Compile-Time AIpattern
Compile-Time AI is the architectural pattern of moving expensive semantic reasoning out of the request path and into a build step — so a token (this era's CPU cycle) is spent once, deliberately, to produce an artifact that is cheap to consume forever after.
Runtime AI (RAG, agentic systems) re-derives meaning from raw material on every query — nothing compounds, every query pays full price. Compile-Time AI applies the compiler discipline (amortize analysis once) to knowledge, context, and agent skills.
Where it appears
- Compile-Time AI blog post (2026-07-12)
- Taxonomy of kib
- Kompile
- Context Compilation Theory
- llm-wiki-compiler
- OVIR
- SkCC
Descend
Dependencies
Foundations
- —
Historical evolution
Named and surveyed across six independent teams (2026-07-12) converging with no cross-citation — evidence of a real convergent trend, not marketing.
Implementation details
Enacted by the Knowledge Compiler SDK as a 9-phase pipeline; SKCE is its reflexive application (compiling its own explanation).
Sources
- [2026-07-12-compile-time-ai-knowledge-compiler-architecture] 2026-07-12-compile-time-ai-knowledge-compiler-architecture
- [2026-07-12-compile-time-ai-knowledge-compiler-architecture] 2026-07-12-compile-time-ai-knowledge-compiler-architecture