Dealing with legacy code is painful.
You fix one thing, and something else breaks.
Small changes often require understanding a large, messy codebase.
I ran into this problem repeatedly, so I built something to help.
What I built
I built EIDOS, an AI tool that fixes messy legacy code automatically.
Most AI coding tools just suggest changes.
EIDOS takes a different approach:
it actually edits your real files and iterates on them until the issue is resolved.
What it can do
- Fix bugs (e.g., SQL injection, type errors)
- Refactor messy code into a cleaner structure
- Add features directly into existing code
- Generate docs and basic tests
The goal is not just suggestions, but actually modifying real code.
Demo
Here’s a quick demo:
https://kaausia45-jpg.github.io/page/
Why I built this
I got tired of dealing with legacy systems where even small fixes felt expensive and risky.
I wanted something that could take over the repetitive and painful parts.
Looking for feedback
This is still early, and I'm exploring how far this approach can go.
If you've dealt with messy codebases:
- How do you usually handle them?
- Do you rely on tools, or mostly refactor manually?
Would love to hear your thoughts.
Top comments (1)
One thing I noticed is that even small fixes in legacy code can take way longer than expected.
Curious if others here have the same experience — what’s usually the hardest part for you?