Category: zk-SQL proving system / verifiable data coprocessor / query-proof infrastructure
Summary: Proof of SQL is best understood as the domain-specific proving system inside the broader Space and Time stack, not just as a feature bullet under the parent chain. Its reusable mechanism is a proof pipeline for SQL queries over cryptographically committed tables: query plans are compiled against tamperproof datasets, witness columns are generated alongside execution, commitments and polynomial constraints are built over those witnesses, and a verifier can check both data integrity and query correctness without re-running the query. That makes it a useful comparison class against general-purpose zkVMs like RISC Zero or SP1: those systems prove arbitrary programs, while Proof of SQL specializes in structured analytics, lower verification costs for query proofs, and developer flows where smart contracts or apps ask for proven SQL rather than proving arbitrary code.
What it does:
Proves that SQL queries were executed correctly against committed data rather than asking contracts or clients to recompute the whole query
Targets both onchain and offchain verification paths, including smart-contract callbacks that consume proven query results
Uses a domain-specific proving flow with query parsing, witness generation, commitment opening, and multilinear sum-check style verification over SQL execution traces
Supports proofs over indexed blockchain data and custom offchain datasets that have been committed into the Space and Time stack
Ships as open-source Rust infrastructure with GPU-oriented proving paths and explicit benchmarks around analytic-query performance
Key claims:
The official README describes Proof of SQL as a high-performance ZK prover that can cryptographically guarantee SQL queries were computed accurately against untampered data and claims sub-second proving for some 1M+ row workloads
The technical docs say a single proof simultaneously attests both that the underlying table data were not manipulated and that the SQL computation itself was executed correctly
The technical docs claim verifier-contract costs under roughly 150k gas, which is materially different from proving stacks that optimize for arbitrary-program flexibility over query-specific verification cost
The smart-contract query flow shows a distinct application control plane: developers request a proof plan, submit it through a QueryRouter path with payment and callback parameters, and consume a proven table result in a contract callback
The repo README explicitly frames the system as faster than generic zkVM or coprocessor approaches for structured SQL workloads because it uses SQL-specific arithmetization, native commitments, and GPU acceleration rather than arbitrary-instruction proving
Whitepaper: No standalone Proof of SQL whitepaper was surfaced during this pass. The clearest primary-source packet is the repo README plus current Space and Time technical docs, alongside the broader Space and Time whitepaper; see ../whitepapers/proof-of-sql-primary-sources-2026-05-12.md.