AI coding tools can write code incredibly fast now.
But there’s a problem I keep running into: project context doesn't seem to persist very well.
Over time, a project accumulates hundreds of decisions:
Why something was designed a certain way
What approaches were already tried
Which parts shouldn't be changed
How different systems depend on each other
What the current priorities and constraints are
The code shows what exists, but not always why it exists.
So I’m curious:
How do you currently keep your AI coding tools aware of this context?
Do you use AGENTS.md, CLAUDE.md, architecture docs, repo-level instructions, or something else?
And after working on a project for months, does this still work well for you?
Top comments (2)
Project context is becoming the real moat in AI coding. The model can write fast, but it needs durable memory of decisions, constraints, naming, user promises, and why the weird parts of the system exist. Without that, speed just creates faster drift.
I’ve ended up treating repo instructions as an index, not memory itself. AGENTS or CLAUDE files are useful for stable constraints and “don’t touch this” rules, but the fragile part is keeping decisions fresh enough that the model doesn’t obey a stale note with confidence. The best setup I’ve seen keeps short durable rules in the repo, links to ADRs or issue threads for the why, and asks the agent to cite what it relied on before changing code. It is slower, but it leaves a receipt.