Claude 能自动改进十类对齐失败。真正值得复用的成果，是一套能发现过拟合、能力退化、作弊和外推失效的评测合同。
Automated alignment researchers need more than strong scores. Anthropic's experiment shows the evaluation contract that makes gains testable.
加州 AB 1651 提醒我们：日常生产需要结果核验账，高风险认证还需要独立的来源披露账。
California AB 1651 shows why AI governance needs one ledger for verified work and another for disclosure at certification gates.
审计 OpenAI Jalapeño 首批基准：1.5 至 1.9 倍每瓦吞吐说明了什么，遗漏了什么，以及全栈推理为何才是真正的战略资产。
An audit of OpenAI Jalapeño's first benchmarks: what the 1.5–1.9x performance-per-watt lead proves, what it omits, and why full-stack inference is the
Coding Agent 降低流程制造成本后，UiPath 把价值放在持久执行、可观测性、治理和按量计费的生产运行时。
UiPath Maestro Flow lets coding agents create workflows, then captures value in durable execution, observability, governance, and metered runtime use.
DiffusionGemma 在 batch 1 下达到 7.1 倍 AR 解码速度，质量、prefill 和并发决定这份收益能否进入真实工作流。
DiffusionGemma reaches 7.1x AR decoding speed at batch one, but quality, prefill, and concurrency determine whether that gain survives.
