# ModelBreeder.com v2.4.0 Positive-Side Implementation Notes

This release records a product-direction correction: ModelBreeder.com is the constructive, benefit-centered side of the model-breeding idea. It should teach people how to build useful model ecologies that compound capability, improve local sovereignty, reduce compute waste, preserve human generativity, and make institutions stronger.

The public site should not become a warning catalog. Repeated cautions, pessimistic language, and negative-case narratives belong on Cognivirus.com. ModelBreeder.com may keep precise engineering boundaries, but it should describe them as fitness proofs, evaluation checkpoints, scope notes, compatibility requirements, and release evidence.

Accepted implementation moves from the latest reports:

- preserve the plain PHP, root-extractable, file-backed site target;
- keep `/docs` as durable source memory and `.uai/long-term-memory.uai` as the pointer ledger;
- add a positive-side directive to active `.uai` memory;
- rename high-visibility safety language toward fitness proof and adaptive governance;
- add Genome and FitnessVector schema guidance;
- add a fitness/novelty selection guide;
- add a zero-dependency Rust browser LLM improvements page;
- add a Cognivirus boundary page so risk-focused negative cases are routed to the companion domain.

This release does not remove evaluator independence, rollback, provenance, scope boundaries, or human review. It reframes those mechanisms as constructive infrastructure for measured improvement rather than as the site's main story.