alt Hacker News“Context rot” with AI coding tools is definitely real. After a few sessions the agent forgets earlier decisions, and you end up repeating the same planning conversation again.
Storing the plan and discussion as Markdown in Git is an interesting approach. It basically treats the agent’s reasoning as part of the project history, not just the final code.
Curious if others here are doing something similar to keep context across sessions.
Replies
I have a log of commits and the decisions and changes for each commit. I use it as a source of data/context when building out plans.
I agree with the other commenter who said they don't build anything without a plan. I would double down on that and say that you need to overplan, and regularly toss out plans as you do research/discovery.
[dead]
I'd been using this workflow for a while, but this post I found on HN a couple of weeks ago really solidified it: https://boristane.com/blog/how-i-use-claude-code/
> The workflow I’m going to describe has one core principle: never let Claude write code until you’ve reviewed and approved a written plan. This separation of planning and execution is the single most important thing I do. It prevents wasted effort, keeps me in control of architecture decisions, and produces significantly better results with minimal token usage than jumping straight to code.