98.6%
Regression pass rate
Programs at a frontier AI lab.
DirectionalCUSTOMERS · PROOF, NOT POSTERS
Numbers, not adjectives.
Anonymized by company type. Specific on the work. The artifact, the metric, and the moment a decision moved — kept together.
PROGRAMS
Anonymized
Case studies attributed by company type, not by name.
PROOF
Specific
The work, the artifact, the metric that moved.
PATH
Private review
Public proof first. Private reference path on request.
Measured, not promised · evidence-backed proof console
What teams left with. In their numbers.
Three readings from three programs. Attributed by type, not by name.
4.2h
Time to approval
Reviewer pickup to sign-off, a regulated decisioning program.
Directional27
Incidents averted
Caught at the gate across an enterprise rollout.
DirectionalEvidence · reviewed runsWindow · buyer-definedBaseline · customer-owned
ON THE RECORD · ANONYMIZED BY TYPE
The work, the artifact, the metric that moved.
Four programs. Four readings. Attributed by company type — not by name. Deeper review available under a private reference path.
CASE · POST-TRAINING
A frontier AI lab
Specialist review on one record.
RLHF adjudication and regression capture stayed attached to the same packet. Eval findings compounded into the next release gate without being re-stated.
11,408
cases reviewed in one release
↳ EVIDENCE PACKET
CASE · CLINICAL OPS
An academic medical center
Second reads went to the right reader.
Uncertain inferences routed back to the radiologist qualified for the body region, with the rubric and override chain attached to the case.
4.2h
average pickup to sign-off
↳ REVIEW RECORD
CASE · RELEASE GOVERNANCE
A regulated decisioning program
Sign-off became a repeatable gate.
Release review moved from a screenshot debate to one inspectable record per run. Examiner questions get one answer per audit cycle.
27
incidents caught at the gate
↳ SIGNED PROOF
CASE · HUMANOID PROGRAM
A top-three humanoid program
Capture, review, export — one chain.
Demonstration capture, teleop review, and export-format lineage stayed attached to the same workflow that ships the policy.
98.6%
regression pass rate
↳ REPLAY SUITE
On the record
“Release review went from a three-week scramble to a repeatable gate. When an edge case slips, we catch it, we save it, and it never ships again.”
BRING THE WORK
Your release, your reviewer, your record.
Bring the release, review queue, or escaped issue that matters most. We will show how it runs — and what it leaves behind.