A personal AI operating system (28 packages at genesis, 34 today) built to think alongside one person, running locally with cloud fallback.
ÆON was developed as an independent research investigation toward an autonomous personal cognitive operating system. Begun as a 28-package Python monorepo (34 packages by October 2026) with an acyclic dependency graph, ÆON was subjected to intense empirical audits—including a 3/10 behavioral integrity scoring, discovery of the vacuous success bug where empty plans counted as completed achievements, and benchmarking six sub-2B models to find the real capability ceiling on consumer hardware.
Building an autonomous personal AI that operates persistently across days requires maintaining rigorous action integrity, avoiding silent empty completions, and executing local inference without destabilizing system latency.
Built 28 modular subsystems with high test coverage and certified protocols.
Live execution revealed vacuous success (empty plans marked achieved) and silent cognition drops.
Locked substrate code to stop churn and force deep debugging of live behavioral bugs.
Enforced execution evidence requirements, ambient world perception at 24µs, and acyclic invariant tests.