Boris Cherny's reflection got today's highest engagement (7226 likes): 'Something I have been thinking about: in the past, the best engineers I knew spent a lot of time automating their work in various ways. Better vim/emacs automation...' — the thread argues that AI coding agents are the natural extension of something the best engineers have always done. The engineers who adopt AI fastest are the ones who were already building automation tooling.

Thariq's prompting philosophy (2754 likes, 248 retweets): 'ideal prompting technique is: thin prompts, thick artifacts + context, thin skills.' This distills the shift from verbose prompting to providing rich context and letting the model do the heavy lifting. His follow-up: 'software engineering is the profession of automation' (1619 likes).

Steipete on PR review (572 likes): 'If I put up a PR for an iOS codebase I don't know and a code reviewer rejects it because it doesn't use the right framework, or if a designer builds a new feature...' — making the case that the same standards should apply to AI-generated code. He also found 'This is really clever' (1534 likes) and noted '5.6 is relentless' (201 likes).

Sottiaux's three-tweet day: (1) 'We've had no 5h limit in Codex plus and pro for a few days. Do you think it is better or are you finding it difficult to manage the usage included in the weekly?' (3287 likes) — the 5-hour limit removal is a major UX shift. (2) File deletion investigation (2225 likes): 'On file deletions. We've investigated a handful of reports where GPT-5.6 unexpectedly deleted files. What we have found is that this most commonly occurs when...' — transparency on a real safety issue. (3) 'Now that we merged ChatGPT and Codex, what should we merge next? What's the double or nothing move.' (1594 likes).

Altman on the silent mode debate (2542 likes): 'amazing to me that some people want the silent version' — expressing surprise that users prefer the non-verbose mode. Peter Yang identified the biggest missed opportunity (283 likes): 'ChatGPT Live and Codex are two incredible products that don't talk to each other. This is OpenAI's biggest missed opportunity imo.' He went on a walk with ChatGPT Live and wished Codex context carried over.

Rauchg shared Vercel Sandbox metrics (218 likes): 'Growing DAUs at 100% m/o/m, 3.5M+ sandboxes created per day, best-in-class Active CPU pricing model, powering Notion, Airtable, Meta, Z...' — Vercel is becoming infrastructure backbone for agent sandboxing. His British pounds joke 'Easy money (British pounds)' (535 likes) continues his currency meme series. He also highlighted Web Analytics API use cases (199 likes).

Levie hosted an enterprise IT dinner (488 likes): 'Change management remains the #1 challenge.agent adoption in the enterprise is happening but slower than the hype suggests.' He also praised open-weights innovation (223 likes): 'The future of AI is going to be a mix of frontier intelligence and open weights.'

Zara on agent-friendly company design (134 likes): 'If you want agents to actually work inside a company, you have to design the company so they can read it. Shopify did this with an agent that had no private channels...' — the insight that organizational transparency is a prerequisite for agent productivity. She also shared how coding agents became her creative outlet (239 likes): 'GitHub is basically my Sketch/Figma.'

Josh Woodward announced Gemini Spark expansion (494 likes): 'Gemini Spark is headed to more Ultra subscribers around the world plus 4 new features: 1) Spark can open and edit Google Docs, acting on feedback from last week.' He also released Google's first Gemini Southeast Asia Report (472 likes): 'Active users more than doubled in the last year, 70% of prompts are submitted in local languages.'

Matt Turck's Argentina formula (111 likes): '1) trail 10-15 minutes before the end, 2) bring on Lautaro Martinez, 3) win, 4) repeat.' Nikunj's Paris immigration complaint (93 likes): 'After spending 3 hours in the immigration line in Paris, I will never complain about SFO and US immigration ever again.' Garry Tan called to 'bring back the em dash' (61 likes). Swyx noted 'you can just tweet things into existence' (140 likes).

Chris Pedregal (Granola CEO, $1.5B valuation, 60-70 people) on AI & I with Dan Shipper. Key topics: (1) Startups are knife fights whether going well or not. (2) Meeting notes are not the end-all value — the real prize is inventing AI-native work interfaces. (3) Dan's pirate/architect framework for product teams. (4) Pre-generation strategy: Granola pre-generates millions of meeting briefs, only ~10% opened, but critical when needed. (5) 'Codex-native apps' paradigm — bring your agent to the app. (6) Token spend tracking started only last week, triggered by Fable. (7) The missing UI paradigm for multi-context agent work. (8) ~50% of weekly Granola users use agentic features.