Limits
| Limit | Why it matters |
|---|---|
| Narrow: 3 of 12 on a fresh blind set | A REAL grade means "no known pattern", not "honest code" |
| Rule-based | A fake written to avoid the rules will pass; a determined author can hide fabrication |
| Python only | Other languages get text checks at most |
| Today's improvements are not live yet | They need a gateway deploy |
| Our benchmark, our blind sets | Every case was written by us; no independent party has measured it |
| Semantics it cannot see | A redactor that returns its input, a check of only a length, a typed-in percentage: the code is plausible, the meaning is wrong |
FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.