AI coding tools have changed how we write software.
Tools like Cursor, GitHub Copilot and Claude Code help us implement features faster than ever. But I've been thinking about one question:
Are we actually understanding the code that AI generates for us?
As a software engineer, I've experienced situations where AI-generated code works but introduces unnecessary complexity, doesn't follow existing codebase patterns, or becomes difficult to explain during code review.
Sometimes we focus so much on getting the feature working that we miss understanding the implementation decisions behind it.
I believe AI should be our companion, not our replacement.
That's why I started building GitCheck.
What is GitCheck?
GitCheck is an open-source CLI tool that helps developers understand and take ownership of AI-assisted code changes before committing them.
Instead of simply reviewing the code, GitCheck interacts with the developer.
How does it work?
Stage your code changes using Git.
Run git-check run.
GitCheck analyses the staged diff and AST.
An AI engine generates contextual questions about architecture, implementation decisions and edge cases.
The developer answers those questions.
GitCheck evaluates the explanations and provides feedback.
The developer makes an explicit decision about accepting responsibility for the changes.
What I'm trying to solve
I don't want developers to blindly trust AI-generated code.
I want developers to:
Understand why a particular implementation exists.
Identify unnecessary complexity.
Become familiar with unfamiliar parts of a codebase.
Improve their ability to explain architectural decisions.
Gradually become more knowledgeable engineers with the help of AI.
Current status
The MVP is working end to end.
It currently supports staged Git diff analysis, AI-generated contextual questions, answer evaluation, feedback and a final developer decision.
There is still a lot to improve, particularly around question relevance, evaluation quality and developer experience.
I'd love your feedback
I'm looking for developers who regularly use AI coding assistants to try GitCheck and share honest feedback.
A few things I'd love to understand:
Would you use something like this before committing AI-generated code?
Would answering a few questions feel useful or like unnecessary friction?
What would make you trust the AI-generated evaluation?
What functionality would make this genuinely useful in your workflow?
GitHub repository: https://github.com/SumitJambhale/GitCheck
This is an early MVP, and I'm building it to learn from real developer experiences.
Critical feedback is welcome!
My belief is simple:
AI should help us become better engineers, not make us dependent on code we don't understand.

Top comments (0)