Hi HackerNews! I've been building this product for 2(!) years now. It's called DiffEnder, and it's a way for humans to review code with the help of AI.
As AI agents increasingly write more code, we are transitioning to a reviewer role of work produced by AI. Code reviews are how we update our theories about how the codebase works, which in turn helps us write good prompts for AI agents. It’s a virtuous cycle.
On the other hand, I’ve seen friends vibe code their way to a v1 product, and hit a wall, because their codebase is a rat’s nest, and they have no working theory of its functionality. They can’t point the LLM to a handful of files, explain what they do, and ask for incremental changes. No more!
You can sign up today at https://diffender.com. Happy to answer any questions you may have!
