Why this repo exists

Every time I start an AI coding session, it wakes up knowing nothing about my project. Not what I decided last week. Not why I named things the way I did. Not which approach I already tried and threw out. So I re-explain. And the next session I re-explain again.

That got old.

The fix turned out to be simple, and it is not a smarter prompt or a bigger model. It is a place to remember. The context, the decisions, the conventions: all of it lives in the repository as plain markdown, indexed so an agent can find what it needs instead of re-deriving it. The repository is the memory. When I switch tools, the memory comes along, because it never left my repo.

I have been working this way across a playtest tool I build and some small games, mostly alone. Over time it turned into ten habits that reinforce each other: committed memory, a personal layer that spans projects, capturing gotchas the moment they bite, researching before the hard-to-reverse calls, letting the agent grill my plans instead of agreeing with them, reviewing with several agents that each look for something different, and treating every settled decision as a cache entry that makes the next review cheaper. That last one is the piece I care about most. The more you write down, the less each future change costs to review.

This repo is where I am putting the method in public. It is not a product. There is no service to sign up for and nothing to install. It is files you can read, copy, and keep.

I do not have all the answers. AI gets things wrong, the network takes real upkeep, and the human is still the bottleneck by design. I am writing this down because it works for me, and because writing it down is the whole point.

More to come. If you want to follow along, the feed is in the README.