Theorem of Context Fragility
Theorem T5
AI-executed development is bounded not by model capability but by the quality of the context made available to the model. Stale or contradictory context is the primary failure mode.
Implication
Context management — the assembly, curation, and invalidation of the information an AI system uses to make decisions — becomes a first-class engineering concern. "Stale plans are toxic context."
Full proof and discussion
A complete derivation — including proof sketch, worked examples, edge cases, and references to supporting research — is in progress. This page will be expanded as the theory matures.
In the meantime, the statement and implication above are the compressed form; they are sufficient to derive the main methodological consequences worked out in RACE Programming.