🌐 双语
Archive

AI Builders
Digest

2026-07-29 10 builders · 23 tweets · 0 podcasts · 0 blogs

🔥 热点话题

OpenAI 重置 Codex 与 ChatGPT Work 用量限制,详解 GPT-5.6 Sol 效率问题OpenAI Resets Codex & ChatGPT Work Limits, Explains GPT-5.6 Sol Efficiency Issues

OpenAI Codex & ChatGPT 负责人 Thibault Sottiaux 宣布已为所有 ChatGPT Work 和 Codex 用户重置用量限制,并恢复五小时限额。他详细解释 GPT-5.6 Sol 比预期更快消耗额度的原因:模型更愿意长时间工作、进行额外工具调用、协调复杂工作流;在相同推理强度下比前代更努力;程序化工具调用(code mode)带来更多并行响应和缓存 token。团队已优化工具等待与网页搜索处理,预期典型使用可延长约 18%。他承认能力与效率不同步,本应更早识别长尾高消耗用户问题。另有一句:「有一天我们创造了重置按钮,其余皆成历史。」

OpenAI Codex & ChatGPT lead Thibault Sottiaux announced a full reset of usage limits for all ChatGPT Work and Codex users, plus restoration of the five-hour limit. He detailed why GPT-5.6 Sol drained quotas faster than expected: the model is far more willing to work longer, make extra tool calls, and coordinate complex multi-tool workflows; it works harder at the same reasoning effort than prior models; and programmatic tool calling (code mode) produced more responses per turn and higher cached input tokens. Improvements to tool-wait handling and web search should extend typical usage by around 18%. He acknowledged that capability and efficiency do not always advance in lockstep and that the team focused too much on median usage before launch.
OpenAI Codex & ChatGPT lead Thibault Sottiaux announced a full reset of usage limits for all ChatGPT Work and Codex users, plus restoration of the five-hour limit. He detailed why GPT-5.6 Sol drained quotas faster than expected: the model is far more willing to work longer, make extra tool calls, and coordinate complex multi-tool workflows; it works harder at the same reasoning effort than prior models; and programmatic tool calling (code mode) produced more responses per turn and higher cached input tokens. Improvements to tool-wait handling and web search should extend typical usage by around 18%. He acknowledged that capability and efficiency do not always advance in lockstep and that the team focused too much on median usage before launch. He also noted: "One day we created the reset button and the rest is history."
查看原文 →查看原文 →

Swyx:AI 原生 IC 与管理者市场严重分化Swyx: Furious Bifurcation Between AI-Native ICs and Traditional Managers

AI 工程师与播客主理人 Swyx(smol.ai、Latent Space 等)指出当前招聘市场出现前所未见的剧烈分化:对 AI 原生个人贡献者 / player-coach 是巨大牛市,对「某部门负责人」类管理者则是巨大熊市。他总结:管理 10 个 agent 一年经验,已经超过管理 10-100 人十年经验。

AI engineer and Latent Space host Swyx observed a furious bifurcation in the hiring market: a huge bull market for AI-native individual contributors and player-coaches, and a huge bear market for traditional "heads of X" managers. His oversimplification: one year of experience managing 10 agents now outweighs ten years of experience managing 10-100 people.
AI engineer and Latent Space host Swyx observed a furious bifurcation in the hiring market: a huge bull market for AI-native individual contributors and player-coaches, and a huge bear market for traditional "heads of X" managers. His oversimplification: one year of experience managing 10 agents now outweighs ten years of experience managing 10-100 people.
查看原文 →

🛠️ 开发者工具与技巧

OpenAI 开源安全漏洞扫描 CLI 与 TypeScript SDKOpenAI Releases Open-Source Security Vulnerability CLI and TypeScript SDK

OpenAI Codex & ChatGPT 负责人 Thibault Sottiaux 宣布发布 CLI 与 TypeScript SDK,用于发现、验证并修复代码中的安全漏洞。功能包括扫描仓库、审查变更、跟踪发现结果,以及在 CI 中运行安全检查。

OpenAI Codex & ChatGPT lead Thibault Sottiaux announced the release of a CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities in code. It supports scanning repositories, reviewing changes, tracking findings over time, and running security checks in CI.
OpenAI Codex & ChatGPT lead Thibault Sottiaux announced the release of a CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities in code. It supports scanning repositories, reviewing changes, tracking findings over time, and running security checks in CI.
查看原文 →

Peter Yang 用 Claude Design + Claude Code 打造 TastemakerPeter Yang Builds Tastemaker with Claude Design and Claude Code

实用 AI 教程作者 Peter Yang 厌倦了 IMDb 和 Letterboxd 的局限,自己用 Claude Design 与 Claude Code 做出了 Tastemaker——一个统一管理电影、电视剧与游戏评分、列表与推荐的精美个人主页。他录制了完整流程视频:从 design.md 与 HTML 规格,到 Claude Design 原型,再到 Claude Code 迭代。因 API 限制,首发仅限前 100 个免费 profile。

Practical AI tutorial creator Peter Yang built Tastemaker after growing tired of IMDb and Letterboxd limitations. The product lets users rate and review movies, TV shows and games, create lists, and discover what to watch or play next—all in one profile. He documented the full process using Claude Design and Claude Code (design.md + HTML spec → prototyping → iteration) and is launching a limited free release for the first 100 profiles due to API constraints.
Practical AI tutorial creator Peter Yang built Tastemaker after growing tired of IMDb and Letterboxd limitations. The product lets users rate and review movies, TV shows and games, create lists, and discover what to watch or play next—all in one profile. He documented the full process using Claude Design and Claude Code (design.md + HTML spec → prototyping → iteration) and is launching a limited free release for the first 100 profiles due to API constraints.
查看原文 →查看原文 →

Dan Shipper 用 ChatGPT 语音模式完成 Codex 历史长文Dan Shipper Writes Full Codex History Using Only ChatGPT Voice Mode

Every CEO Dan Shipper 花了一上午在沙发上完成 Codex 的完整历史长文。他通过深度访谈 Thibault Sottiaux、AJ Ambrosino 与 Greg Brockman 收集素材,全程用 ChatGPT for Work 的语音模式完成组织、时间线构建、写作、编辑与修订——手从未碰过键盘或鼠标。他称其为创意工作的「goated」工具,文章将于几周后在 Every 发布。

Every CEO Dan Shipper spent a morning writing the definitive history of Codex entirely via ChatGPT for Work voice mode. Drawing on in-depth interviews with Thibault Sottiaux, AJ Ambrosino and Greg Brockman, he organized the timeline, wrote, edited and revised without ever touching keyboard or mouse. He called the experience "absolutely goated for creative work"; the piece will appear on Every in a few weeks.
Every CEO Dan Shipper spent a morning writing the definitive history of Codex entirely via ChatGPT for Work voice mode. Drawing on in-depth interviews with Thibault Sottiaux, AJ Ambrosino and Greg Brockman, he organized the timeline, wrote, edited and revised without ever touching keyboard or mouse. He called the experience "absolutely goated for creative work"; the piece will appear on Every in a few weeks.
查看原文 →

🌍 其他动态

Amjad Masad 关注分布式数学证明与开放信Amjad Masad on Distributed Math Proofs and Open Letters

Replit CEO Amjad Masad 想象一个类似 SETI@Home 的项目:人们捐赠算力去搜索数学证明,而非外星信号。他同时关注是否有人正在起草反对销毁珍本书籍的公开信,并转发了达到 1300 Elo 的相关进展。

Replit CEO Amjad Masad floated the idea of a SETI@Home-style project where people donate compute to search for math proofs instead of alien signals. He also asked who is writing an open letter against destroying rare books and highlighted a milestone reaching 1300 Elo.
Replit CEO Amjad Masad floated the idea of a SETI@Home-style project where people donate compute to search for math proofs instead of alien signals. He also asked who is writing an open letter against destroying rare books and highlighted a milestone reaching 1300 Elo.
查看原文 →查看原文 →查看原文 →

Aaron Levie 推荐 Zuckerberg 的 AI 愿景文章Aaron Levie Highlights Zuckerberg’s AI Vision Piece

Box CEO Aaron Levie 转发并称赞 Mark Zuckerberg 关于 AI 的一篇文章,称其为「伟大的作品与愿景」。

Box CEO Aaron Levie shared and praised a piece by Mark Zuckerberg outlining a vision for AI, calling it a "great piece and vision."
Box CEO Aaron Levie shared and praised a piece by Mark Zuckerberg outlining a vision for AI, calling it a "great piece and vision."
查看原文 →

Aditya Agarwal 回顾十年前的个人洞察Aditya Agarwal on a Decade-Old Personal Insight

South Park Commons 普通合伙人、前 Dropbox CTO Aditya Agarwal 回忆十年前在迷茫中获得的两个深刻洞见:想在世界上建造新事物,以及想与聪明而高强度的人共事。如今这两点都已变得充裕。

South Park Commons GP and former Dropbox CTO Aditya Agarwal reflected on two deep personal insights he gained a decade ago while lost in uncertainty: he wants to build new things in the world, and he wants to work with brilliant, intense people. On days like this, both feel abundant.
South Park Commons GP and former Dropbox CTO Aditya Agarwal reflected on two deep personal insights he gained a decade ago while lost in uncertainty: he wants to build new things in the world, and he wants to work with brilliant, intense people. On days like this, both feel abundant.
查看原文 →

Peter Steinberger:服务大模型很难Peter Steinberger: Serving Large Models Is Hard

OpenClaw 与 OpenAI 相关建设者 Peter Steinberger 转发并评论:「服务大模型很难。」

OpenClaw and OpenAI-affiliated builder Peter Steinberger simply observed: "Serving large models is hard."
OpenClaw and OpenAI-affiliated builder Peter Steinberger simply observed: "Serving large models is hard."
查看原文 →

Madhu Guru 观察到「纯手写代码」的工程师Madhu Guru Spots a Software Engineer Coding by Hand

Meta AI 高级总监 Madhu Guru(前 Google Gemini/Veo 负责人)调侃看到一位软件工程师今天在工作:没有 Claude、没有 Wispr、没有 Tab 键,只是用双手写代码。

Meta AI Senior Director Madhu Guru (formerly led Gemini and Veo at Google) noted seeing a software engineer at work with no Claude, no Wispr, no tab key—just writing code with bare hands.
Meta AI Senior Director Madhu Guru (formerly led Gemini and Veo at Google) noted seeing a software engineer at work with no Claude, no Wispr, no tab key—just writing code with bare hands.
查看原文 →