Architecture Philosophy: Rule-First Design

The Core Question When building an engine to verify AI-written code, architects hit a brutal dilemma. Where do rules live, and who enforces them? Axiom's answer is simple. Rules must precede code. Rule-First Principle Traditional software development works like this: Write code first. Run tests to catch bugs. Add rules to prevent similar bugs. Repeat forever. This reactive approach shatters catastrophically when verifying AI-generated code. LLMs excel at crafting outputs that pass tests while concealing logical flaws—they're geniuses at creating "plausible wrong answers." Their probabilistic nature means bugs reappear in subtly different forms, bypassing test-based detection. Without rules locked in first, the engine eventually gaslights itself. It starts lowering verification standards to accommodate AI hallucinations. A verification engine that discovers rules from code inspection ends up in a self-justifying loop. It validates code against constraints derived from that same code. So