5/12 Say it twice: OSC never claims code is safe. It records what was checked and what came back. Coinkite's clean AI run on vulnerable code is proof that "clean" ≠ "safe", a data point, not a guara...
6/ 12 "none-found" is a first-class result, not a lesser one. The most useful entry OSC can hold is often "model X ran over scope S, found nothing", exactly the record that would've made Coinkite's pre-hack audit auditable after the fact.
1