I was in the middle of working with Claude on a project, and started getting an internal server error that ended up being an outage. I was concerned because I was on a call with a client at the time, and was looking to get some help on a specific question that he had. Then I looked up at the top of my VSCode, and saw the Codex extension that I set up a few weeks ago. I switched out of my Claude Code window and asked Codex to get caught up on what I was working on. In less than 5 minutes, I was back to having a fully operational AI assistant with all of the necessary context needed to help me with that question and several others that came down the pipe. This is where the ICM shows its value. Claude was down for over an hour and 45 minutes, but I didn't lose productivity because my workspace wasn't reliant on the AI I was using. This is a key differentiator between a durable system and brittle agents. Don't get held hostage by AI models. Build the structure and adapt as needed.

I spent yesterday building out four old Weekly Comps. #4 through #7. I didn't submit any of them, and every one of them closed months ago. It was probably the most useful day I've had in here. What I didn't appreciate until I tried it: the winners and honorable mentions are still sitting in the stacks. That's an answer key. When you enter a comp live, you build blind and find out afterward how you did. When you build an old one, you read the brief, build your version, and then go compare it against work that already cleared the bar. And you can borrow, out loud, with credit. That changed what I built. On #5 the winner used a schedule-lock, so I adapted it into a number gate: no scripts, no frameworks, nothing until the user says one number, unhedged. On #7 I took the ordered short-circuit gate flow from the winner's two entries. On #6 I took "announce your interpretation before you act" from Mayston. Each one still needed something of its own, and having the winner in front of me made that easier to find rather than harder. I could see what was already covered and go somewhere else. The one I'd point at is #4, The Agency. The brief asks for a real-estate agency OS, and the winner built exactly that, publicly. I almost rebuilt it. Then it occurred to me that a five-folder Austin real-estate system on my GitHub, sitting right next to an existing public winner, reads as derivative. And I don't know real estate, so every handoff between specialists would have been invented. So I built the same architecture on my own back office instead. Same five folders, different vertical. The handoffs were real because I actually run them. Then I added a validator that exits non-zero on a malformed handoff card, and it caught two genuine defects in my own examples while I was writing them. None of this gets submitted anywhere. It's four repos I didn't have on Friday. If you joined late, or you looked at a comp and thought you didn't have time that week: the brief is still there. So are the winners. Go build the old one.

It’s so unreal how much is being sold to people out there off of simple misunderstanding of the tech. Once you spend some time in this community, you start to realize how much of this is everywhere. If you enjoy the video, please head to the YouTube page and like/subscribe for the algorithm 😁

I explain the details about it on today's high tea . I built a Claude skill that turns a process, an idea, or a messy folder into an ICM workspace. The folder structure does the orchestration. Numbered folders carry the order, the hierarchy scopes context, and plain markdown files hold state. One agent walks the right files at the right time and does the work a multi-agent setup would. Repo: github.com/RinDig/icm-architect 📦 What it does Two modes. 🔨 Build. You describe your work and it pulls out the structure already sitting in how you talk about it. The stages, the points where you stop and check, what stays the same every run versus what is new. Then it picks one of five proven forms and scaffolds the smallest workspace that carries the job. ♻️ Restructure. Point it at a folder, repo, or vault you already have. It reads every file, sorts each one by role, shows you a migration map, waits for your yes, then moves and checks the result. 🧩 The five forms Pipeline, umbrella, record library, knowledge bundle, context map. They mix and nest, so most real workspaces use more than one. ✅ The walk test Every result gets checked cold. An agent with no memory has to open the root, find its way, act, and report status from the files alone. If it can't, the structure gets fixed until it can. ⚙️ How to use it You can honestly just tell claude to download it from the link, but if you're using codex or something else it will just have to restructure the claude.md to agents.md Or if you want to do more Hands-On install Claude Code: drop the folder in ~/.claude/skills/icm-architect/, then say "ICM this" or "build me a workspace for X." Claude apps: zip the folder and upload it under Settings, then Capabilities. Fork it, break it, tell me what you built. 👇

Thursday, July 16 at 7:00 PM ET. Quick version for anyone who hasn't been following: The Lyceum is Eduba's 12-week AI certification program and the first credential we've ever issued. Over 3,000 people are on the waitlist and seats per cohort are limited. What we'll cover in the hour: 01 / The structure. 12 weeks, three sprints, nine live sessions, 18 hours of instruction, 12 instructors per cohort. 02 / The cohorts. Technical, Business, and Creator. Same core curriculum, weighted differently. We'll walk through how to pick yours. 03 / The competition. $250,000+ in prizes across the tiers and how your capstone feeds into it. 04 / The certification. What you have to do to earn it and what it actually certifies. 05 / The investment. What it costs, how payment works, and who should not enroll. Then live Q&A until the questions run out. One more thing. At the end of the session we're doing something for the people actually in the room. It's capped at a small number, it goes in the order people claim it, and we're not putting it in writing. Be there and stay to the end. The session is live only. No recording going out. Thursday · July 16 · 7:00 PM ET skool.com/live/XM7969jTG7L Come with the hard questions. Bring the skeptical ones too. That's what the hour is for.