Most AI developer tools today focus on one goal: generating better code.
That’s an important problem, but I think another challenge is becoming just as significant.
Context doesn’t survive.
When you start a new AI session, you often have to explain your project again:
- Why was this system designed this way?
- What’s the current direction?
- Which architectural decisions are intentional?
- Which ones should probably be revisited?
Git tracks source code.
Documentation explains isolated concepts.
Chat history disappears.
None of them provide a persistent understanding of a project’s reasoning.
That’s why I’m building Contorium.
The core idea is straightforward:
Git stores code history. Contorium stores project intelligence.
Project intelligence includes things like:
- Decisions
- Design rationale
- Assumptions
- Project state
- Timeline
- Relationships between decisions
- Knowledge validity
One feature I’m exploring is Decision Lifecycle.
Instead of recording an architectural decision once and forgetting about it, Contorium tracks whether that decision is still trustworthy as the project evolves.
A decision can move from VALID to WARNING, then to NEEDS_REVALIDATION if the assumptions behind it are no longer true.
This isn’t about replacing developers or making autonomous decisions.
It’s about giving both humans and AI a persistent memory of the project.
As AI becomes a permanent part of software development, I think project memory will become just as important as code generation.
That’s the direction I’m exploring with Contorium.
https://www.contorium.dev/
https://github.com/ContoriumLabs/contorium
Top comments (0)