AI coding assessment
Hire engineers for how they work with AI.
Chiron is an AI coding assessment for the AI-pair era: candidates do real engineering work with an AI pair, and we score the code, the reasoning, and how they direct the AI.
Why a new screen
The work evolved. The technical screen didn't.
The work changed
Most engineers now write code with an AI pair every day. The screen that ignores that is testing a job nobody does anymore.
Banning AI tests the wrong thing
A closed-book leetcode round measures recall under surveillance — not whether someone can ship correct software with the tools they actually use.
Allowing AI isn't enough either
Letting candidates use AI only matters if you can see how they use it. Output alone can't tell a sharp operator from a lucky paste.
What an AI coding assessment should measure
Score the work, the reasoning, and how they direct the AI.
The code
Concurrency, retries, cancellation, edge cases — the things that file bug reports in production.
The reasoning
At checkpoints, the candidate walks through trade-offs. We score the alternatives they raised — and the ones they missed.
The AI direction
The AI pair produces some deliberately wrong code. We watch what the candidate accepts, rejects, or fixes. Disclosed up front.
The integrity check
When the code is strong but the reasoning is thin, the divergence shows. A built-in tell, not an afterthought.