Launching today
Steerhaus
Solve coding challenges by directing an AI.
5 followers
Solve coding challenges by directing an AI.
5 followers
The AI writes the code. You prove you know how to drive it.
Interactive
Free
Launch Team / Built With
5 followers
5 followers
Hey Product Hunt, I'm Andrei, the solo founder and only engineer behind Steerhaus. I designed, built, and shipped every part of it myself.
You don't write the code. You drive the AI and get scored on how well you steer.
It started from one question: technical interviews still make you write code by hand, but that is increasingly different from how many engineers work today. We collaborate with AI, but almost nothing measures how well we do that.
So I built Steerhaus.
You get a real bug in a real codebase. Instead of editing code yourself, you prompt an AI to fix it. Every attempt is verified against a hidden test suite running in an isolated Docker sandbox. Your score reflects not only whether you passed, but how effectively you guided the AI.
There are around 50 challenges, a global leaderboard, and a multiplayer mode where everyone solves the same bug.
It's free to try. I'm running everything on a small server for now, so there are a few daily limits to keep it online.
Try it at steerhaus.dev. It only takes a few seconds to get started.
I'll be here all day answering questions about the product, the scoring system, or how it's built. Thanks for checking it out!
Love the concept of making AI-generated code feel like it needs a real driver behind the wheel. One thing I'd love to see is a replay or diff view showing what the AI changed between attempts, so I can learn from the patterns it picks up and get better at steering it on the next run. Would make the whole thing feel way more educational too.
finally tried steerhaus and i love that it makes me actually think through what the AI is producing instead of just rubber stamping it. the little quiz after each generated snippet caught me off guard in a good way, felt like a real code review.
Really like the concept of making AI coding actually accountable. One thing that would help me trust the scores more is showing the exact diff or reasoning behind why a submission was marked down, so I know what to actually work on next instead of just seeing a number.