๐Ÿ‘‹ Welcome! This is the actual system I use to run a multi-track content project with Claude โ€” across sessions, and across model versions โ€” without losing context or re-explaining myself every time.

I built this after realizing the real problem wasn't "the AI forgets things." It's that most people never built a system for what the AI should remember. This fixes that, and it works with any AI model โ€” Claude, ChatGPT, Gemini, whatever comes next.

Made by Ethanusis ยท pickenough.com


๐ŸŽฏ The Core Idea

Two documents, not one:

1. Master Doc โ€” current state only. What's done, what's next, what decisions were made and why (briefly). This is what you reattach every single session.

2. Archive Doc โ€” everything else. Detailed reasoning, old logs, finished discussions. You almost never open this โ€” it exists so the Master Doc doesn't get bloated.

The split matters because of one thing: AI context windows are limited, and models compress/summarize as conversations get long. Every time something gets summarized, small details can drift a little. If everything lives in one giant document, that drift compounds fast. Keeping the Master Doc lean is what keeps the drift low.


๐Ÿ“‹ How to Set It Up

  1. Create two documents (Notion, Google Doc, plain Markdown file โ€” doesn't matter): Master Doc and Archive Doc
  2. In the Master Doc, keep only:
  3. Everything else โ€” long explanations, finished debates, detailed research โ€” goes in the Archive Doc, with a short pointer left in the Master Doc ("full reasoning in Archive, section X")

๐Ÿ” The Session-Start Prompt

Paste this at the start of every new conversation, with both docs attached:

Here's my Master Doc [attached] and Archive Doc [attached].
Pick up from the "Next Steps" section and continue from there.
If anything in here seems outdated or inconsistent, flag it before
proceeding โ€” don't just assume it's still accurate.

That last line matters more than it looks. It's what catches version drift before it snowballs into a real mistake.