We've been talking about this for a while, and it's finally running. Every Tuesday on the AI Breakthrough call, we're setting aside 30 to 45 minutes for the Hot Seat. Depending on how the time goes, that's one, two, or three members in the chair. You bring a real problem, the room goes to work on it. Here's why I'm excited about it, and it's not really about the person in the seat. It's about everyone else. When one member puts a genuine, unsolved problem in front of this group, all of us get to watch smart operators reason through something real instead of something hypothetical. That's the part you can't get from a course. Half the value of a hot seat goes to the people who never said a word. There's also a second participant. Agent | Solution Architect, one of our AI agents, reads every submission and writes its own full answer before the call. Then it seals it. Nobody sees it, including you. On the call you speak first, the room digs in, and only then do we open the envelope and compare. And when the room beats the agent, which happens, that gap gets captured and turned into an actual upgrade before the next Tuesday. So the agent isn't testing us. We're training it. Every hot seat makes the next one sharper, and your thinking ends up in every proposal it writes after yours. I've attached a PDF that walks through the whole thing end to end: the loop diagram, what the agent reads and writes, how the improvement cycle works, and what makes a strong submission. Worth two minutes before you write yours. 👉 Submit here: https://hotseat.agenticsociety.com/submit Takes about two minutes. Closes Monday 6:00 PM ET for the next day's call. And if you don't get picked right away, nothing is thrown away. You keep your spot and move up the list every week you wait. Getting in early genuinely works in your favor. Bring the thing you're actually stuck on. Not the polished version. That's what this room is for. Who's first? 🔥

𝗜 𝘄𝗿𝗼𝘁𝗲 𝘁𝗵𝗲 𝘀𝗰𝗿𝗶𝗽𝘁 𝘁𝗵𝗮𝘁 𝘄𝗮𝘀 𝘀𝘂𝗽𝗽𝗼𝘀𝗲𝗱 𝘁𝗼 𝗽𝗿𝗲𝘃𝗲𝗻𝘁 𝘁𝗵𝗶𝘀. 𝗜𝘁 𝗱𝗲𝗹𝗲𝘁𝗲𝗱 𝟮𝟵 𝗿𝗲𝗮𝗹 𝗳𝗶𝗹𝗲𝘀 𝗮𝗻𝗱 𝗰𝗼𝘀𝘁 𝗺𝗲 𝘁𝗵𝗲 𝗿𝗲𝗽𝗼. The setup is the one most of us land on, because every step of it is sensible. Your Company Brain lives in Google Drive so it syncs everywhere. The desktop app makes it a normal folder. Your code ends up in there too, so the brain and the work sit together. Then you work from two machines and point AI agents at all of it. Every choice is fine. Together they put 𝗮 𝗱𝗮𝘁𝗮𝗯𝗮𝘀𝗲 𝗶𝗻𝘀𝗶𝗱𝗲 𝗮 𝗳𝗶𝗹𝗲 𝘀𝘆𝗻𝗰𝗲𝗿. Here is the collision. Your git folder is a database. Google Drive does not merge conflicts, it renames them, adding a " 2" or a " (1)". It does that everywhere, including inside the git folder. Git never puts a space in its own internal filenames, so the moment Drive renames one, git cannot read its own history. It never announces itself as a sync problem. It shows up as git having amnesia: "bad sha1 file", "badRefName", a missing object, or a push that just gets refused. Nothing hints that sync did it. 𝗡𝗼𝘄 𝘁𝗵𝗲 𝗽𝗮𝗿𝘁 𝘁𝗵𝗮𝘁 𝗰𝗼𝘀𝘁 𝗺𝗲 𝘁𝗵𝗲 𝗿𝗲𝗽𝗼𝘀𝗶𝘁𝗼𝗿𝘆. The obvious fix is a script that deletes every odd-looking file inside the git folder. I wrote that script. I was proud of it. It ran on every single commit. What I did not know is that Drive sometimes renames the ORIGINAL rather than adding a copy beside it. So those "junk" files are sometimes your only copy of a real git object. My clever cleanup deleted 29 of them. The repo could no longer rebuild its own history. Re-clone. The work survived only because it had been pushed. 𝗧𝗵𝗲 𝗰𝗼𝗿𝗿𝗲𝗰𝘁 𝗿𝘂𝗹𝗲 𝗶𝘀 𝗼𝗻𝗲 𝗹𝗶𝗻𝗲, 𝗮𝗻𝗱 𝗶𝘁 𝗶𝘀 𝘁𝗵𝗲 𝗼𝗽𝗽𝗼𝘀𝗶𝘁𝗲 𝗼𝗳 𝘄𝗵𝗮𝘁 𝘂𝘀𝘂𝗮𝗹𝗹𝘆 𝗴𝗲𝘁𝘀 𝘄𝗿𝗶𝘁𝘁𝗲𝗻: • Suffixed file exists, original is MISSING, so 𝗿𝗲𝘀𝘁𝗼𝗿𝗲 𝗶𝘁. Drive renamed your original. • Suffixed file exists, original is there, so delete the duplicate. The original is fine. Nothing inside the git folder gets deleted unless a correctly named file survives next to it. And the habit that matters more than any script: 𝗧𝗵𝗲 𝗿𝗲𝗺𝗼𝘁𝗲 𝗶𝘀 𝘁𝗵𝗲 𝘀𝗼𝘂𝗿𝗰𝗲 𝗼𝗳 𝘁𝗿𝘂𝘁𝗵. 𝗧𝗵𝗲 𝘀𝘆𝗻𝗰𝗲𝗱 𝗳𝗼𝗹𝗱𝗲𝗿 𝗶𝘀 𝗮 𝗿𝗲𝗽𝗹𝗮𝗰𝗲𝗮𝗯𝗹𝗲 𝘃𝗶𝗲𝘄 𝗼𝗳 𝗶𝘁. 𝗡𝗼𝘁𝗵𝗶𝗻𝗴 𝗶𝗿𝗿𝗲𝗽𝗹𝗮𝗰𝗲𝗮𝗯𝗹𝗲 𝗲𝘃𝗲𝗿 𝗹𝗶𝘃𝗲𝘀 𝗼𝗻𝗹𝘆 𝗶𝗻 𝘁𝗵𝗮𝘁 𝗳𝗼𝗹𝗱𝗲𝗿.

Quick housekeeping that will get you faster answers. Going forward, please send any question or support request to [email protected]. Here's why: when you email one of us individually, your message lands in one person's inbox. If they're traveling, in a session, or heads-down building, you wait - sometimes days. ask@ goes to the whole team. Whoever's closest to the answer picks it up, so you get a real response as fast as we can get it to you. Slack isn't going anywhere. Our Slack community is alive and staying alive. Keep using it for what it's great at - thinking out loud, sharing wins, swapping builds, and helping each other in real time. Just know that a support question posted in a channel (or DM'd to one of us) can slip past us. If you need an answer, email it. One inbox. One team watching it. Faster answers. 📩 [email protected] - save it as a contact so it's there when you need it.

Hey folks! If you've been wanting to put your first real agent to work but weren't sure where to start, this one's for you. I put together a complete, do-it-yourself guide for spinning up the Project Manager agent — the same one we run internally to keep the whole team on task. It's built in layers so you can go as deep as you want: ✦ Level 1: Use it — what the agent actually does for you day to day (the 10am standup, the tracker, the 5pm nudge). Plain English, zero setup. If you read nothing else, read this. ✦ Level 2: Build it — the step-by-step to stand up your own. Five steps, the exact walls to watch for, and a "you'll know it worked when" check at every stage. ✦ Starter Kit — copy-paste the agent's whole "brain" file and a setup checklist, right off the page. ✦ Appendix — for when you're ready for more: adding the rest of your fleet, keeping costs down, watching that everything's actually running, and the gotchas we already hit so you don't have to. Two things I baked in on purpose: - It's tool-agnostic. Everywhere it names a tool we happen to use (Slack, Google Drive, Vercel), it gives you alternatives right beside it, so you map it to YOUR stack instead of adopting ours. - It's crawl, walk, run. Start with just the PM, live with it for 30 days, then grow. Everything you need to go further is already in there. 👇 Grab the full guide here: https://os.agenticsociety.com/tutorials/pm-agent