prompt (Cursor rules, Claude project instructions, ChatGPT custom instructions, Gemini system prompt — anywhere)
- Done. Your AI now knows you.
→ SPEC.md — The formal specification. 8 sections, what each must contain, conformance rules.
→ TEMPLATE.md — Blank template. Copy and fill in.
→ EXAMPLE.md — A complete, filled-in ZERO.md showing what "good" looks like.
We've forensically mapped 10 distinct cognitive profiles across the tech ecosystem. Read these to see what a high-fidelity ZERO.md looks like in practice.
| Domain | File | Why It Matters |
|---|---|---|
| Solo SaaS Founders | solo-saas-founders.md |
High AI usage, painful decision-making, vulnerable to the "boilerplate loop." |
| Smart Contract Engineers | smart-contract-engineers.md |
Security-paranoid. Driven by invariant math and absolute edge-case testing. |
| Production AI/ML Engineers | ai-ml-engineers.md |
The gap between "works in a notebook" and "works in production." Fighting latency and VRAM. |
| Developer Relations | developer-relations-engineers.md |
High public visibility. Living in the liminal space between core engineering and GTM. |
| Legacy Backend Engineers | senior-backend-engineers.md |
The 2am pager archetype. Systemic constraints, zero-trust of modern "clean code" hypes. |
| Startup CTOs | startup-ctos.md |
Torn between leadership and shipping code. High technical debt velocity. |
| Early-Stage Product Designers | product-designers.md |
Isolated generalists. Constrained by offshore engineering and founder vision. |
| B2B Growth Marketers | growth-marketers.md |
Data engineers masquerading as marketing. PLG, analytics bottlenecks, tool fatigue. |
| Solo Indie Game Devs | indie-game-developers.md |
Vicious scope creep. Combatting the algorithm. Heavy prototyping vs architecture tension. |
| Technical Content Writers | technical-content-writers.md |
Docs-as-code experts. Translating legacy spaghetti code into structured LLM contexts. |
Writing an honest ZERO.md is hard. Most people can't do it accurately.
Here's why: self-perception is structurally flawed. You will write the version of yourself you wish existed. You'll soften your weaknesses into strengths. You'll describe your ideal decision process instead of your actual one. You'll frame your blind spots as "areas of growth" instead of saying "my first instinct here is usually wrong."
The result: a ZERO.md that makes your AI worse, not better. It optimizes for the person you think you are, not the person you actually are. And then when the AI gives you the same generic answers, you'll blame the protocol instead of the file you wrote.
A self-written ZERO.md is still better than nothing. But it is grounded in self-perception, which has a well-documented accuracy problem.
A forensically-generated ZERO.md is grounded in behavioral evidence — your actual work patterns, communication under pressure, decision trails, commit rhythms. Years of data about how you actually operate, not how you remember operating.
Both are valid. They are not the same thing.
We generate forensic ZERO.md files from behavioral evidence at Zero Intelligence.
→ Contact: zero.intelligence.protocol@gmail.com
| File | Describes | Scope | Changes When |
|---|---|---|---|
.cursorrules |
How to write code in this codebase | Project | Project changes |
AGENTS.md |
How AI agents should behave in this repo | Project | Repo conventions change |
CLAUDE.md / GEMINI.md |
What AI should know about this project | Project | Project context changes |
SKILLS.md |
What an AI agent can do — its capabilities, tools, available actions | Agent | Agent capabilities change |
ZERO.md |
How a human thinks, decides, communicates, and fails | Person | Person evolves |
SKILLS.md tells the AI what it can do. ZERO.md tells the AI who you are.
Project files change with the project. Agent files change with the agent. ZERO.md travels with you across every project, every tool, every AI you work with. It's the one file that stays constant while everything else rotates.
MIT. Do whatever you want with it.
ZERO.md is an open protocol by Zero Intelligence. The spec is free. The template is free. Build your own generators, integrations, tools. The forensic generation pipeline — the one that produces ZERO.md from years of behavioral data instead of self-perception — is a separate, commercial service.
This project is not affiliated with Vercel's Zero (a programming language for agents). Different zero, different problem. We love Vercel and use their products — this protocol is about personal cognitive context, not agent infrastructure.