Sottiaux's closing triple: 'I think I can officially say: we are so back' (14421 likes, top of the day); 'All reset for everyone. Enjoy the week with Astra' (8803) — another banked reset, this time just to enjoy the week; and 'Looking at this 28 page deck right now where Codex is keeping track of all the individual launches… Brought to you by an incredible team (and Astra)' (1882). Steipete's PR theater of the absurd (608 likes): 'Did a PR to one of our upstream projects and they requested some minor changes. What's even the point with this workflow? You already wrote the prompt, why make me ping my agent again so your agent then merges?' — human-to-human GitHub etiquette colliding with agent-to-agent reality.

Shipper's trio: gmail archaeology — 'i was looking through my gmail messages from the early 2000s. i have incredible news for you 13 year old dan….you came up with a name :)' (415); the childhood-dream post — 'hard to express how much building something like this was my childhood dream… and how out of reach it was until last week' (65); and the Mastercard riff (18): 'there are some things more tokens can't make. for everything else there's Fable and Astra.'

Amjad's functional-equivalence thesis (730 likes): 'I don't think we've reached AGI but what we have is functionally indistinguishable from AGI. Because we have a relentless programmer that doesn't get bored or tired. So any problem that can be cast as a coding problem is virtually solved.' Levie's double: the AI jobs prediction is playing out opposite — automation lands in areas with non-finite demand (427 likes), and the building advice (101): 'with rapidly accelerating AI progress, you should likely be building with a vision that contemplates at least a few orders of magnitude of capability improvement or token volume… The best opportunities right now are those where you can [add] value to customers today with things that are barely only possible, but with a full mission that feels nearly impossible.'

Nikunj's bottleneck shift (98 likes): 'As building gets easier, what to build becomes the important bottleneck. You might not need many of them, but the good ones [PMs] are extremely well setup for the AI native organizations.' Peter Yang's field note (17 likes): 'Yeah I don't know if I'm doing something wrong but Astra seems worse at triggering my skills automatically and following my instructions in them.' Plus his Gemini-for-music note (5) and AI-stitched family videos (10), and Nan Yu's optical illusion (44).

Rauchg's trio: 'Your software factory needs high quality video recording. 𝚊𝚐𝚎𝚗𝚝-𝚋𝚛𝚘𝚠𝚜𝚎𝚛 delivers' (1030 likes) — 'We're shipping more functionality derived from the lessons of automating the new bottleneck in software engineering: review, testing and QA'; oss-grants v2 (678 likes): $1,000 each to 35 contributors across themes — ① agent skills & tools ('Skills are now a valuable form of software') like 𝚙𝚐𝚋𝚘𝚝, 𝚔𝚗𝚒𝚙, 𝚞𝚗𝚕𝚒𝚐𝚑𝚝𝚑𝚘𝚞𝚜𝚎, ② local AI like 𝚜𝚕𝚘𝚝𝚜𝚝𝚛𝚎𝚊𝚖, 𝙻𝚘𝚌𝚊𝚕𝙰𝙸, 𝚌𝚘𝚕𝚒𝚋𝚛𝚒; and 'Lots of cool stuff shipping this week' (201). Askell's thought experiment (1017 likes): 'It would be cool to set up an email address that autonomous AI models could reach out to if they were looking for moral guidance. But it would require a reverse captcha that can detect that you're neither a human nor an AI being instructed to break it by a human.'

Garry Tan's trio: 'Children need to be teach-yourself-on-YouTube-maxxing' (465); RFS demystified (172): 'RFS are just hunches on what might be next, and conversation starters… The main determiner of a startup idea's success are the specific founders building specific tech for specific customers. The rest is categorization'; and the SF politics jab (126). Amjad's London opening (120): the Mayor of London Sadiq Khan helped open Replit's first international office — 'He calls himself an AI realist: clear-eyed about the disruption AI will create, but optimistic about the impact if we give people the tools and skills to participate.'

A re-run in today's feed (first covered in our 2026-08-28 issue): AI & I's 'Best of the Pod' with the CEO/CIO/managing partner of Walleye, a ~$10B AUM hedge fund (and Every consulting client). The episode opens with the legendary all-hands memo: 'Using ChatGPT is not cheating. That's a non applicable idea from academia… As a hedge fund, we should be ashamed to leave money on the table by ignoring tools that make us faster, smarter, and more effective. Not using these tools is like refusing to use the Internet in 1995.' Core threads: the owner-operator's conviction ('if we ultimately get disrupted by AI, that's on me'); the March 2023 TMT analyst who demoed tools to replace his own job — which became Current, the internal fundamental-investing agent platform (usage spikes during earnings); the Sword in the Stone / Merlin analogy — living backwards in time, seeing the endpoint more easily than the steps; weekly demos with a third of the 400-person firm joining, flipping fear from 'afraid to make mistakes' to 'afraid to fall behind'; adoption stats (~75% weekly active ChatGPT users, ~1/3 using AI coding tools like Windsurf); LLM-driven sentiment analysis on unstructured data as pure alpha; recording every Zoom/call into a data lake with LLMs processing the morning risk call for memory and prediction; writing as thinking — 4-5 hour memos becoming 15 minutes ('I don't think we need to waste time trying to be clever anymore'); the tools don't negate thinking — they give you more time to think; the robber-barons/Gilded Age analogy — skills going obsolete within a single lifetime, faster this time; 'I'm not a VC, thank God'; cowboys and barbed wire; human nature as the stationary constant ('history doesn't repeat itself, it rhymes'); quant vs human intuition in low-large-numbers fuzzy situations; first principles as incentives alignment, hating fluff, intellectual honesty, measuring everything; and the firm as microcosm — 'you have to learn these tools or you're just not gonna be competitive. At some point, if you don't, that's your choice.'