Watch what actually happens when you open a new AI chat.

You re-explain who you are. What you're working on. Who it's for. How you write. What you already decided last week and don't want to relitigate. Then you ask your question — and by then half the conversation is spent on setup.

Then you switch models, or the thread gets too long, and you do the whole thing again from zero.

The fix isn't a longer prompt. It's writing your context down once, in a form you can paste anywhere.

A minimal version you can build today, in a plain text file:

1. Who you are — one paragraph. Role, what you actually do day to day, what you're responsible for.

2. What you're working on — the current project, its goal, its constraints, and the deadline if there is one.

3. How you communicate — the register you want back. Short or thorough. Blunt or diplomatic. British or American spelling. Whether you want options or a recommendation.

4. What's already decided — the things you do not want re-argued every session. "We're not doing X. That's settled."

5. Where you left off — three lines, updated at the end of a working session. What got done, what's next, what's blocking.

Paste sections 1-4 at the start of a new chat. Paste section 5 when you're resuming. That's it.

Two things that make this work better:

• Keep it in plain text or Markdown, not inside one assistant's memory. Anything stored inside a single product can't move with you.

• Write it for a competent stranger, not for a machine. If a new contractor could read it and get to work, a model can too.

And one boundary worth setting deliberately: decide up front what you will never put in a context file. Client names, unreleased work, financials, anything you'd regret pasting into a tool you don't control. Write that rule down too.

Try the simple version above for a week. If you want a complete system for building reusable context, preserving your voice, moving work between AI tools, and keeping that context current, that’s exactly what we built Posoulo Context for.