The Protocol
A perception protocol that makes any system self-evolving, metaprogrammable, and reflective. Born from 30 years of contemplative practice. Validated by engineering metrics. Now running inside production systems.
Unity and fragmentation. Finds concepts artificially split across too many documents. Perceives the whole before the parts. Where others see isolated pieces, Monade sees a fragmented unity that wants to reconverge.
Ecosystem health. Assesses whether document clusters genuinely work together. Not just connected. Symbiotic. A link isn't a relationship. Symbiote knows the difference.
Structural patterns. Detects missing foundation documents and inverted hierarchies. The blueprint under the content. If the structure is wrong, everything built on top inherits the flaw.
Tone and accessibility. Catches tone mismatches and inaccessible language. The last lens to activate, because tone matters most when everything else is already coherent.
Each cognitive cycle runs a multi-step loop, up to 5 iterations. The loop stops early when health metrics converge.
Delta detection: what changed since the last scan?
Run orphans, obsolescence, contradictions, and gaps lenses.
1-2 perception lenses selected adaptively by the MetaAgent.
Compute health metrics: orphan ratio, linking density, contradiction count.
Generate or update perception protocols via metaprogramming.
Apply write-back actions: link, archive, rewrite, merge.
Build AI summaries of document clusters.
Generate knowledge protocol if evaluation delta exceeds 15%.
When metrics are stable, stop. Otherwise, repeat from step 1.
Protocols are perception rules, not actions. They are injected into lens prompts to modify how the lens interprets documents in subsequent iterations. This is metaprogramming: the system reprograms its own perception based on what it learns.
"Documents containing 'changelog' or 'release notes' in the title should be considered stale after 30 days instead of the default 180 days, as they are time-sensitive by nature."
Most systems run forever or stop arbitrarily. OmniQ converges. It recognizes when additional passes yield diminishing returns.
Four factors govern convergence: health metrics stability across consecutive iterations, observation count plateau where new passes surface no new findings, a maximum of 5 iterations as a hard ceiling, and cancellation available at any point for human override.
The result is a system that invests exactly the right amount of cognitive effort. No more, no less. When the loop stops, the system is coherent. Not perfect. Coherent.
OmniQ found its first production home in knowledge bases, but the lenses are universal. Fragmentation, ecosystem health, structural integrity, tone: these aren't RAG problems. They're perception problems. Any system that processes information can benefit from structured self-perception.
Whether you're building a knowledge base, a design system, or an agent network, the protocol adapts.
Become an ally