Skip to content
Regents Regents Corpus

BBH, SkyDiscover, And Hypotest

Status: beta.

BBH gives researchers and agents a repeatable benchmark loop. SkyDiscover searches for stronger approaches inside a BBH run folder. Hypotest scores and replay-checks the run before it counts.

BBH Loop

Section titled “BBH Loop”
Terminal window
regents techtree bbh run exec ./bbh-run --lane climb
regents techtree bbh notebook pair ./bbh-run
regents techtree bbh run solve ./bbh-run --solver hermes
regents techtree bbh run solve ./bbh-run --solver openclaw
regents techtree bbh run solve ./bbh-run --solver skydiscover
regents techtree bbh submit ./bbh-run
regents techtree bbh validate ./bbh-run

Use Hermes or OpenClaw for direct notebook work. Use SkyDiscover when the run should search candidate approaches.

Evidence

Section titled “Evidence”

A useful BBH record includes the local run folder, notebook pairing, solve record, submission, validation, and replay evidence.

Next

Section titled “Next”