Self Evolve 是 AI Agent 自进化证据地图：先判断系统改了什么、谁验证、能否保留和回滚，再进入论文、项目、benchmark 与证据边界。English mirrors now cover the core evidence path; long-tail parity is still in progress.https://agent-evolution.com/https://agent-evolution.com/blog/anthropic-dynamic-workflows-agent-swarm-evolve/https://agent-evolution.com/blog/anthropic-dynamic-workflows-agent-swarm-evolve/把 Claude Code dynamic workflows 放回 Self Evolve 主题：它不是普通多 Agent，而是让任务组织、验证队列、并行子代理和安全边界一起变成可审计的 Agent-Swarm Evolve。Sat, 30 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/anthropic-may-2026-agentops-platform-shift/https://agent-evolution.com/blog/anthropic-may-2026-agentops-platform-shift/把 Anthropic 2026 年 5 月的 Opus 4.8、Dynamic Workflows、Stainless 收购、Claude containment 和 Series H 融资放回 AI Agent 自进化主题：模型竞争正在变成 AgentOps 基础设施竞争。Sat, 30 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/agent-frameworks-evolution-layer/https://agent-evolution.com/blog/agent-frameworks-evolution-layer/从英文论文第六章拆解主流 Agent 框架：它们提供运行时、角色、对话、流程、prompt 编译或状态图，但自进化需要额外的评估、记忆、更新和治理层。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/evaluation-benchmarks-for-self-evolving-agents/https://agent-evolution.com/blog/evaluation-benchmarks-for-self-evolving-agents/从英文论文第五章拆解自进化 AI 的评估问题：代码、数学、Agent、开放式 benchmark、Star 传播信号、过程指标和推荐评估协议。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/evolutionary-code-and-algorithm-discovery/https://agent-evolution.com/blog/evolutionary-code-and-algorithm-discovery/从英文论文第四章拆解进化式代码和算法发现：LLM 作为优化器、语义变异器、程序搜索器，以及 AlphaEvolve、Darwin Gödel Machine、OpenEvolve 的工程意义。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/five-evolution-loops-for-ai-agents/https://agent-evolution.com/blog/five-evolution-loops-for-ai-agents/从英文论文第二章拆解 Self Evolve 的 Five Evolution Loops：Specification-to-Execution、Search、Evaluator、Reflection、Population，以及如何组合成真实 Agent 系统。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/future-roadmap-for-self-evolving-ai-agents/https://agent-evolution.com/blog/future-roadmap-for-self-evolving-ai-agents/从英文论文第八章拆解未来方向：评估瓶颈、长期记忆漂移、安全与对齐、五大循环组合、生产挑战和 Self Evolve 两到三年路线图。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/self-evolving-ai-introduction-static-to-evolving-agents/https://agent-evolution.com/blog/self-evolving-ai-introduction-static-to-evolving-agents/把 Self-Evolving AI Agents 英文论文第一章拆成一篇可读博客：解释什么是 AI 自进化、它和在线学习/AutoML/普通 Agent 的边界，以及为什么真正的问题是可验证的自我修改。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/self-improvement-methods-for-llm-agents/https://agent-evolution.com/blog/self-improvement-methods-for-llm-agents/从英文论文第三章拆解 LLM/Agent 自我改进方法：推理时修正、反思记忆、训练期自博弈、RL、语言梯度，以及什么时候该用 prompt、fine-tuning 或 reinforcement learning。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/user-painpoints-production-self-evolving-agents/https://agent-evolution.com/blog/user-painpoints-production-self-evolving-agents/从英文论文第七章拆解自进化 Agent 的用户痛点：幻觉、循环、工具误用、上下文溢出、调试困难、成本、部署、监控与治理。Tue, 26 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/adas-to-dgm-evolution/https://agent-evolution.com/blog/adas-to-dgm-evolution/追踪从 ADAS（ICLR 2025）到 Darwin Gödel Machine 的技术演进：从图灵完备搜索到开放式进化归档。UBC Jeff Clune 团队如何让 Agent 自己设计自己。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/agent-evolution-layer/https://agent-evolution.com/blog/agent-evolution-layer/Reflexion 如何把任务失败转化为自然语言记忆，让 Agent 在不更新权重的情况下持续改进。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/ai-self-evolution-from-concept-to-implementation/https://agent-evolution.com/blog/ai-self-evolution-from-concept-to-implementation/什么是 AI Self Evolution？本文从 Self Evolve 视角梳理自我进化智能体的核心概念、技术路线与实现路径，帮助读者先建立可复核判断框架。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/ai-self-evolution-map/https://agent-evolution.com/blog/ai-self-evolution-map/用工程视角梳理自进化智能体如何改进提示词、工具、记忆、代码、工作流与策略，并给出可变对象、反馈信号、验证门和回滚边界。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/alphaevolve-deep-dive/https://agent-evolution.com/blog/alphaevolve-deep-dive/分析 Google DeepMind AlphaEvolve 的双模型架构、MAP-Elites 质量多样性搜索和 4×4 complex-valued matrix multiplication 的 48-scalar result，并标出复核边界。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/code-self-improvement-playbook/https://agent-evolution.com/blog/code-self-improvement-playbook/一套轻量实践流程：让编码智能体自调试、写测试、修复失败并重新验证，同时保留基线、回归切片、失败记录和可审计改进证据。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/cross-domain-research-map/https://agent-evolution.com/blog/cross-domain-research-map/梳理 AutoML/NAS、进化计算、LLM 自我改进与 Agent 框架之间的深层连接，揭示 Self Evolve 的技术全景。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/evolutionary-computation-llm-survey/https://agent-evolution.com/blog/evolutionary-computation-llm-survey/导读进化计算与 LLM 融合的代表性进展：从 OPRO、FunSearch 到 AlphaEvolve，从 LLaMEA 到 OpenEvolve。本文是阅读入口，不是系统综述协议。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/how-to-build-self-evolving-agent/https://agent-evolution.com/blog/how-to-build-self-evolving-agent/面向开发者的 Self-Evolving Agent 构建笔记。从最小闭环开始，逐步添加评估器、记忆、进化策略和回归防护，并标出哪些环节仍需复核。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/llm-as-evolutionary-optimizer/https://agent-evolution.com/blog/llm-as-evolutionary-optimizer/解读 OPRO、FunSearch、ReEvo 和 LLaMEA 四篇核心论文，理解 LLM 如何充当变异器、重组器和候选生成器。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/llm-five-self-improvement-paradigms/https://agent-evolution.com/blog/llm-five-self-improvement-paradigms/对比分析 Self-Refine、Reflexion、Agent Symbolic Learning、RISE 和 Absolute Zero 五种 LLM 自我改进范式的方法、优劣与适用场景。Self Evolve 核心技术解读。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/open-source-self-evolving-projects/https://agent-evolution.com/blog/open-source-self-evolving-projects/从 Self Evolve 项目索引中选取 10 个覆盖不同机制的开源自进化 AI 项目，涵盖进化式代码优化、Agent 进化框架、反思记忆与自评判训练。本文说明它们适合回答什么问题，也标出证据边界。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/researcher-network/https://agent-evolution.com/blog/researcher-network/梳理 AI 自我进化领域的核心研究者、实验室与合作网络，从 Jeff Clune 的开放式进化到 Google DeepMind 的 AlphaEvolve。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/self-evolve-project-index/https://agent-evolution.com/blog/self-evolve-project-index/从 OpenEvolve、AgentEvolver、Reflexion 到 Self-Refine，理解自进化 AI 项目的四类实现路线。Fri, 22 May 2026 00:00:00 GMThttps://agent-evolution.com/blog/selfevolve-paper-deep-dive/https://agent-evolution.com/blog/selfevolve-paper-deep-dive/详细解读 SelfEvolve 论文的方法、公式与实验结果。探索 LLM 自生成知识与迭代自调试如何实现代码自我改进，以及与 Self Evolve 生态的关联。Fri, 22 May 2026 00:00:00 GMT