3:07 a.m. · Checkout is down · $18,400 lost every minute
An AI can fix it in seconds.
RecoveryPact makes sure the fix doesn’t cost more than the outage.
Two AIs negotiate every recovery: one speaks for the database, one speaks for the business. A separate safety authority decides what may actually run, and recovery only counts when customers can buy again.
Below are real recorded runs from our benchmark, replayed side by side. Same incident, same live database, same AI model. Pick one.
Why this is new
Two AIs that disagree on purpose
The Database AI wants the database healthy fast. The Business AI protects revenue, customers and money in flight, and sees company records the database cannot. They argue until the evidence settles it.
Agreement is not permission
Even when both AIs agree, a separate rule-based safety authority checks the exact action against the live system. It approves once, for two minutes, or refuses. It cannot be talked around.
The business defines “fixed”
Typical agents declare success when the database looks healthy. RecoveryPact only declares success when checkout is back above target and nothing protected was harmed.
It learns, carefully
After each incident an Auditor AI writes the report and a lesson. Memory can stop a mistake from being repeated. It can never authorize an action on its own.
How one recovery works
- ObserveDatabase signals, plus company records through Zetaris
- ProposeDatabase AI picks from actions that were actually observed
- ChallengeBusiness AI demands evidence, weighs costs, accepts or escalates
- AuthorizeRule-based authority approves one exact action, or refuses
- VerifyCheckout back above target, nothing protected harmed
- LearnAuditor AI writes the report and stores the lesson in memory
Built with
- NVIDIA NemotronAll three AIs and the memory embeddings
- ZetarisOne query across the workload registry, change calendar and live orders
- Meterless H-MEMRanked, self-correcting memory with an append-only ledger
- CursorDevelopment environment