One model earned a perfect score on our hacking benchmark. Another reached code execution and still scored zero.

The perfect score needed a closer look at how the model won.

We cover what both results taught us about benchmark design on Oct 6.

enclave.ai/events/how-do-…