EvidenceStrong architectural inferencev1.22.1
In plain English
This page shows what kind of support exists for each claim: real systems, experiments, early evidence, architectural reasoning, open questions, or speculative scenarios.
- 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 Source Map
Evidence levelStrong architectural inferenceTechnical label: Strong architectural inference
This page maps uploaded source reports to the risk-side pages created in v1.22.0. It does not independently verify every external link listed inside those reports. It explains how the reports were used by Cognivirus.com.
Source report groups
| Report group | Public route or summary | Risk-side use |
|---|---|---|
| Controlled evolution theory | Source report page | Genome, FitnessVector, novelty, speciation, and multi-parent merge risk. |
| ModelBreeder architecture resources | Architecture projects and Architecture resources | Browser-native architecture, external project leads, client-side execution boundary. |
| Adaptive ecologies part 2 | Architecture resources part 2 | Champions, specialists, challengers, independent evaluation, lineageThe parent-child history of models, adapters, datasets, or releases. Open glossary definition DAGs, resource ledgers. |
| Browser ecology dashboards | Ecology dashboard source report | Dashboard risk, artificial-life analogues, local visualization, browser state. |
| Zero-dependency runtime | Zero-dependency Rust LLM source report | Wasm runtime, adapterA small add-on that changes or specializes model behavior. Open glossary definition payloads, tokenizer/parser risk, caches, allocation, edge execution. |
| Apex Threat review | Apex real instances and Evidence levels | Bound the risk side with external evidence labels and what-is-not-proven panels. |
Claims supported by this source map
| Claim | Evidence level | Boundary |
|---|---|---|
| Candidate populations create review-scale pressure. | Strong architectural inferenceA conclusion or output produced from data. Open glossary definition | Supported by architecture reports, not a measured incident rate. |
| Fitness and novelty become selection pressure. | Strong architectural inference | Supported by controlled-evolution design logic, not proof of malicious intent. |
| Multi-parent merging increases lineage complexity. | Strong architectural inference | ProvenanceA record of where a component or behavior came from. Open glossary definition risk follows from graph complexity; exact risk depends on implementation. |
| Adapters and LoRAA common kind of small adapter used to specialize large models. Open glossary definition can be supply-chain surfaces. | Security-framework consensus | Bounded by Apex source map; not every adapter is dangerous. |
| Local-first browser execution changes inventory and rollbackReturning a system to an earlier known state. Open glossary definition. | Strong architectural inference | Local privacy benefit remains valid; risk is governance visibility. |
| Runtime caches and custom parsers require controls. | Strong architectural inference | Zero-dependency reduces third-party code but does not remove local defects. |