Commit History
We offer enterprise-grade, proprietary, verified, provenance-clean training datasets and verifiable-reward (RLVR) environments built from real software commit history, pull requests, issues, tickets, and rich repository metadata.
Training Data
- 1,000,000,000+ tokens (o200k_base)
- 13,889 commits
- 43,635 source files at head
- Pull requests and full commit history per repository
- Linked issue and ticket metadata
- 185 MB of structured GitHub API metadata
Contact us via Signal/text: 650-880-9229
Verifiable Rewards (RLVR)
- 159,518 verification checks
- 43,288 verified passing-check rewards
- 65,568 path-level reward signals (26,257 positive)
- 10,866 commit-level reward signals
- 1,676 strict-verified commit rewards
Exclusive and non-exclusive licenses cover our compiled dataset packages, verification artifacts, manifests, and curation methodology, not third-party rights. Trained models and outputs remain yours.