For labs

Professional work where models still fail.

Models have mapped the public internet and most of code. Legal, finance and support are where they still fail, and where nobody has built good graders. That work lives in the AI products professionals actually use.

Open sea seen from above
What we deliver

Curated datasets

Sessions with expert corrections, redacted with two independent passes and a human spot-check, and the source approves every sample before anyone sees it. Each dataset ships with a datasheet: source period, fields, outcome coverage, redaction method and measured redaction error rate, known limits, license scope.

Tasks with graders

We scope real work into tasks with hidden graders and reference answers. We try to break every grader with shortcut answers before it ships.

Recurring feeds

Only after a source is signed; volume and cadence quoted after a shape review.

How evaluation works

Samples go to named teams under evaluation-only terms, with a per-team copy we can trace. Rights reviewed with counsel before anything moves. Licenses set scope (train, eval or both), term and exclusivity.

Tell us the domain, the form and the volume you need.

Tell us what you need