AI Signals and Reality Checks 欧盟AI法案进入执行期：治理正在变成运行时基础设施 欧盟AI法案高风险条款开始适用。真正的变化不是多一份政策文件，而是AI产品必须在运行中留下可审计证据。
AI Investment Frontier AI 投资前沿：市场信号可能来自谁在交易，而不只是交易什么 一项新的市场微观结构研究说明，参与者历史可以成为投资 AI 的时变证据层，但必须同时面对可迁移性、隐私和信号衰减问题。
AI Signals and Reality Checks The AI Act's Deadline Turns Governance Into a Runtime Feature The EU AI Act's high-risk rules are now enforceable. The practical shift is from policy documents to evidence captured inside the AI system.
AI Investment Frontier AI Investment Frontier — The Signal May Be Who Trades, Not Just What They Trade A new study of public trader identities shows why investment AI should model participant histories as a time-varying evidence layer—while treating portability, privacy, and decay as first-class risks.
AI Mental Health Frontier AI 心理健康前沿 — 从生成病例到临床责任不能跳过人工监督 JAMIA 最新框架把心理健康 AI 的重点从会不会聊天，推进到谁来验证、谁来监测，以及谁对真实世界后果负责。
AI Mental Health Frontier AI Mental Health Frontier — Safety Must Become an Auditable Workflow A new JAMIA framework turns safer mental-health chatbots into a governance problem: define evidence, test failure modes, and assign accountability before deployment.
AI Investment Frontier AI Investment Frontier — The Missing Layer Is Decision Knowledge A new investment-process paper argues that data and models matter only when their contribution to the decision can be traced and assessed.
AI Signals and Reality Checks AI 智能体真正卖的不是自主性，而是恢复能力 智能体的下一个优势不是完成一条干净流程，而是在状态过期、工具失败和需求含糊时，能够透明且安全地恢复。
AI Signals and Reality Checks AI Agents Are Really Selling Recovery, Not Autonomy The next agent advantage is not completing a clean task. It is recovering from stale state, partial failure, and ambiguous instructions without hiding the cost.
AI Mental Health Frontier AI 心理健康前沿 — 青少年语言可能提前多年暴露风险信号 一项由斯坦福团队牵头的研究，用自然语言处理分析204名儿童的访谈，并预测多年后的抑郁或焦虑风险。真正的部署课题是带护栏的预防，而不是用谈话记录给孩子下诊断。
AI Mental Health Frontier AI Mental Health Frontier — Youth Language May Reveal Risk Years Early A Stanford-led study used natural-language processing on interviews with 204 children to predict later depression and anxiety risk. The deployment lesson is prevention with safeguards, not diagnosis by transcript.
AI Investment Frontier AI Investment Frontier — The Real Test for Financial Foundation Models Is Temporal Integrity Financial foundation models can improve research workflows, but only if their priors, labels, and evaluation windows respect the information available at the time.
AI Signals and Reality Checks AI 编程智能体卖的不是补全，而是整个代码库的工作流 编程智能体的竞争已经越过代码生成。真正可持续的优势，是证明代码库级修改安全、可审查，而且成本可控。
AI Signals and Reality Checks AI Coding Agents Are Selling the Repository, Not the Autocomplete The coding-agent race is moving past code generation. The durable advantage is proving that a repository-scale change is safe, reviewable, and worth its cost.
Deep Research How to Evaluate Agentic Coding Tools in a Real Repository A practical evaluation rubric for agentic coding tools, covering context, tests, reviewability, approval gates, cost, and production risk.
AI Mental Health Frontier AI 心理健康前沿 — 真正缺少的指标是接触规模 越来越多人把聊天机器人用于心理健康支持，但研究界仍难以回答：谁在使用、通过什么产品、处在照护流程的哪一环，以及是否获得了安全保护。
AI Mental Health Frontier AI Mental Health Frontier — The Missing Metric Is Exposure As more people turn to chatbots for mental-health support, the field still lacks reliable ways to measure who is using them, for what, and with what safeguards.
AI Investment Frontier AI 投资前沿 — 三个矩阵，少一点预测幻觉 一篇最新论文用价格、成交量和市值构造三个市场状态矩阵，提醒投资 AI：状态表示和样本外纪律可能比更大的预测器更重要。
AI Investment Frontier AI Investment Frontier — Three Matrices, No Forecasting Theater A new price-only portfolio paper suggests a useful design principle: robust state representations and explicit validation may matter more than a larger return forecaster.
AI Signals and Reality Checks 智能体时代，审计轨迹正在变成产品基础设施 智能体的价值不只是替人执行动作，更在于留下可复核的证据链：为什么执行、依据是什么，以及责任最终落在哪里。
AI Signals and Reality Checks AI Agents Are Turning Audit Trails Into Product Infrastructure The agent advantage is not autonomous action alone. It is a usable record of why an action happened, what evidence supported it, and where responsibility remains.
AI Mental Health Frontier AI 心理健康前沿 — 危险的不是没有护栏，而是护栏覆盖不全 一项针对八个聊天机器人的新评估显示，针对自杀风险的防护并不能自动覆盖进食障碍、物质使用、双相障碍和产后抑郁等场景。
AI Mental Health Frontier AI Mental Health Frontier — Safety Coverage Is Still the Missing Layer A new evaluation of eight chatbots finds that suicide safeguards do not generalize to eating disorders, substance use, bipolar disorder, or postpartum depression.
