Five programmes

Verification-first studio

For teams building agentic products where a correctness claim matters, in settings such as fintech, health operations, infrastructure, security, and autonomous coding.

It is a time-bounded programme with AKVC Labs: access to the verification platform, a claim ledger, formal specification of one to three core invariants, a dashboard of calibrated evidence, and an audit-log integration.

Verified-claims diligence

For organisations assessing companies that build AI agents. The deliverable is a claim-scoped report written by a person, not access to the scoring model.

Each claim in scope carries an evidence file, a calibrated support score, formal-check results where the claim can be formalised, analyst commentary, a statement of limitations, and a reference into the audit log. Scores are probabilistic assessments of specific claims. They are not guarantees, and they are not a rating of the company.

Evidence-weighted syndicate

Materials for a prospective deal include an evidence file and calibrated scores, so a reader can see the support behind specific claims.

Registering interest asks for information about the programme. It does not reserve a place, and this page does not solicit participation.

Infrastructure tollbooth

For builders of agent platforms that need verification, evidence scoring, and an audit layer. AKVC Labs runs that layer: hash-chained audit logs, proof-check services, and claim ledgers.

Proof bounties

For organisations and research groups with a claim they want machine-checked. A fixed-scope engagement run by AKVC Labs with formal-methods specialists.

The work is to formalise the claim, attempt a proof or a counterexample in Lean, and deliver the proof artefact with a written report.