In plain English
This page is part of the technical reference. It keeps the expert detail but starts with a plain-language summary for first-time readers.
- Why this matters: AI risk can come from the whole arrangement, not one obvious model.
- What to look for: data, memory, routes, adapters, tools, evaluators, updates, and rollback paths.
- Technical version below: the expert terminology remains available and is linked through the glossary.
ModelBreeder Risk-Side Refocus v1.22.0
This package refocuses the ModelBreeder-derived material toward Cognivirus.com's role: risk analysis, source boundaries, and control surfaces. The constructive possibility material remains preserved, but new links and pages make the risk-side translation explicit.
Added routes
- ModelBreeder Risk Side
- ModelBreeder Risk Translation
- Model Breeding Risk Boundaries
- ModelBreeder Risk Review Worksheet
- ModelBreeder Risk Source Map
Added data
Updated memory intent
The .uai hot memory, long-term memory, report pointer index, source-report corpus, totem, receiver brief, startup packet, and next-recursive prompt now say that ModelBreeder.com can explore possibilities while Cognivirus.com must emphasize risk surfaces, controls, and proof boundaries.
Boundary preserved
This implementation does not claim that ModelBreeder.com is unsafe. It does not claim that every model-breeding, adapterA small add-on that changes or specializes model behavior. Open glossary definition, browser, or dashboard project is dangerous. It maps the same mechanisms into risk-review obligations.