Screening · the assessment
A first round built like the job.
One ~60-minute session in a real browser IDE, alongside Lyra. The candidate extends a realistic full-stack app — not a toy puzzle — and Chiron reads the code, the reasoning, and the AI direction at once.
Inside the assessment
Frictionless entry
No account, no password. A 6-digit email code and the candidate is in.
Lyra-led onboarding
A named AI host walks greet → consent → task brief. Expectations set, consent given, into the work.
A real workspace
Container-backed repo, database, and test runners. The clock starts only when everything is ready.
A deep assessment library
Frontend / backend / full-stack tracks, junior → staff: form engines, data tables, job queues, rate limiters, caches, outbox patterns.
Planted errors
Calibrated, subtle bugs that mimic real LLM failure modes — off-by-ones, wrong edge cases, hallucinated methods. Never syntax errors.
Durable & resumable
A pausable clock and graceful waiting screens. Disconnect or refresh and the work is safe.
An exam blueprint running underneath a natural pairing session — the assistant is rented and swappable; the Examiner is everything Chiron owns.