A way to keep your decisions and context in one place when working with AI.
Two markdown files. One holds stable information about your project (the Core). The other is a running log of what you've done and decided (the Addendum). Both travel with you — copy them into any AI chat, any model, any tool.
Every entry in the Addendum includes a short note about where the information came from: who was involved, why you made the decision, when it happened, and which AI model you were using. We call these receipts.
That's it. No apps, no accounts, no platform.
AI forgets everything between sessions. You end up re-explaining the same context, the same decisions, the same constraints — over and over.
These files fix that. Load them at the start of a session, and the AI knows what you're working on, what you've already tried, and what rules you've set for yourself.
When you switch models (ChatGPT to Claude, or vice versa), the context travels with you. When you come back to a project after weeks away, the history is there.
The receipts matter because memory drifts. Six months from now, you won't remember why you made a decision. The receipt tells you: this is what I knew, this is who I talked to, this is what the AI suggested, this is what I chose.
Your project's stable identity. Update this rarely — only when something fundamental changes.
What goes here:
- What the project is (one paragraph)
- Who it's for
- How you talk about it (voice, terms to use, terms to avoid)
- Constraints and guardrails
- Health and pace rules (so you don't burn out)
Think of this as the brief you'd hand someone joining your project. It should answer: what are we doing, why, for whom, and what are the rules?
Your running log. Update this often — after decisions, after sessions, after anything worth remembering.
What goes here:
- Dated entries (one date heading per day, numbered entries under it)
- Each entry has: what happened, why it matters, what comes next
- Receipts: who · why · source · date · model
Think of this as a work diary with timestamps and provenance. When you make a decision, log it. When you learn something, log it. When something fails, log it.
Starting a new project:
- Copy the Core template
- Fill in the sections (skip any that don't apply)
- Copy the Addendum template
- Make your first entry: "Project started. Here's what I'm trying to do."
Starting an AI session:
- Paste the Core into your chat (or attach it as a file, or add it to they AI project files)
- Paste the recent Addendum entries (you don't need the whole history every time)
- Work normally
After a session:
- If you made decisions or learned something important, add an Addendum entry
- If something in the Core changed (new constraint, new direction), update it
Switching AI models:
- Same process — paste both files into the new chat
- The context travels because it's in files you control
Every Addendum entry should note where the information came from. Keep it short.
The fields:
- Who: You, a collaborator, an AI, a source you read
- Why: What triggered this decision or entry
- Source: Link, document name, conversation reference
- Date: When this happened
- Model: Which AI you were using (if relevant)
Example:
Decided to use Babylon.js for the 3D engine.
Receipt: Me + Claude · researching WebXR options · Babylon.js docs comparison · 2025-01-15 · Claude Sonnet
You don't need every field every time. Use what's useful. The point is: future-you can trace back how you got here.
The Core + Addendum split works for most projects. But some work benefits from extra documents:
For coding/development projects:
- Add a Technical Core (stack decisions, architecture, dev rules)
- Add a Dev Log (session-by-session coding progress, dead ends)
- Keep strategic thinking in the main Core/Addendum; keep code details separate
For content projects (YouTube, writing, etc.):
- Add metrics to your Addendum entries (views, engagement, what you're measuring)
- Add review checkpoints (check results at 7 days, 30 days)
For outreach/pipeline work:
- Add a Tracker file (one entry per target/contact, status updates)
- The Addendum becomes your activity log; the Tracker becomes your CRM
General rule: Start with two files. Add more only when the original two get cluttered or when you're mixing very different types of work (strategy vs. code, for example).
Keep the Core short. If it's longer than 2-3 pages, you're probably including things that belong in the Addendum or a reference document.
Don't log everything. The Addendum is for decisions, lessons, and things you'll want to find later. Routine work doesn't need an entry.
Date your entries. You'll thank yourself later.
Update the Core when you learn something durable. If an Addendum entry keeps being relevant for weeks, it might belong in the Core.
It's okay to skip days. This isn't a diary. Log when there's something worth logging.
Use plain language. If you can't explain a decision simply, you might not understand it yet.
That's fine. The files are still there. When you come back to the project — weeks or months later — you'll have a record of where you left off.
The worst case is: you have some markdown files you don't use anymore.
The best case is: you never have to re-explain your project from scratch again.
Do I need special tools? No. Any text editor works. Any AI chat works. The files are plain markdown.
What if my AI already has memory features? Use both. Platform memory is controlled by the platform. These files are controlled by you. They also work across platforms, which built-in memory doesn't.
How long should entries be? Short enough to scan, long enough to understand later. A few sentences to a few paragraphs. If you're writing pages, consider splitting into multiple entries.
Can I change the format? Yes. These templates are starting points. Adapt them to how you work. The only thing that matters is: you can find what you need later, and you know where it came from.
- START.md — You're reading it
- Core_Template.md — Blank template for your project's stable identity
- Addendum_Template.md — Blank template for your running log
Copy them. Rename them for your project. Start writing.