In plain English
This page covers the high-risk pattern where small adapters, routes, memory, evaluators, and descendants can reinforce each other across time. It is a risk model, not a build guide.
- 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.
Algorithmic reproduction taxonomy for apex ecologies
Several reports use biological terms such as mitosis, meiosis, apoptosis, and pathogenesis. Cognivirus.com keeps these terms explicitly metaphorical. They describe information-flow and governance patterns, not living organisms.
Taxonomy
| Metaphor | Public-safe meaning | Relevant safety question |
|---|---|---|
| algorithmic mitosisA proposed metaphor for near-copy successor creation involving artifacts, runtime packages, memory states, or deployment patterns. It is not a biological claim and not a replication instruction. Open glossary definition | near-copying of an artifact, runtime package, memory state, or deployment pattern | Who is allowed to create a successor, where may it run, and how is it counted? |
| algorithmic meiosisA proposed metaphor for recombination among compatible models, adapters, task vectors, prompts, or routes. The safety concern is behavior that appears only after recomposition. Open glossary definition | recombination of compatible weights, adapters, task vectors, prompts, code paths, or routes | What new behavior appears only after recombination? |
| algorithmic mutation | bounded change to a candidate through fine-tuning, pruning, quantization, prompting, or policy edits | Which prior evidence is invalidated? |
| algorithmic deprecation | retiring, pruning, compressing, or merging components to reduce cost or confusion | Did deprecation remove safety-relevant constraints or auditability? |
| cognitive pathogenesis | a pattern carried by memory, social proof, procedure, or narrative rather than one model file | Does the behavior persist through human and organizational channels? |
What this does not mean
It does not mean AI models are alive. It does not mean current systems are conscious. It does not equate model adaptation with literal computer-virus replication. It does not imply that all modular systems are unsafe.
Why the taxonomy helps
The metaphors are useful when they expose missing governance questions. For example, a lineage graph may show that an adapter has a parent, but it may not show whether a behavior was reintroduced through synthetic data, a memory snapshotA saved state of what the AI system remembers. Open glossary definition, a router statistic, or a human procedure.
Required records
A reproduction-aware ecology should record:
- parent artifact identifiers;
- exact compatibility assumptions;
- transformation type;
- evaluator versionThe exact version of the evaluator used for a test or release. Open glossary definition;
- runtime composition;
- memory snapshot;
- synthetic-data lineageThe parent-child history of models, adapters, datasets, or releases. Open glossary definition;
- permission changes;
- UTC timestamp;
- no-opThe decision not to change the system. Open glossary definition and rejection outcomes.
Boundary
The taxonomy is an assuranceConfidence, backed by evidence, that a system meets safety or governance requirements. Open glossary definition tool. It is not a design recipe for autonomous replication.