Last updated: 2026/02/25 07:00
Watch the full episode: https://www.youtube.com/watch?v=9jgcT0Fqt7U
Watch the full episode: https://www.youtube.com/watch?v=9jgcT0Fqt7U
Watch the full episode: https://www.youtube.com/watch?v=9jgcT0Fqt7U
Builders Unscripted spotlights the stories behind real projects and the mindset that makes them possible: you can just build things. Prior to joining OpenAI, Peter Steinberger sat down with Romain Huet, Head of Developer Experience, to talk about OpenClaw, his journey in open source, and how he builds with Codex.
"The Codex app lets you go further, do more in parallel, and go deeper on the problems you care about." -gdb
Build faster, cheaper, and with lower latency using prompt caching. This Build Hour breaks down how prompt caching works and how to design your prompts to maximize cache hits. Learn what’s actually being cached, when caching applies, and how small changes in your prompts can have a big impact on cost and performance. Erika Kettleson (Solutions Engineer) covers: • What prompt caching is and why it matters for real-world apps • How cache hits work (prefixes, token thresholds, and continuity) • Best practices like using the Responses API and prompt_cache_key • How to measure cache hit rate, latency, and token savings • Customer Spotlight: Warp (ttps://www.warp.dev/) led by Suraj Gupta (Team Lead) to explain the impact of prompt caching 👉 Prompt Caching Docs: https://platform.openai.com/docs/guides/prompt-caching 👉 Prompt Caching 101 Cookbook: https://developers.openai.com/cookbook/examples/prompt_caching101 👉 Prompt Caching 201 Cookbook: https://developers.openai.com/cookbook/examples/prompt_caching_201 👉 Follow along with the code repo: http://github.com/openai/build-hours 👉 Sign up for upcoming live Build Hours: https://webinar.openai.com/buildhours 00:00 Introduction 02:37 Foundations, Mechanics, API Walkthrough 12:11 Demo: Batch Image Processing 16:55 Demo: Branching Chat 26:02 Demo: Long Running Compaction 32:39 Cache Discount Pricing Overview 36:03 Customer Spotlight: Warp 49:37 Q&A
この記事では、WebMCPという新しい標準について説明されています。WebMCPは、AIがウェブサイトとインタラクションするための構造化されたツールを提供し、従来の遅いボットスタイルのクリック操作を排除します。ScottとWesは、WebMCPの機能やブラウザ統合について議論し、命令型APIと宣言型APIの違いについても触れています。また、WebMCPの利点やトークン効率についても言及されており、これがウェブにおけるAIの重要な瞬間になる可能性があると述べています。 • WebMCPはAIがウェブサイトと効率的にインタラクションするための新しい標準である。 • 従来のボットスタイルのクリック操作を排除し、構造化されたツールを使用する。 • 命令型APIと宣言型APIの違いについて議論されている。 • WebMCPのブラウザ統合が可能である。 • トークン効率が向上し、AIの利用が促進される。
Make that idea happen, no matter how wild it seems, with ChatGPT 🎧: The Letter Live at The Fillmore by Joe Cocker
Sharpen your skills and hit the road with ChatGPT. 🎧: Brain by The Action
Russ Miles joins the show to unpack why developer platforms fail and how to rethink platform engineering through the lens of flow of value rather than factory-style developer productivity metaphors. Russ explains why every organization already has an internal developer platform, and why treating it as platform as a product changes everything. The conversation explores cognitive load and cognitive burden, how to design around strong feedback loops, and why the OODA loop mindset helps teams make better decisions closer to development time. They discuss the risks of overloading pipelines and CI/CD systems, the tension between shipping fast and handling security vulnerabilities in a regulated environment, and how to “shift left” without simply dumping responsibility onto developers. Drawing on lessons from Rod Johnson, the Spring Framework, TDD, and modern software engineering as described by Dave Farley, Russ reframes platforms as systems that support experimentation through the scientific method. The episode also touches on AI assisted coding, developer focus, and how thoughtful developer experience and DX surveys can prevent burnout while improving value delivery.
Javi walks through a logging refactor and shows why Codex's self-verification is a step change: the model runs the app, finds the right session, and proves logs still flow. Takeaways: - Codex can validate its work by running tests and launching the app. - It excels at broad refactors that touch many files. - The model can find session IDs and query tools on its own. - Verification collapses a risky manual loop into minutes. When the agent can prove correctness, you can move faster with less risk. Chapters: 00:00 Why Codex has been a step change 00:18 Self-verification: run tests and launch the app 00:52 The task: a logging refactor across many files 01:10 The risk: do not break observability 01:28 How this used to be verified manually 01:35 Ask the model to verify logs end-to-end 01:50 It finds the session ID and queries logs MCP 02:03 Proof: logs still pipe, task done fast
Deep research in ChatGPT is now powered by GPT-5.2. Now in deep research you can: - Connect to apps in ChatGPT and search specific sites - Track real-time progress and interrupt with follow-ups or new sources - View fullscreen reports Rolling out starting today. https://chatgpt.com/features/deep-research/
Alexander Embiricos (a Product Manager on the Codex team) shows how he uses Codex skills to make a small product change, diagnose a Buildkite failure, and improve the skills so the next PR goes faster. Takeaways: - Skills are a shortcut for repeated workflows like Buildkite logs. - When a skill fails, fix the root cause and update the skill. - The real win is compounding: the codebase gets easier over time. This is the loop: ship the fix, then teach the workflow. Chapters: 00:00 PM context: careful before tagging engineers 00:16 A confusing button and a quick team check 00:32 Delete the button, then hit a PR failure 00:46 Use the Buildkite skill instead of digging through logs 01:09 Install the Buildkite token 01:20 Update the skill so it works next time 01:53 Inductive loop: feedback - fix - improve the skill 02:06 Compounding payoff: Codex gets better over time
How should advertising work in an AI product? Asad Awan, one of the ad leads at OpenAI, walks through how the company is approaching this decision and why it’s testing ads in ChatGPT at all. He explains how ads are built to stay separate from the model response, keep conversations with ChatGPT private from advertisers, and give people control over their experience. Chapters 00:00:29 — Mission and principles 00:04:01 — Separation between ads and answers 00:07:31 — Who will see ads 00:08:52 — Internal input and decision-making process 00:11:06 — Controls and how ads will work 00:15:53 — Guardrails for sensitive conversations 00:17:33 — Skepticism about ads 00:20:26 — Helping small businesses 00:24:13 — Future of ads
We build the tools. You build the future. Start building with Codex. https://openai.com/codex/
Joey demonstrates multitasking with Codex worktrees: delegate a drag-and-drop feature in one worktree while continuing local work, then review and apply both PRs. Takeaways: - Worktrees let you delegate tasks and keep moving in parallel. - Keep local momentum while Codex works in the background. - Use quick questions and comments to correct issues mid-flight. - The mindset shift is from line-by-line edits to architecture and flow control. Parallel workflows turn waiting time into progress. Chapters: 00:00 Work trees enable parallelism 00:20 Example feature: reorder pinned tasks 00:33 Kick off drag-and-drop in a work tree 00:52 Keep working locally while Codex runs 01:27 Spot a bug: the branch is created twice 01:53 Provide Figma context and check other PRs 02:06 Multiple PRs finish in parallel 02:23 Apply the drag-and-drop changes 02:40 Review the result 02:47 Mindset shift: architecture over individual lines 03:04 Context switching and good stopping points
Build something that matters with ChatGPT. Director : Bharat Sikka Music : “Ek Din Bik Jayega" by Mukesh, Poornima, Lata Mangeshkar [Saregama Music]
Work on your form and technique with ChatGPT. Director : Bharat Sikka Music : “Aanewala Pal Janewala Hai" by Kishore Kumar [Saregama Music]
Our smartest model got an upgrade. Claude Opus 4.6 plans more carefully, stays on task longer, and works more autonomously, so you can do more with less back-and-forth. Read more: https://anthropic.com/news/claude-opus-4-6
Across the US, people are using ChatGPT to do more. As the Ortega’s family-run tamale shop in Los Angeles expands, ChatGPT helps them organize schedules, respond to customers, and even spin up a searchable website tracker in a single afternoon—freeing them up to focus on what they’re building together.
No matter what you are building, ChatGPT can help. In Reno, Richard Lane uses ChatGPT to evolve an 86-year-old salvage yard without losing what makes it work. By testing ideas, streamlining decisions, and bringing his team along, ChatGPT makes it possible for one of the fastest metal shops in Reno to keep thriving.
Millions of people daily use ChatGPT to build the things that matter to them. In South Carolina, the Sharps rely on ChatGPT to help run their fourth-generation family farm. From planning to problem-solving, it supports the everyday decisions that help a local operation thrive.
Ed Bayes from the Codex team shows how the Codex app pairs with Figma out of the box: prompt with a Figma link and have a working prototype in minutes. Takeaways: - One-click install for Figma with the Figma skill. - Pasting a Figma link is enough to kick off a strong first pass. - Codex can pull from your design system and get 80-90% there. - Interactive prototypes are key for building dynamic behavior. Design-to-code is faster, and AI UX gets easier to stress test. Chapters: 00:00 One-click MCP install with Figma out of the box 00:14 Paste a Figma link into Codex 00:38 Watch hot reload progress in real time 00:53 Compare against the original design system 01:02 Polish the last 10-20% 01:25 Why this is faster for AI-driven UI 01:36 Stress test with interactive prototypes
Ads are coming to AI. But not to Claude. Keep thinking. Read more about why: https://www.anthropic.com/news/claude-is-a-space-to-think
Ads are coming to AI. But not to Claude. Keep thinking. Read more about why: https://www.anthropic.com/news/claude-is-a-space-to-think
Ads are coming to AI. But not to Claude. Keep thinking. Read more about why: https://www.anthropic.com/news/claude-is-a-space-to-think
Ads are coming to AI. But not to Claude. Keep thinking. Read more about why: https://www.anthropic.com/news/claude-is-a-space-to-think
Andrew from the Codex engineering team shows how he uses automations in the Codex app to take care of the least fun parts of his job. In this walkthrough, you'll see automations that: - Summarize yesterday's commits into a morning pulse - Upskill Codex overnight by fixing skills and scripts - Update personalization and AGENTS.md to reduce misunderstandings - Triage top Sentry issues with memory across runs - Keep PRs green by fixing CI failures and resolving merge conflicts These automations run on a schedule, carry context forward, and help engineers stay focused on the work that needs their attention the most. Chapters: 00:00 Automating the "unfun" work 00:18 Morning commit pulse 00:47 Upskill: improve skills overnight 01:20 AGENTS.md updates 01:48 Sentry triage automation 02:55 Merge conflicts and CI pain 03:22 Keeping PRs green 04:05 Auto-fixing CI and conflicts
The Codex app is a powerful command center for building with agents. • Multitask effortlessly: Work with multiple agents in parallel and keep agent changes isolated with worktrees • Create & use skills: package your tools + conventions into reusable capabilities • Set up automations: delegate repetitive work to Codex with scheduled workflows in the background Download for macOS here - https://openai.com/codex Windows support coming soon.
When you look at scientific tooling, a lot of it hasn’t changed in decades. That’s we recently launched Prism: a free, AI-native environment for scientific writing and collaboration, designed to mean less time in your editor and more time doing research. Physicist & Research Scientist Alex Lupsasca joins Kevin Weil (VP, OpenAI for Science) and Victor Powell (Product, Prism) to walk through what it looks like when ChatGPT works inside a LaTeX project with full paper context. You’ll see Prism: • polish writing with reviewable edits • generate a clean diagram from a whiteboard photo • spin up multiple chat threads to tackle citations and math checks in parallel Explore Prism and try it on your next draft: https://prism.openai.com
Introducing the Codex app—now available on macOS The Codex app is a powerful command center for building with agents. • Multitask effortlessly: Work with multiple agents in parallel and keep agent changes isolated with worktrees • Create & use skills: package your tools + conventions into reusable capabilities • Set up automations: delegate repetitive work to Codex with scheduled workflows in the background Available starting today on macOS with Windows coming soon. And for a limited time, Codex is available through ChatGPT Free and Go subscriptions—and we’re doubling rate limits for Plus, Pro, Business, Enterprise, and Edu users—across the Codex app, CLI, IDE extension, and cloud. Download the app → openai.com/codex
On December 8, the Perseverance rover safely trundled across the surface of Mars. This was the first AI-planned drive on another planet. And it was planned by Claude. Engineers at NASA Jet Propulsion Laboratory used Claude to plot out the route for Perseverance to navigate an approximately four-hundred-meter path on the Martian surface. Read the full story: https://www.anthropic.com/mars
It’s great to be back behind the mic! In this episode of JavaScript Jabber, I’m joined by Dan Shapir and our guest Jack Harrington from Netlify and TanStack
この記事では、Wes BosとScott TolinskiがAIを活用した超特化型の個人ソフトウェアの構築について議論しています。彼らは、個人エージェントやホームオートメーション、JSONをデータベースとして使用する方法、そして大規模言語モデル（LLM）がどのように迅速でカスタムなアプリを実現し、膨大なSaaSを置き換えるかについて探求しています。また、ClawdbotプロジェクトがMoltbotに改名されたことにも言及しています。 • AIエージェントを使用して、個人向けの超特化型アプリを構築する方法を探る。 • JSONをデータベースとして利用する手法を紹介。 • 大規模言語モデル（LLM）が迅速でカスタムなアプリを実現する方法を解説。 • 膨大なSaaSを置き換える可能性について言及。 • プライバシーに関する考慮事項も取り上げられている。
As AI begins to meaningfully accelerate scientific discovery, we’re taking an early step to reduce friction in day-to-day research work with Prism. Prism is a free workspace for scientists to write and collaborate on research, powered by GPT-5.2. Prism offers unlimited projects and collaborators in a single, cloud-based, LaTeX-native workspace, and is designed to expand access to scientific tools. By reducing version conflicts, manual merging, and mechanical overhead, Prism helps teams spend less time managing files and more time engaging with the substance of their work. We’re excited to learn from researchers using Prism today and to continue building toward tools that help science move faster — together. Prism is now available on the web to anyone with a ChatGPT personal account. Coming soon to ChatGPT Business, Team, Enterprise, and Education plans. Try today: openai.com/prism
Sam Altman answers questions and discusses the future of AI with builders from across the AI ecosystem.
Your connected tools are now interactive inside Claude. Manage projects in Asana, draft messages in Slack, build charts in Amplitude, and create diagrams in Figma—without switching tabs.
この記事では、Kent C. DoddsがMCP（Model Context Protocol）とコンテキストエンジニアリングについて解説し、AI駆動のツールを効果的に構築するために必要な要素を探ります。具体的な実例やUIパターン、パフォーマンスのトレードオフについて議論し、ウェブの未来がチャットにあるのかブラウザにあるのかを考察します。MCPの最適化や効率化の手法、MCP UIの重要性、MCPサーバーの開発フローについても触れています。最終的に、MCPの構築におけるHTMLの返却やレンダリングのタイミング、ツールの呼び出し方についても説明されています。 • MCP（Model Context Protocol）とコンテキストエンジニアリングの重要性を解説 • AI駆動のツールを構築するための具体的な実例を紹介 • MCPの最適化や効率化の手法について議論 • MCP UIの重要性とその構築方法を説明 • ウェブの未来がチャットかブラウザかについて考察
