What it claims, checked live
Each line is a claim from the product page. The check behind it runs against the fleet’s own database on every visit (cached five minutes) and hourly as a contract that opens an issue when it fails.
| Check | Claim and what the check saw |
|---|---|
| Pass | It finds its own work: issues with a code task for its code loop were filed in the last 7 days. 556 issues with a code-task line filed in the last 7 days |
| Pass | It scores its decisions: outcomes were checked against a prediction logged before the decision, in the last 7 days. 282 scored in 7 days, all time 787 of 1072 matched |
| Fail | It runs continuously: the tick recorded its steps in the last 2 hours. 0 tick steps in the last 2 hours, last never |
| Pass | It keeps a trust score per kind of action, refreshed in the last 3 hours. 10 surface and action classes, refreshed 2026-11-10T16:08:45Z |
| Pass | It watches its own failure modes: every recorded failure mode family has active detectors. 636 failure modes recorded, 610 active detectors |
| Pass | It loses nothing: every issue that entered the ledger is open, closed with a disposition, or a recorded delete. 566 open, 6104 closed with a disposition, 0 silent drops |
| Pass | It records its own decisions: autonomous decisions were logged with a rollback path in the last 7 days. 739 decisions logged in 7 days |
Who directs it
Rowan Brad Quni-Gudzinas is a systems researcher studying autonomous AI systems. Rowan sets the platform’s operating rules and evaluates what it does; coding agents write the code. Work and record: rwnq8.github.io, orcid.org/0009-0002-4317-5604, github.com/rwnq8, linkedin.com/in/rowan-quni, qnfo.org/about.
Where it falls short
One installation (QNFO). Hundreds of issues are open against a target of zero. The prediction match rate is scored by the system itself. The core still shares a database with QNFO’s research applications. The full list, the evidence and the spin-off plan are on the product page.
Machine-readable: /api/platform-claims.