A few days ago I shared my win: $1,500 for 10 long-form videos. A lot of you asked how I did it, so here are a few lessons: 1. I didn’t try to hard sell. I gave him a free sample before he committed to anything, hence making his decision risk free. 2. I saw an opportunity and took it. He posted an IG story that he was looking for editors, I took the chance and replied instantly offering him a free sample. 3. I trained Claude very well for the first video so that the remaining 9 videos were fully automated. I made a full breakdown video, including how I priced it: https://youtu.be/6FXZgLMDF8c My total time on the project was under 3 hours, and I never opened any editing app.

My son was born this week, Azrael is his name. Creation is a wild thing, I decided to spend some time (the little time I have) writing with just my own two hands as a celebration of the human life. No ai used in the script or writing, I hope y'all enjoy it. If you do please like and share and comment on YouTube.

I've been building an Operational Second Brain for over three months, and I use it every day. It's one big ICM container that holds all my work. It has no L2 at the root, and inside it live the classic ICM processes I run. What it does: 📋 manages projects and tasks 📚 holds all my knowledge, organized by department 💾 manages memory and brain profile 👥 tracks the people I work with and the roles they hold in each project 🧩 It's also an ICM expert. I put in everything I learned studying here, so when the system evolves, it evolves according to ICM methodology and with ICM processes. 🛡️ Integrity control mechanisms are planned. The goal is a proper harness on the agent's behavior, so that as the system grows the structure doesn't degrade and stays ICM. 📏 On size: ICM says keep things small, and this is a big container. I think I managed to adapt it so it stays efficient. After three months of daily use it works really well for me. 🔄 When I notice the agent's behavior degrading, I run commands that make it reread its behavior rules, and it goes back to the expected behavior pretty well. Determinism in agents is impossible, but with the right harness I'm finding you can get controlled behavior. And the actual work and outputs come from classic ICM workspaces anyway. ⚠️ The one real problem I see is the root AGENTS.md. It has grown to about 900 lines, roughly 25k tokens. It's read once, at session start, and that's what makes the system behave the way I expect. I'm working on shrinking it. It's the part that goes most against ICM 🫣 Have you built or used systems like this? How would you handle it, and what do you think?

My friend sent me a video of a runway show he saw. I asked my video worker to make a 45 second run way show with a background, then I also asked him to just make the stage with a background scene, as if I was showing what we could pay in the background of the show. I am kind of taken back by the results from just that request I gave the worker. What do you think?

Subject: Barto, Italy 🇮🇹 Threat level: Low (only dangerous with espresso) Background: 20+ years in IT infrastructure & cybersecurity. Built SOCs, migrated clouds, led teams. Spent two decades telling people “don’t click that link.” Point of entry: A YouTube video. I clicked “just one more” and woke up in a Skool community. The security guy fell for the oldest exploit: autoplay. Current activity: Building a lot of side projects (too much maybe) with Claude. Objective: Learn as much as I can about ICM and build things that work while I sleep. Recommended action: Say HI BARTO! :-) Status: Contained. Happy to be here lads