从叠加假说、稀疏自编码器到归因图，机械可解释性在 2026 年第一次跑进生产线：Anthropic 能画出 Claude 的思维回路，OpenAI 从 GPT-4 抽出 1600 万特征，Goodfire 本周把内部探针做成低价监测产品。本文梳理核心进展、落地场景与「SAE 是否找到真实特征」的根本性质疑。

Sparse AutoEncoders (SAEs) offer a promising way to inspect language model representations, but it is still unclear what…

While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autono…

Physical AI is moving rapidly from research to large-scale deployment. By 2035, ABI Research projects an installed base …

Welcome back to TechCrunch Mobility, your hub for the future of transportation and now, more than ever, the role AI is p…

本站「RSI 与 Agent 自进化」专题入口页：把站内 11 篇原创深度与 14 条一手动态收进同一张地图——先给 30 秒定性（RSI 改"改进能力"、自进化改"任务表现"），再按概念／全景／证据／判定／事件／工程／治理七层分层索引，附三条按时间预算划分的阅读路线（30 分钟 / 2 小时 / 半天）、一页速查卡、收录标准与更新日志。