AI Agent 的 Harness 机制学习思考 – 人人都是产品经理
woshipm.com2026 年 2 月，OpenAI 官方博客发布了一篇震撼业界的文章：《Harness Engineering: Leveraging Codex in an Agent-First World》。 文章讲述了一个看似不可思议的实验：一个仅有 3 人的工程师小
EvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification
arxiv.orgAnthropic proposes the concept of skills for LLM agents to tackle multi-step professional tasks that simple tool invocations cannot address. A tool is a single, self-contained function, whereas a...采纳率从7.9%到54%：快手智能Code Review的三阶进化
mp.weixin.qq.com万字长文干货，建议收藏！op7418/Claude-to-IM-skill: Bridge Claude Code / Codex to IM platforms — chat with AI coding agents from Telegram, Discord, or Feishu/Lark.
github.comBridge Claude Code / Codex to IM platforms — chat with AI coding agents from Telegram, Discord, or Feishu/Lark. - op7418/Claude-to-IM-skillSkillNet: Create, Evaluate, and Connect AI Skills
arxiv.orgCurrent AI agents can flexibly invoke tools and execute complex tasks, yet their long-term advancement is hindered by the lack of systematic accumulation and transfer of skills. Without a unified...
Real Money, Fake Models: Deceptive Model Claims in Shadow APIs
arxiv.orgDistributed Systems Reading List
dancres.github.io
How OpenAI uses Codex | OpenAI
openai.comA detailed look at how OpenAI teams use Codex to understand code, refactor systems, improve performance, boost velocity, and enhance engineering workflows.Agentic Memory: Learning Unified Long-Term and Short-Term Memory...
arxiv.orgLarge language model (LLM) agents face fundamental limitations in long-horizon reasoning due to finite context windows, making effective memory management critical. Existing methods typically...Memory in the Age of AI Agents
arxiv.orgMemory has emerged, and will continue to remain, a core capability of foundation model-based agents. As research on agent memory rapidly expands and attracts unprecedented attention, the field has...
After two years of vibecoding, I'm back to writing by hand
atmoio.substack.comAgents write units of changes that look good in isolation. They are consistent with themselves and your prompt. But respect for the whole, there is not.EvoSkill: Automated Skill Discovery for Multi-Agent Systems
arxiv.orgCoding agents are increasingly used as general-purpose problem solvers, but their flexibility does not by itself confer the domain expertise needed for specialized tasks. Recent work addresses...生成率从8%到60%：快手智能测试用例生成系统的四阶进化
mp.weixin.qq.com快手构建智能用例生成系统，实现从手动编写到审核校验的模式转变
Making etcd incidents easier to debug in production Kubernetes
cncf.ioWhen Kubernetes clusters experience serious issues, the symptoms are often vague but the impact is immediate. Control plane requests slow down. API calls begin to time out. In the worst cases…World Models
arxiv.orgWe explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial...MicroGPT explained interactively
growingswe.comWalk through Karpathy's 200-line GPT from scratch. Tokenize names into integers, watch softmax convert scores to probabilities, step through backpropagation on a computation graph, explore attention heatmaps, and see a tiny model learn to generate plausible names.Unit Testing Guidelines
geosoft.no
Harness engineering: leveraging Codex in an agent-first world | OpenAI
openai.comBy Ryan Lopopolo, Member of the Technical Staff
Compound Engineering: Make Every Unit of Work Compound Into the Next
every.toA system for AI-assisted software development where each piece of work makes subsequent work easierAI Agent Systems: Architectures, Applications, and Evaluation
arxiv.orgAI agents -- systems that combine foundation models with reasoning, planning, memory, and tool use -- are rapidly becoming a practical interface between natural-language intent and real-world...The Path Ahead for Agentic AI: Challenges and Opportunities
arxiv.orgThe evolution of Large Language Models (LLMs) from passive text generators to autonomous, goal-driven systems represents a fundamental shift in artificial intelligence. This chapter examines the...That's all for now. Come back later for more.