Public claims should carry their evidence with them.
QueryRook research is our proof engine: papers, claims, benchmark runs, artifacts, limitations, and reproducibility bundles that make the product stronger every time it is tested.
Evidence-native Postgres control planes
Burn-in testing for autonomous database agents
Autonomous rollback and proof-governed DDL
Trust-minimized Conduit operations
Workload DNA and fleet intelligence
No claim without a scope and a caveat.
We use academic-style disclosure, but we keep it operational. Every report should tell a buyer what was measured, what failed, what improved, and what should not be inferred.
- ACM artifact review and badging
- SIGMOD/VLDB reproducibility expectations
- TPC-style full disclosure reporting
- Versioned QueryRook claim registry
- Raw artifacts before polished charts
Short notes build the research corpus.
QueryRook now has a daily publishing lane for Postgres optimization, proof-backed index changes, database-agent safety, and AI boundaries. Each note carries a question, evidence, and caveat.
Read daily notes ->Weaknesses found in papers become product work.
The goal is not to make QueryRook look perfect. The goal is to make every result inspectable enough that flaws become visible while we can still fix them.
Run a validation pilotOne slow-query packet, built for Postgres criticism.
The proof lab shows the public shape of a QueryRook recommendation: fingerprint, EXPLAIN evidence, shadow index proof, lock posture, rollback SQL, and caveats.
Open proof lab ->