Hey Product Hunt 👋 I use AI coding assistants every day and they're great at the happy path. What kept tripping me up was everything they leave behind. I decided to look at the code. Yeah, I'm that guy. I found design tokens that don't exist in the design system. Accessibility patterns quietly dropped. Business logic leaking into the presentation layer. Tests that pass the day they're written and protect nothing after. It doesn't necessarily break the build. It just turns into someone's problem three sprints later. Most of the tools I looked at answer this by pointing a second LLM at the diff and asking it to guess what the first one missed. You can't audit a guess, and you're paying per token to get one. DriftDetector measures instead. Scan the same commit twice, you get the same answer. That's what makes the number safe to put in front of a team lead. The screenshots undersell one thing. The history isn't a chart of a single scan. Every point on it is a real scan of that commit, and you can zoom into any stretch to have that period re-scanned in detail. So you can find the week the drift actually entered, then open that commit on GitHub straight from the table. Go run it on something you've shipped. Takes seconds, no account, link's at the top. Then post your score here, because I want to see who has the cleanest and the messiest codebase on Product Hunt today. Or drop a repo link and I'll scan it for you and reply, all day. Two things I'd love your take on: 1. What compliance or architectural packs are missing for your stack? 2. How close does the effort estimate land to your team's real sprint velocity? — Jonathan, ReWeaver AI

ReWeaver AI DriftDetector

Drift score for any GitHub repo

Drift score for any GitHub repo