Text rules do not compile into agent behavior. We tested it. The data is public.
0/9 text rules followed. 9/9 compiled rules enforced. A 43-day longitudinal study across 8 concurrent AI agents proving the compiler gap is real.
How corrections compound over time. Session-over-session persistence data showing that compiled rules do not decay, while text-based rules degrade to zero within sessions.
A synthesis paper drawing on 100+ sources to prove the compiler gap is a universal phenomenon across all AI agent systems, not an artifact of a single deployment.
Not a demo. Not a benchmark. A longitudinal study of real human-AI collaboration.
The variable was not the rule content. It was the delivery mechanism.
Only one compiles.
Store what agents know. Retrieval does not equal behavior change.
Improve what agents read. Better prompts still get ignored.
Orchestrate what agents do. Workflows without behavioral enforcement.
Compile corrections into structural rules. What you say becomes what agents do.