`me.md`

Use AI to distill your digital life into a personal context document. Eliminate the cold-start problem forever.

The problem

Every new AI conversation starts from zero. You say "I'm a developer" and the AI doesn't know if you're a junior building your first React app or a twenty-year veteran who's founded three companies.

The quality of AI assistance is directly proportional to the quality of context it has.

Without context, AI defaults to generic. It over-explains things you already know. It under-explains things you don't. It suggests tools you'd never use and misses the ones you rely on. It writes in a voice that isn't yours.

You can fix this in every single conversation by re-explaining your background. Or you can build it once and let AI do the heavy lifting.

What is `me.md`?

A me.md is a personal context document — a structured markdown file that captures who you are: your history, skills, projects, beliefs, communication style, and current priorities. Provide it to AI as system context, and every conversation starts with understanding instead of ignorance.

Think of it as the document a trusted colleague would write after working closely with you for years, distilled into something an AI can consume in its context window.

The technique uses AI itself to build this document. You point it at your existing digital life — notes, journals, project files, emails, whatever you have — and it processes everything through a multi-stage distillation pipeline, progressively compressing thousands of files into a single, comprehensive profile.

What it is not

Not a CV. A CV is what you want employers to see. A me.md is what you'd want a brilliant assistant to know — including your weaknesses, contradictions, and how you actually think.
Not a diary. It captures the arc of your life, not every detail. Target: 3,000–8,000 words.
Not static. It should be refreshed as your life evolves.

The core idea: hierarchical distillation

Your digital life is too large to process in one go. A notes vault, email archive, or project collection might contain tens of thousands of files spanning decades. No AI context window can hold all of that.

The solution is a multi-stage pipeline where AI agents work at different levels of granularity, progressively compressing information until it fits a usable context budget:

Folder Annotation

Label your folders so downstream agents know what they're looking at and whose content it is.

↓

Inventory & Triage

Crawl your files. Classify by relevance. Build a manifest of what to process and what to skip.

↓

Folder-Level Extraction

Process each folder as a batch using parallel agents. Extract personal-context facts into folder summaries.

↓

Thematic Synthesis

Group folder summaries by theme. Merge overlapping information. Resolve contradictions and show evolution.

↓

Final Composition

Assemble into one coherent document. Edit for flow. Enforce the context budget.

Each stage reduces volume by roughly 10x. Thousands of files become dozens of summaries, which become a handful of thematic drafts, which become one document.

What it gives you

No more re-explaining. Your career history, tech stack, communication preferences — all there from message one.
Better judgement calls. The AI knows whether to explain a concept or assume you know it.
Continuity across tools. The same document works with any AI — paste it into a system prompt, attach it to a project, or include it in a CLAUDE.md file.
Personalised tone. The AI matches your communication style rather than defaulting to generic helpfulness.
Depth from breadth. The pipeline surfaces patterns you might not notice yourself — recurring themes, evolving interests, the thread connecting disparate projects.

What goes in it

The structure should emerge from your content, but these sections are a good starting point:

Identity Name, location, languages, key biographical facts

Timeline Major life chapters and turning points, chronologically

Career & Work Professional trajectory, roles, companies, achievements

Projects What you've built — active, completed, and abandoned

Skills & Expertise Technical and non-technical, with honest proficiency levels

Interests & Passions What you care about beyond work

Beliefs & Values Worldview, principles, strong opinions

Communication Style How you think, write, and prefer to be communicated with

Current Context What's active right now — projects, priorities, challenges

The hard parts

Building a me.md from a large corpus isn't just summarisation. There are real architectural challenges:

Signal vs noise. A journal entry about a life decision is worth more than a utility bill. The pipeline must triage and weight content by relevance.
Deduplication of meaning. The same project might appear in your journal, CV, a README, and an annual review. The pipeline must merge, not concatenate.
Context budget. The final document must be 3,000–8,000 words. That's a compression ratio of 1000:1 or more. Every sentence must earn its place.
Temporal coherence. Your life has a timeline. The document should capture how things evolved — career shifts, moves, changing interests — not just a flat snapshot.
Accuracy. With multiple summarisation layers, there's a risk of facts drifting, details being lost, or different periods being conflated. Each stage must preserve specifics.
Attribution. A folder of markdown files could be your writing, saved articles, course notes, or client work. Without knowing whose content it is, agents will misattribute.

Using your `me.md`

As custom instructions. Most AI platforms allow system prompts or custom instructions. Paste your me.md there and every conversation starts calibrated.
As a project context file. Tools like Claude Code support project-level context files (CLAUDE.md). Include your document so AI assistance is always personalised.
As conversation context. For one-off conversations, paste it at the start.
As a team resource. A redacted version helps AI tools understand each team member's expertise and communication preferences.

Privacy

Your me.md will contain personal information. If it stays local, include whatever you're comfortable with. If stored in cloud systems, consider maintaining two versions — a full version for local use and a redacted version for cloud contexts.

Keeping it current

A me.md decays. Stale context actively misleads. Update the Current Context section monthly, projects and skills quarterly, and do a full review annually. You don't need to re-run the whole pipeline — just edit the relevant sections.

Build yours

The guide walks through each pipeline stage in detail, with the techniques and prompts you need to process your own digital life into a me.md.

Read the guide