Google Labs' experiment (23081 likes, top of the day): '🚨NEW EXPERIMENT🚨 Playground is an experimental gaming platform that lets you create your own games with zero coding experience. If you can think it, you can play it.' Altman's double: the custom UI launch (8052): 'ChatGPT can now generate a custom UI for you!' and the personal note (548): 'Have been waiting for this one for a long time, and I would hate to have to go back to the old version of Chat!' Claude's trio: the price cut (10304): 'One more thing: we're halving the price of cache reads on Claude Sonnet 5.5, to $0.10 per million tokens. That makes Sonnet 5.5 around 20% cheaper to run on most long-running work'; the Haiku 5.5 alignment note (1507); and the launch (1462): 'Haiku 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure.' Albert's context (349): 'For reference, Haiku 4.5 came out on October 15th, 2025 so these two columns are less than a year apart. and oh yeah Haiku 5.5 is also much faster and 75% cheaper…'

Sottiaux's triple: the encore (4525 likes): 'Confirmed landed across all accounts. How are we doing so far?'; the silent reship (433): 'Day 3 (encore): We silently re-shipped codex cloud. It's pretty good now'; and 'Will be there by EOD PST' (2203). Thariq's trio: the failure mode (1296): 'the most common failure case I see is when people are working outside their domain of expertise and don't know how to be precise with their prompts and plans, so they have to spend a lot of turns iterating imprecisely'; the game-dev creator thesis (453): 'I think it's probably an incredible time to be a game dev content creator, for many people making games is a form of having fun and they're willing to pay for it!'; and the follow-up (148): 'and because agents make it easier, almost everyone is operating outside their expertise at some point — but luckily, you can just ask the model to teach you what you don't know.' Rauchg's double: 'gm from the best city in the world' (784); and the engineering thesis (605): 'One thing the engineering community will (re)discover is that every program can be hardened and optimized basically ad infinitum. There are real costs in every direction: time, attention, opportunity.' Steipete's essay (671): 'While everyone's talking about agents, I've been exploring how teams can use them to work better together. Here's my take, from OpenAI DevDay 2026.' Levie's compute manifesto (812): 'The compute needed for the stage of AI we're about to enter is going to be insane… This will take a mix of inference volume that is going to keep growing by orders of magnitude.'

Amjad's trio: the math verdict (1095 likes): 'Math was bound to fall first. The purer the field the easier it is for AI to crack'; the consumer clarification (513): 'That's because Replit is not a consumer app. Most of our creators are building businesses or working on one'; and the desktop-security partnership (224): 'Desktop AI apps are great but expose users to supply-chain attacks and catastrophic mistakes by agents. We're building a powerful desktop experience with focus on security & reliability. Excited to partner with @Microsoft on this and be an early adopter of @nvidia's OpenShell.' Peter Yang's trio: the Suno take (40): '.@suno is so damn good honestly. It's probably going to overtake Spotify at some point'; the AI-art timeline (45): 'AI already solved images, music, and video. It's going to solve gaming soon'; and the SaaS warning (6): 'It seems like anything you ship can now be decompiled and rebuilt by AI… Traditional SaaS is especially vulnerable, since it's expensive and full of features built for humans that agents don't need.'

Shipper's double: the Dots field report (94): 'OpenAI's Dots launched last week and they've taken over @every. They protect your attention… Over the weekend, my Dot posted my model-testing results to Slack and brought back replies. I could keep working without opening Slack and disappearing down a rabbit hole. My weekend stayed pretty clear and calm'; and the recursion joke (49): 'tfw your ChatGPT app is controlling your Claude app is controlling your ChatGPT app.' Aditya's double: the SWE essay (179): 'What should an ambitious software engineer do today? Ten years ago, being a great software engineer gave you an unusual advantage: a rare skill you could apply to almost any industry…'; and the Waymo chat (52). Cat Wu's PM use case (246): 'One of my favorite PM use cases for Claude is asking "who used <feature> the most last week? make me an artifact of the top 10 by usage, then reach out and schedule 15 min to chat." It's the fastest way to get user feedback!' Plus Nan Yu's layout joke (21), Cherny's trio (57/55/28), Turck's Ben Affleck bit (158) and the mind-blowing beta (40), Zara's trio (139/108/3), and Nikunj's trio (152/4/2).

A brand-new AI & I by Every (published 2026-10-07, new to this digest): Dan Shipper and Willie Williams — the head of platform who led the engineering — tell the full story of why Every traded personal agents for one company agent. Cold open: week two of OpenAI's Dots — real use cases at Every (school-email triage, Slack access requests, doing launch-video work by phone, model-testing results posted to Slack while the weekend stays calm); the permission structure is 'completely a mystery… extremely annoying'; and the hot take: 'in a year this is the default way you interact with all your AIs… my dot is named Boo, he's super cute, but Boo will probably be dead' — cute characters get thrown away the moment something performs better. The history: from the original OpenClaw phase (Claudius/Clawbot lineage, the January Think Week in Panama, a model drop that made every claw suddenly way better, 'so many bots in our Slack, absolute chaos'), to Plus One (the one-click claw maker, RIP) — Shipper wanted agent-per-person, Williams wanted one-agent-for-everyone, and 'in retrospect I was wrong': OpenClaw was too fast-moving and vibe-coded to build a platform on, and through that process all the personal claws died (RIP R2C2 — the Mac mini got unplugged and nobody restarted it). The takeaways: seeing other people use AI transforms; tinkerers love customization, operators want something that just works; 'instead of everyone investing in their own agents, why not take all that energy and put it into one agent the whole company benefits from.' The bet: the professional/personal split solidifies — one agent per company ('it needs to have been there a year to really get up to speed… I don't see us repeating the chaos of every worker gets their coworker'), plus personal agents that personalize (Muse/Instinct/Dots). Engineering management in the age of AI: 'ironically the most similar job' — leverage from seeing the field (Tend feeds consolidating Slack/email/customer reports, rumor-mill triage, field-marshal view), manager still a force multiplier on people ('my token usage falls off a cliff — there's no billion-token move; we still need to sit down and have a conversation'), dashboards watched by an agent every five minutes for anomalies, an agent as the always-on first on-call. The future org: company agent → team agents (Claudia, Bill) → personal dots as portals to your codex — 'the shadow org chart, but top-down: make something great at the top, split it out when you need specialization.' Willie's personal model: role-specific coaches in GrokBot (a swim coach, a therapist; transactional stuff stays a Codex thread) and why 'Willie's swimming coach is reaching out' beats remembering threads. Plus: the feed-based world (everyone can have their own feed now; frontier alerts in the Every Agent; Jev + Decisions API making proactivity cheap) and X as an explore/exploit problem — 'my cognitive defenses have to be up… it would be great if an agent could take the brunt and I get the nuggets.'