# Uploaded source dossier: AI Code Beading and Model Breeding

This locally authored summary preserves the site interpretation of a user-supplied PDF on teleodynamic AI, code beading, browser-executed small models, quantization, and model breeding.

## Direct findings in the dossier

The dossier emphasizes modular small models, browser and WebAssembly constraints, extreme quantization, dynamic skill composition, routers, cascades, adapters, and persistent structured memory for agents. It frames code beading as structured, queryable memory rather than unstructured markdown notes.

## Cognivirus interpretation

The dossier supports a risk lens in which memory, routers, adapters, and edge deployment become part of the cognitive ecology. The important safety implication is not that browser AI is inherently dangerous; it is that cheap local execution and persistent structured memory create more places for behavior to persist after any single model session ends.

## Limits

The dossier mixes technical discussion with broad conceptual synthesis. It should be used as an architectural source note, not as proof of deployed self-replicating AI.
