Roadmap
What's next.
Everything below is public. Help pick priorities by upvoting issues on GitHub.
Now · Q2 2026
Launch + 100 services
- Scale attestor set to 6
- Ship CLI 1.1 w/ GPU proof acceleration
- Onboard 40 services to Verified tier
Next · Q3 2026
Live inference attestation
- Per-request proof mode (not just offline eval)
- 3 new benchmark suites per month
- Publisher revenue share for referred traffic
Later · Q4 2026+
Decentralized governance
- DAO for benchmark admission
- SOC 2 Type II
- Cross-chain verifiers (Base, Arbitrum, Solana)
Active proposals
PMP-01
Quorum attestation · 3-of-5 attestors must agree for "sensitive" benchmarks (SWE-bench, BrowseComp). Proposed
PMP-02
Live-traffic proofs · Attest on a 1% sampled slice of production requests. Draft
PMP-03
Per-language leaderboards · Separate MBPP boards for Python/TS/Rust. Discussion
PMP-04
Confidential attestation · TEE-based runs for publishers who can't share transcripts (e.g., PII-laden). Discussion
Non-goals
- We won't take a commerce cut of listed services. Ever.
- We won't build closed/proprietary benchmarks that we alone can run.
- We won't bundle opinions into scores — the pinned methodology speaks for itself.