递归自我改进路线图：当前仅见片段，全链路尚未实现
Beautiful roadmap paper on Recursive self-improvement.
对递归自我改进现状的系统性梳理，清晰界定能力边界，Agent 研究者必读以校准预期。
Beautiful roadmap paper on Recursive self-improvement.
关于递归自我改进（Recursive self-improvement）的精美路线图论文。
Concludes, we are already seeing pieces of RSI, but full recursive self-improvement is not here yet.
结论是，我们已经开始看到 RSI 的片段，但完整的递归自我改进尚未到来。
Most self-improving AI still cannot improve how it improves
大多数自我改进型 AI 仍无法提升其自身的改进能力
Says that most things called "self-improving AI" today only automate parts of the improvement process.
指出当今被称为“自我改进型 AI”的大多数系统仅自动化了改进过程的某些部分。
Genuine recursive self-improvement would mean the AI can persistently improve not just its outputs, prompts, tools, or code, but eventually the mechanism that decides how future improvements are discovered, tested, and kept.
真正的递归自我改进意味着 AI 能够持续改进其输出、提示词、工具或代码，并最终改进决定如何发现、测试和保留未来改进的机制。
AI is already very strong at answering knowledge and reasoning questions, but still much weaker at doing long, multi-step tasks with tools, software, and changing environments.
AI 在回答知识和推理类问题方面已经非常强大，但在使用工具、软件以及在变化环境中执行长流程多步骤任务方面仍然较弱。
The paper maps progress across 5 levels, from executing human-designed improvements to changing the improver, evaluator, or research policy used in later rounds.
该论文将进展划分为 5 个层级，从执行人类设计的改进，到改变后续轮次中使用的改进者、评估者或研究策略。
That last step makes the process recursive: a successful update changes how future updates are discovered or judged.
最后一步使该过程具有递归性：成功的更新会改变未来更新的发现方式或评判标准。
The survey finds broad evidence for lower levels, while experience-driven learning and deployment adaptation are more domain-dependent and end-to-end L5 evidence remains concentrated in bounded prototypes.
调查显示，较低层级的证据广泛存在，而经验驱动的学习和部署适应更依赖于具体领域，端到端的 L5 证据仍集中在有限的原型系统中。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力