Verdict · from Riken Patel's AI Lab

Ship AI changes without holding your breath.

Run real evals against a golden set — deterministic checks and an LLM-as-judge — then gate the release on the result. Edit any output and re-run to watch the gate flip.

Runs in-browser No sign-up Streams live
verdict.rikenpatel.work/demo
Evals today
0
Pass rate
0%
p95 latency
0.0s
0%
pass rate
last 12 runs · gate open
Deterministic + judge

Regex, JSON-schema, and string checks where the answer is exact; an LLM-as-judge only where it's subjective.

Release gate

Below your pass-rate threshold, the gate blocks the merge — exactly like a failing CI check.

Cost & latency too

A prompt that doubles tokens or p95 is a regression as real as a wrong answer — tracked every run.

Everything runs client-side

See it run for yourself.

No sign-up, no API key. Open the console and launch a real multi-agent run in your browser.

Launch live console