Start with the outcome
AXIOM begins with the desired end state, constraints, non-goals and the evidence required to call the work done.
PUBLIC DOCUMENTATION
How AXIOM works.
A compact view of the operating model behind scoped execution, authority, verification and continuity.
Decompose into bounded work
The core loop is Recover → Define Done → Decompose → Schedule → Execute → Verify → Replan → Loop. An Atom is the largest useful unit that remains safely bounded, independently verifiable and recoverable.
Current truth is explicit
Historical material can be searched for learning and recovery, but it never silently becomes current authority. Protected owner gates remain outside autonomous execution.
Execution is not completion
Completion requires evidence appropriate to the claim: readback, tests, hashes, runtime evidence, browser verification or human QA where machines are insufficient.
Failures stay attributable
AXIOM distinguishes pre-execution failures, partial mutation, verification failures, dependency changes and route or capacity failures. Retry requires evidence that retry is safe.
Recovery is part of the architecture
Known-good rollback, checkpoints and provider-independent control paths are preserved so one failed dependency degrades its branch instead of stopping the whole system.
Reuse → Verify → Extend
Working components, source and proven routes are recovered before new systems are created. Rebuild is the last option, not the first reflex.