Kimi K2.7-Code claims 30% fewer thinking tokens and a drop-in API swap path, but independent benchmarks show kernel ...

AI coding agents boost code output by 180% but shipping rises only 30%, MIT finds. Why private data access beats benchmark ...

AI coding benchmark MirrorCode published its full results June 26, showing Claude Opus 4.7 autonomously rebuilt a 60,000-line interpreter and scored 56% overall — completing tasks that take human ...

“Mostly right is the wrong bar,” Pearl CEO Andy Kurtzig says, as research tests top AI models against professional judgment.

Value stream management involves people in the organization to examine workflows and other processes to ensure they are deriving the maximum value from their efforts while eliminating waste — of ...

Xiaomi's HarnessX autonomously rewrites AI agent harnesses mid-execution, delivering +14.5% avg performance gains — and +44% ...

Most AI coding benchmarks still ask the question: did the agent produce code that passes the current tests? This is a useful question, but it is too narrow. Software development is iterative.

Microsoft's new vulnerability-scanning system, codenamed MDASH, scored 88.45% on the CyberGym benchmark, surpassing single-model systems from Anthropic and OpenAI by using more than 100 specialized AI ...

Anthropic reports 65% of its product team's code is AI-generated by Claude, a statistic often misinterpreted as broad ...

Researchers are racing to develop more challenging, interpretable, and fair assessments of AI models that reflect real-world use cases. The stakes are high. Benchmarks are often reduced to leaderboard ...

Two Chinese AI tools now match Anthropic's Mythos in cybersecurity vulnerability detection. Both are freely available, ...