Andrew Chen

We just shipped Pulse and I'm so excited about what this means for how we use ChatGPT. When we started, the core idea was simple: what if ChatGPT could talk first? Not waiting for you to ask the perfect question, but proactively sending you things that would actually help based on what it knows about you. My own usage of ChatGPT has changed since we started building this. Last week I asked about bike maintenance and Pulse remembered I'd been looking at weekend activities, so it found me a bike route I didn't know existed near my place. When I was debugging Pulse code, it surfaced fresh RL papers from Arxiv because it knows what's relevant to my work. The magic for me is that ChatGPT now works for me even when I'm not actively using it. Its value no longer stops at the boundary of what I typed. It connects the dots between my past conversations, proactively does work on my behalf, and surfaces what I need, often before I even think to ask. This is v1. We're working to make it more useful, more timely, and get it to more people soon. Would love to hear what feels useful and what doesn't. Your feedback directly shapes where we take this!

Finally, Astral Codex

@bnj @openclaw “Stop jungling and group with team”

Made a little @openclaw tamagotchi that sits on my desktop He occasionally comments on what he sees (mercilessly)

this may be the best seahorse response

I know this is pretty well established at this point, but Codex 5.3 is a much more effective model than Opus 4.6. I went back and forth on both for a bit, but haven’t touched Opus at all now for a full week. First model to get me off of Opus… ever. Good job Codex team.

You can just build things.

Windows has been achieved internally

After playing around with GPT-5.3-Codex after crashing-and-failing with GPT-5.2-Codex, it is with great confusion that I announce that it is a Sonnet 4.5 -> Opus 4.5 leap in coding.

@GaelBreton 2 weeks is an eternity!

GPT-5.2 Pro and Thinking spot the issue in Anthropic’s analysis skill, Opus 4.5 and Gemini 3 Pro do not (Its a not huge issue, but interesting) (The 5.2 models have the strongest statical reasoning in my experience & 5.2 Pro is best if you want pedantic, correct second opinions)

millennial gamers are the best prepared generation for agentic work, they've been training for 25 years

GPT-5.2 pro with astonishing 29,2% in frontier math. ngl thats impressive. Especially because this is one of the toughest benchmarks of all.

One blog post that I am still inspired by today is Context Engineering by Manus. It feels very intuitive and aligns with how things should be. manus.im/blog/Context-E…

Skill file: --- name: ask-questions-if-underspecified description: Clarify requirements before implementing. Do not use automatically, only when invoked explicitly. --- # Ask Questions If Underspecified ## Goal Ask the minimum set of clarifying questions needed to avoid wrong work; do not start implementing until the must-have questions are answered (or the user explicitly approves proceeding with stated assumptions). ## Workflow ### 1) Decide whether the request is underspecified Treat a request as underspecified if after exploring how to perform the work, some or all of the following are not clear: - Define the objective (what should change vs stay the same) - Define "done" (acceptance criteria, examples, edge cases) - Define scope (which files/components/users are in/out) - Define constraints (compatibility, performance, style, deps, time) - Identify environment (language/runtime versions, OS, build/test runner) - Clarify safety/reversibility (data migration, rollout/rollback, risk) If multiple plausible interpretations exist, assume it is underspecified. ### 2) Ask must-have questions first (keep it small) Ask 1-5 questions in the first pass. Prefer questions that eliminate whole branches of work. Make questions easy to answer: - Optimize for scannability (short, numbered questions; avoid paragraphs) - Offer multiple-choice options when possible - Suggest reasonable defaults when appropriate (mark them clearly as the default/recommended choice; bold the recommended choice in the list, or if you present options in a code block, put a bold "Recommended" line immediately above the block and also tag defaults inside the block) - Include a fast-path response (e.g., reply `defaults` to accept all recommended/default choices) - Include a low-friction "not sure" option when helpful (e.g., "Not sure - use default") - Separate "Need to know" from "Nice to know" if that reduces friction - Structure options so the user can respond with compact decisions (e.g., `1b 2a 3c`); restate the chosen options in plain language to confirm ### 3) Pause before acting Until must-have answers arrive: - Do not run commands, edit files, or produce a detailed plan that depends on unknowns - Do perform a clearly labeled, low-risk discovery step only if it does not commit you to a direction (e.g., inspect repo structure, read relevant config files) If the user explicitly asks you to proceed without answers: - State your assumptions as a short numbered list - Ask for confirmation; proceed only after they confirm or correct them ### 4) Confirm interpretation, then proceed Once you have answers, restate the requirements in 1-3 sentences (including key constraints and what success looks like), then start work. ## Question templates - "Before I start, I need: (1) ..., (2) ..., (3) .... If you don't care about (2), I will assume ...." - "Which of these should it be? A) ... B) ... C) ... (pick one)" - "What would you consider 'done'? For example: ..." - "Any constraints I must follow (versions, performance, style, deps)? If none, I will target the existing project defaults." - Use numbered questions with lettered options and a clear reply format ```text 1) Scope? a) Minimal change (default) b) Refactor while touching the area c) Not sure - use default 2) Compatibility target? a) Current project defaults (default) b) Also support older versions: c) Not sure - use default Reply with: defaults (or 1a 2a) ``` ## Anti-patterns - Don't ask questions you can answer with a quick, low-risk discovery read (e.g., configs, existing patterns, docs). - Don't ask open-ended questions if a tight multiple-choice or yes/no would eliminate ambiguity faster.

📢 Confession: I ship code I never read. Here's my 2025 workflow. steipete.me/posts/2025/shi…

This is the biggest jump in Image Arena that we've seen since Nano Banana GPT-Image-1.5 has taken #1 on Image Arena with a significant lead Huge congratulations to the team at @OpenAI for this achievement!

@Yampeleg We decided to multiply the matrices differently and it worked.

ChatGPT’s Pulse is quite good.

@__nmca__ Will consider if the model is still considered frontier by end of year

New ChatGPT Atlas release out! Just hit "update" in the top right. chatgpt.com/atlas - extensions import - icloud passkeys - new downloads ui - setting to use control + tab to cycle to most recently used tab - select multiple tabs at once (shift + click) - ability to set google as default search - "insert" button from Ask ChatGPT sidebar - faster Ask ChatGPT sidebar and by popular demand: vertical tabs! more coming early next week.