Kimi K3's 79% repair score refers to successful attempts in a specific software-testing setup. AlphaSignal's benchmark account reports 53 successes in 67 attempts across a suite of 13 tasks. Each attempt began in a fresh sandboxAn isolated environment where AI-generated code or agent actions run without being able to touch anything real.Full definition → without network access. Success required passing visible tests, hidden tests and the regression suite, while leaving test files unchanged. This definition checks whether a patch survives multiple forms of verification. The percentage counts repeated attempts on that suite, not the fraction of arbitrary software bugs the model can fix.
Terms in this piece · Glossary
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
postPrompt cache TTL is the hidden line item in long coding-agent sessions
articleFull breakdown: Claude-only apps often wait until the host speaks the new revisi
articleDo not ask whether the agent follows the rule. Ask what stops it when it does no
articleFull Breakdown: When can a change ship unread? When something cheap and hard to 
