Sottiaux's Friday double: (1) 'Oops... I did it again. Enjoy reset usage limits for all paid users for Codex and ChatGPT Work. Super grateful for an incredible team who is iterating at light[ning] speed.' (9072 likes). (2) 'GPT-5.6 Sol confirmed to be an extremely good model' with a screenshot (10876 likes — today's highest X engagement). He also noted 'And transitively might also have reset other rate limits out there. Let's see. You're welcome if so.' (1159 likes).

Claude's blockbuster announcement (28134 likes — possibly the highest single-tweet engagement in Follow Builders history): 'Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits. Pro and Team Standard users will continue to have access.' Plus a follow-up: 'We know this has been frustrating, and we want to give you more certainty about what your plan includes.' (3108 likes).

Steipete built a codexbar icon editor with Codex (306 likes): 'ya'all made me go crazy with codexbar icon customization issues, so I built an editor. (by me, I mean codex).' He also shared pain watching Codex use browser + computer use: 'It's both amazing and painful to watch codex use browser + computer use to open Chrome, go to my PR, tap on comment and wrangle with the macOS picker — all TO UPLOAD [a file]' (317 likes). Plus a philosophical question: 'Are we still talking loops or did we shift to graphs yet?' (2502 likes).

Swyx's practical automation tip (1019 likes, 77 retweets): 'btw if you havent set your {codex | claude | gemini | devin} automations to autoresearch how to improve your seo/aeo every week you are really truly missing out.' He also noted 'alpha only gone when we all stop asking these level of questions' (10 likes) and posted 'still true' (4 likes).

Levie on cost economics (331 likes): 'This post is key. The cheaper AI gets, the more opportunity there is for the entire ecosystem — especially including end-customers — to benefit. Everything is [going to be fine].' Madhu Guru on enterprise bottlenecks (264 likes): 'The reason enterprises struggle to go beyond basic chat bots is the talent gap to build harnesses and evals.' She also noted that enterprises won't consume Kimi directly — they'll get it through Google Cloud for enterprise guarantees (85 likes).

Altman shared a simple endorsement (2028 likes): 'this is cool:' — amplifying something noteworthy. Peter Yang noted 'Codex browser use has finally been defeated' (29 likes) and previewed an upcoming episode with Thariq (33 likes), plus a relatable pain point: 'spending an entire day staring at screens trying to manage agents fries my brain. Much better to be walking around outside talking' (26 likes).

Zara's building-in-public tip (206 likes): 'If making content feels like extra work, show the work already happening inside your product. A tiny screen recording, the first version of something rough.' She also captured a cultural shift (141 likes): 'Just a few years ago, many people were uncomfortable with recording meetings. Now it's just assumed that all business meetings are recorded (not for humans, but for AI).'

Thariq on token economics (634 likes): 'building prototypes of mockups, schemas, data models, proof of concepts, etc. is the best way to avoid spending tons of tokens before realizing you don't want [what you asked for].' A practical engineering insight — validate the shape of the output before burning compute.

Rauchg's Friday ship post (70 likes): 'Sandbox data for downloads. Happy Friday! Time to ship more agents.' Plus praise for shipping velocity: 'Ship like @shadcn' (193 likes). Masad shared the Replit community's chess exploration: 'This is a sick exploration of chess history. The Replit community is ChessMaxxing' (94 likes).

Sachin Katti (OpenAI Head of Industrial Compute, ex-Intel CTO, ex-Stanford professor) on MAD with Matt Turck. Recorded at RACE conference in Paris. Key topics: (1) 'Demand far outstrips compute supply today — anything we can bring online, we consume immediately.' (2) Data centers as giant factories turning electrons into tokens, liquid-cooled football-field-scale. (3) OpenAI spending ~$50B on compute this year; industry ~$700B. (4) Jalapeno custom silicon optimizing tokens-per-watt. (5) Inference is now the majority of compute — even training involves heavy inference (synthetic data, test-time compute). (6) 'Our biggest worry is we can't build fast enough — the physical world doesn't move that fast.' (7) Grid investment commitment: 'We are not taking power away from the grid. We invest in new generation.' (8) AI designing its own chips: 'The world of recursion is not far — AI will design the systems it needs to train the next generation of AI.'