In plain English
This page explains how AI systems can change over time through updates, tests, retraining, memory, and approvals even when no single model rewrites itself.
- 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.
Selection Pressure in the Report Corpus
Across the reports, selection pressure is the mechanism that turns imperfect measurements into persistent behavior. The candidate does not need hostility. The pipeline only needs a score and a memory.
Sources of selection
Selection can come from benchmark score, evaluatorA system that judges whether an AI output or candidate is acceptable. Open glossary definition rating, user adoption, latency, cost, market popularity, canary metrics, developer preference, or organizational release pressure. Every retained component becomes training material for the next cycle.
No-op erosion
The reports reinforce why no-opThe decision not to change the system. Open glossary definition must remain a first-class outcome. If an organization treats “no change” as failure, the system will keep searching for promotable variation even when the safest action is to stop.
Control pattern
Require promotion to beat both incumbent performance and complexity cost. Preserve a no-op log showing when no change was selected and why. Treat no-op as evidence of mature governance, not lack of progress.