全部 AI 动态

我很好奇一个问题： 大家在不同的 Coding Agent 之间到底怎么共享上下文？ 比如我现在在Claude Code里跑完了一个项目，或者项目跑到一半，然后需要在CodeX里继续。 这两块的上下文怎么衔接比较好？有没有优雅的解决方案？

perhaps one of the most durable parts of pstack are its engineering principles: very concise skills (~10-20 loc) that distill down good practices into its core. as models get smarter, i expect most of pstack skills to compress into short principles or workflows (a generic list of steps) and that i will be able to delete a few of them. how it works is that /poteto-mode contains a short index of available principles and tells agents "In your reply, name each principle that shaped a decision and the specific choice it changed." this is why you'll often see a Principles Named or Used section at the end of the agent's response many of these principles were extracted from my sessions correcting agents since the start of this year. some are inspired by problems i see people talk about, like Attack the Premise, inspired by @VictorTaelin's post https://x.com/VictorTaelin/status/2096273196412498057 you can also use the principles as skills if you want greater control, by invoking /principle-<name> https://github.com/cursor/plugins/blob/main/pstack/docs/guide/08-principles.md

【Tripo × Astra / Opus / Argon × Blender】 ３Dモデル完成しました✨ -------------------- 制作スキルはリプ欄で配布いたします! GithubURL添付いたしますので、良ければご活用ください！ -------------------- 制作期間はおよそ２週間（実質１ヶ月ですが、品質微妙で２回作り直しました💦） 簡単に制作ステップを解説すると以下の通りです。 ✅ 顔・身体・髪の毛・表情毎の３面図・アクセサリー類のイラストを作成 ✅上記のモデルをTripoで全て作成 ・全部SmartmeshP2.0 四角形5000～10000くらいのポリゴン数がおすすめ ・ポリゴン数が多すぎると、補正時に歪みやすくなります ✅各パーツをBlenderMCPで結合 ✅簡単な小物・装飾類は直接モデリングで作成してもらう ✅眼球のモデリング・手を補正 ✅リグ挿入・スキンウェイト調整、IK/FK制作 ✅揺れもの調整 ✅シェーダー調整 ちなみに翼の開閉もできるようにしました🔥 これは翼の構造自体をしっかりAstraに考えさせ、羽の配置を一枚一枚しっかりモデリングさせました！ 気合でこれくらいの品質のものまで作れるのはホント凄いですね💦 #Tripo

https://x.com/i/article/2108375208570818560

I built a dream logger. For all those dreams I forget five minutes after waking up. This little @M5Stack CoreS3-SE turns dreams into images using the @OpenAIDevs image-gen API. 🧵

Connecting Codex to hardware helped me build my cat his own TV I have an extremely active cat who unfortunately loves TV a little too much (he tries to jump into it) I've always wanted to give him something more stimulating, so I decided to build him his own

this is Felix, the detective in a mini mystery game i built for nights when i host dinner. and yes: he lives on an M5StickS3 from @M5Stack. i can ask him Qs about the case, pitch my lil theory, and get clues printed on an actual receipt 🧵 I built it with Codex, ImageGen, Decisions API & GPT-Transcribe:

Vibe coded hardware is here, and amazingly fun! This week I'm covering how to use @OpenAIDevs Codex, @CloudflareDev Workers and the Zectrix Note4C to build your own custom eink display. I made a train tracking screen 🤓🚂 https://jamesdevonport.com/p/can-you-vibe-code-hardware

hey @bot convert the linked essay to an ebook and email it to my kindle so i can read it on my flight

did you know you can submit Grok Bot (and Grok, Cursor) plugins? submit them here! https://x.ai/bot/marketplace/submit-a-plugin

pstack is the new solid

it's easy to fall into the trap of micromanaging your agents instead of correcting the environment that shapes its behavior. pstack 0.15.9 now ships a new /correct skill: if you keep correcting agents for the same mistakes, it finds the pattern and fixes it with architecture, types, and checks. i like to imagine that commits rapidly flowing into the codebase are like a bonsai tree rapidly growing in whatever direction it pleases. what you want to do is not to give in to the chaos, but tame it intentionally with the right constraints 0.15.9 also has improvements to: • /architect: now includes instructions on designing agent friendly architecture • a new /benchmark-checklist skill based on @brendangregg's benchmarking checklist here's a prompt you can give to grok bot to start architecting a better codebase with pstack: /poteto-mode create a new Project agent to refactor and rearchitect my repo so that our architecture is more agent friendly. it should use both the /correct and /architect skill to look through past commits and find the most common pitfalls that agents fall into, whether its with code quality, performance, or bugs. if relevant you can also use the /recall skill to find past conversations for context. it should make a high level plan for me to review. it should also answer its own open questions by prototyping rather than defer to me. come back to me with a plan backed by real data. once i approve the plan, use either the stack or full autopilot playbook (ask me which i want) to execute

ban and delete most if not all code comments! agents love copying what already exists. a seemingly harmless comment explaining some temporary workaround is one of the worst viruses you can have in your codebase. like a weed it'll spread quickly as a tool for agents to quickly get their job done. before you know it, your codebase is overrun with hacks and stale comments. just one more workaround bro you can use pstack's /no-comments skill as a weedkiller. it'll seek out useless comments while keeping the ones that actually should exist (like explaining a browser quirk), and figure out how to solve them on a first principles basis here's a prompt you can try asking @bot: create a project agent that uses /poteto-mode to spawn a /swarm to use /no-comments and identify useless code comments we can prune in my repo. come back with a high level plan of which of these workarounds would be best solved by refactoring and which we can straight up delete https://x.ai/bot/plugin/9717366

https://x.com/i/article/2109541953155313664

https://x.com/i/article/2110601994578866176

Agent Evals 最常见的问题都有什么？ @LangChain 团队 @divyasan0 以 Q&A 形式分享了 Agent Evals 实践指南，把常见问题一次性回答了。她给出的可落地工程流程：从真实 trace 起步、小数据集建立基线、失败案例回流迭代、离线/在线分层评估、代码优先于 LLM 裁判、benchmark 用容器化环境验证最终状态。 QA 博客地址： https://www.langchain.com/blog/common-questions-about-agent-evals 为什么 Agent 难以测试 ？ · 输出不确定性：同一输入，多次运行结果不同； · 答案非唯一：多个不同回答都可能是正确的，无法做字符串精确匹配； · 路径自选：Agent 自己决定执行步骤，即使最终答案正确，路径也可能是“错的”（调用了不必要的工具、未经授权就行动）。 起点：从 trace 开始的最小闭环 一条清晰路径值得直接采纳： 读一批真实 trace → 选定一个行为 → 建 10–20 条小数据集 → 跑基线实验 → 之后所有改动都与基线对比。 其中最有价值的工程习惯是失败回流机制：发现一次失败 → 作为新样本加入数据集 → 改进 agent → 重跑全量。这让数据集随系统演化而增长，评估从“一次性验证”变成“持续改进的飞轮”。未上线的团队可以用内部 dogfooding 产生 trace。 离线 vs 在线：两道互补的闸门 · 离线评估（Offline）：发生在发布前、对数据集、用来测新 prompt/模型/工具防回归 · 在线评估（Online）：发生在生产环境、对真实流量 trace、用来持续监控线上质量 两者形成闭环：线上标出的失败，转化为线下的新样本。数据集组织原则也很实用，同一行为、同一组评估器的样本放一起；输入、评估器或目的不同就另建数据集，避免一个大盘子什么都装。 评估器的优先级：代码 > LLM 裁判 > 人工 · 代码评估器：确定性检查（如“价格是否低于预算”），快、便宜、零歧义，能用则必用； · LLM-as-a-judge：处理代码判不了的主观维度（如“回复是否清晰确认了航班信息”）； · 人工审核：最贵但最可靠的信号，用于抽样校准前两者； · Pairwise（成对比较）：让裁判二选一，比绝对打分更稳定。 评估对象的颗粒度也分层：单步 → 最终回复 → 完整对话 → 环境终态（如容器内记录、文件、测试是否通过）。 LLM 裁判能否信任：三条可操作建议 1. 用人工标注对照校准，调整裁判 prompt 直到两者对齐； 2. 问窄问题，“回复是否提到了每一个中转站？”远好于“这是个好回复吗？”； 3. 用二元判定或分类标签，避免 1–10 打分（数值打分的跨 run 一致性很差）。 Eval ≠ Benchmark：一个重要的概念澄清 Benchmark 本质是 “重型离线 eval”，其特殊性在于： · 任务长链路，涉及数十次工具调用； · 每个任务自带独立的容器化隔离环境； · 评分方式是检查运行后的环境状态（订票记录是否存在、代码测试是否通过），而非比对文本。 文中介绍了开源评估 harness Harbor（Terminal-Bench 的底层引擎），每个任务构建一个容器，结果记录到 LangSmith 实验。自建 benchmark 的路径：从 trace 收集任务 → 定义评分规则 → 对分数做 hill climbing（爬山式迭代优化），甚至可以做成 Karpathy 式的自动化研究循环。 线上评估的成本控制 生产流量全量打分不现实，博客给出三层降本： 1. 采样与过滤：只评估有负反馈的 trace； 2. 代码检查优先：几乎零成本； 3. 换更便宜的裁判模型：小模型裁判在 pass/fail 判定上与人类标注一致，总成本 $0.34 vs Claude Sonnet 4.6 的 $28.17（约 1/83）。对二元判定类任务，裁判模型确实不必用旗舰模型。 不必从零自建 生态已提供相当多现成件：针对幻觉、PII 泄露等常见问题的预置 LLM 评估器模板、开源 agentevals 包（针对消息和工具调用）、可从代码仓库和 trace 直接生成可运行评估的 Eval Engineering Skill、从 issue 聚类自动生成数据集和断言的 LangSmith Engine。且整个体系框架无关；支持 OpenAI Agents SDK、Claude Agent SDK，或用 @ traceable 装饰器裸接。 迁移也不必推倒重来：数据集以 CSV/JSONL 导入，pytest/Vitest 集成保留现有测试体系，已有裁判 prompt 直接粘贴复用。 两个进阶难题 多轮对话：每一轮都看似正常、整体却失败的“累积性失败”是单点评估的盲区。解法是给 trace 附上 thread ID 将各轮聚合，再挂载面向整段对话的在线评估器（如“用户是否被迫重复了自己的需求”）；离线侧则把完整对话存为样本，或用 LLM 模拟用户。 超越最终输出：复杂 agent 需要检查轨迹（消息/工具调用序列）与终态（仓库变更、测试结果）。例如编码 agent 跑完后直接在其工作环境里执行该 repo 的测试套件；这正是 Harbor + 容器化沙箱的设计动机。