Primary paper
Proof of Compute Integrity
The paper describes signed compute events, off-chain event storage, compact on-chain commitments, Kaspa blockDAG anchoring, and later verification or lineage reconstruction.
Whitepaper and technical research
Proof should be inspectable.
PoCI is a framework for AI lineage, timestamp integrity, and compute provenance. The research is published so the architecture, boundaries, and implementation direction can be examined directly—not reduced to a marketing claim.
Core architecture
Four connected loops
01
Event creation
An AI or compute action is represented as a structured event with lineage and execution metadata.
02
Proof generation
The event is hashed and signed, creating a compact commitment without publishing the full private payload.
03
Kaspa anchoring
Selected hashes, signatures, and lineage commitments can be anchored to Kaspa's blockDAG.
04
Verification
Later observers can reconstruct lineage and verify that presented records match the original commitments.
What PoCI provides
Integrity evidence
- Tamper-evident event commitments
- Signed lineage between compute events
- Independent timestamp and anchoring paths
- Verification without exposing full event data
What PoCI does not claim
Important boundaries
- It does not prove an output is correct
- It does not prove a dataset is accurate or ethical
- It does not reveal private model reasoning
- It complements—not replaces—ZKML, hardware attestation, and formal verification
PoCI is being developed in public. Terminology, implementation details, and interfaces may evolve as the research is tested and the application matures.