Real-SWE ran frontier models against licensed, private enterprise codebases (billing, tax, multi-service work) and one number jumped out at me: rollout duration barely moves resolution. 71. 4% of rollouts that finished in under 10 minutes FAILED.
Source: [Dev.to](https://dev.to/cole_halton_42f71d71b809b/giving-a-coding-agent-more-time-barely-helps-5bg3)