How to Build Personal Context for AI

Most people's first instinct is to write a bio. A paragraph about who they are, what they do, maybe a note about communication preferences. Paste it into a system prompt and move on.

It works - briefly. The problem: a bio is prose, and prose is hard for an AI to use precisely. It has to interpret what matters, guess at structure, figure out relevance. The output is inconsistent.

Personal context done well is structured, layered, and built from sources that already reflect who you are - not from what you remember on a given afternoon.

What should personal context include?

Four things: who you are, how you work, what you're doing right now, and who and what surrounds you. Each updates at a different rate.

Identity - who you are

Name, role, company, location. Changes maybe once a year. Every AI tool needs this before it can do anything useful.

Profile - how you work

Professional background, expertise, communication style, the tools you use. Your professional operating manual. Changes when your role or focus shifts - not often.

Mind - what you're doing right now

Active projects, current priorities, decisions in progress. The most dynamic layer. What's true today may not be true in three weeks. Without it, an AI knows who you are but not what you're doing. That gap matters.

Environment - who and what surrounds you

The people you work with, companies you deal with, meetings on your calendar. What separates specific, useful answers from generic ones.

Do I need all four layers to start?

No. Start with Identity and Profile - enough to make a real difference immediately.

They're also the easiest to build. The information already exists in your LinkedIn profile, personal website, email signature. You're not creating anything new, just collecting and structuring what's there.

Mind and Environment come next:

• Mind requires some reflection - what are you actually focused on right now?
• Environment gets richer as you connect more sources (calendar, email, contacts)

A practical sequence: Identity and Profile before you open a tool. Mind before you start a project. Environment when you want AI help with relationships and scheduling.

How do I write a context file manually?

Structured Markdown. Not paragraphs. The goal is a file an AI reads like a form, not like a letter.

Here's the structure - based on how Marcin set his up before Unabyss existed:

## Identity
Name: Marcin
Role: CEO
Company: Unabyss
Location: Krakow, Poland

## Profile
Background: B2B SaaS founder, previously led GTM at two early-stage startups.
  First company exited in 2022.
Expertise: Go-to-market strategy, content marketing, early-stage sales, MCP integrations
Tools: Notion, Linear, Slack, Claude, Cursor
Communication style: Direct, prefers written async over calls, no fluff

## Mind
Active projects: Unabyss post-launch content cluster, investor outreach for EUR 2M seed round
Current priorities: Content publishing, ICP validation, MCP guide distribution
Decisions in progress: Pricing model for Teams tier

## Environment
Key contacts: Stas (CTO), Filip (CRO), Dominik (CPO/CMO), Mateusz (backend)
Companies: Unabyss (own), several SaaS pilot customers in EU
Recurring meetings: Weekly team standup, biweekly investor check-ins

Three rules

• Short fields, not paragraphs. "B2B SaaS founder, previously led GTM at two early-stage startups" beats two sentences of narrative.
• Specific over general. "Go-to-market strategy, content marketing, early-stage sales" beats "business development."
• Mind is a snapshot, not a diary. Current state only. If it happened last quarter, cut it.

First setup: 30-45 minutes. Updates after that: minutes.

What sources should I build from?

Extract from sources that already reflect who you are. They're more accurate than what you'd write from scratch and stay more current.

For Identity and Profile

• LinkedIn (role, background, expertise)
• Personal website or portfolio
• GitHub activity - reflects your actual stack better than a self-reported list
• Old CVs or bios

For Mind

• Notion or your project management tool
• Recent documents you've been working in
• Your own notes about what matters right now

For Environment

• Google Calendar (who you're meeting, how often)
• Gmail or Slack (who you communicate with, which companies)
• CRM if you have one

One thing worth doing: cross-reference across sources. LinkedIn says one thing about your role, Notion says another - that's context drift. The version that reflects what you're actually doing is the one to keep.

How do I keep it current?

The failure mode everyone hits

People set up Identity and Profile carefully, then never touch Mind again. After a month, the AI has accurate background but stale operational context - it knows who you are, not what you're doing.

A 5-minute weekly review fixes this. Add new projects, drop completed ones, update priorities.

Is there a faster way?

Yes. Two limitations with the manual approach: initial setup takes time, and maintenance requires discipline.

Alternative: connect your sources and let extraction handle it. Unabyss connects to LinkedIn, Notion, Gmail, Google Calendar, GitHub, and others - extracts the data, cross-references it, and produces a structured vault automatically in under 90 seconds.

The practical difference:

• Profile built from what you've actually done, not what you remember
• Environment that reflects who you're meeting with this week, not last month
• Source conflicts flagged automatically

Mind still benefits from manual input - it's intentional and often contains things that don't live anywhere connected. But the foundation is handled.

The context is yours either way. Any AI tool can use it: via MCP for tools that support it, via export for everything else.

Set up your context vault with Unabyss

Next steps

• Connect your context to Claude or Cursor → How to Load Your Personal Context via MCP
• All the ways context reaches AI tools → How to Deliver Personal Context to AI Tools
• Back to basics → What Is Personal Context for AI?