The way we build has been on my mind lately. Part of it was @Ruben Aguirre ’s post about memory and drift and the system getting away from him. Part of it was a conversation with @Curtis Hays. Curtis kindly gave me permission to share the following, which is all his: The frame is the one from the thread @Geoff Thilo started. Two layers everybody can see: where the work lives (the folders), and what the model can see when it acts (the context loading). There's a third nobody builds for — discipline. What keeps the first two true after the day you set them up. People conflate organizing files with keeping them true over time. They're not the same job. I keep an index of every client and which version of their brand doctrine is current. I wrote it once. Then a few weeks later I touched it once more for an unrelated change. That was it. Two edits in two months while the actual client work churned underneath it — new versions, new clients, a half-dozen files that moved. By today the index was just wrong. It froze on the day I wrote it and the world walked past it. Here's the part worth noticing. It didn't go stale because I got lazy. It went stale because nothing in the system owned keeping it current. No rule, no gate, no step that said "when a doctrine file changes, the index re-derives." I went looking — there was nothing. It was a hand-built snapshot with no maintenance contract attached. An artifact like that was always going to freeze. The only question was how fast. So that conversation from Curtis got me thinking. I build a lot slower than some people in here - and I’ve already tagged two of them in this post. Part of that is because I spend time worrying about what Curtis calls the discipline. I stop and think about the boring stuff when I build. How is it going to get maintained? Does the skill force it or suggest it? I run a fortnightly reflection for my personal system. For my work one, I run a weekly sweep *and* a fortnightly reflection. I think the discipline is important enough that I baked a version of this into a competition entry in here - adjusted from my own practice to suit use by someone who is not a builder and stop the system from disintegrating over time.

Remember the post from just a couple of days ago (feels like eternity) where I caught my AI hoarding my whole company in a folder it could see and I couldn't? The one where I called it a filing cabinet that kept everything in its pockets? It's fixed. And it got fixed because of this skool community, so this is the well deserved follow-up. Quick recap for anyone who missed it: my AI had spent weeks quietly stashing skills, workflows, and 149 facts about my business in its own private home folder, outside my repo, where my team couldn't see it and a dead laptop would have erased it. I ranted. You showed up. The evidence is here: https://www.skool.com/cliefnotes/my-ai-has-been-hoarding-my-whole-company-in-a-folder-i-cant-see-and-it-knows ------------------------------------ What we actually built (three layers) ------------------------------------ 1. A router. A plain table that tells the AI, for every kind of task, exactly which folder the output goes in. It doesn't get to "decide" anymore. It reads the row. 2. A cleanup. We went through all 149 buried facts one at a time and moved the ones that belonged in the repo into the repo. 95 of them. The drawer went from 149 down to 52. Most of what was in there never belonged there in the first place. 3. A hook. This is the important one. It's a tiny script that runs before every single save, and if the AI tries to stash anything in that hidden folder, it blocks it cold. I sat there and watched it block my own AI in real time, trying to do the exact thing that started this whole mess. I have never been so happy to get told no. ------------------------------------ The giant thank you: @Mira Bradshaw ------------------------------------ @Mira, I need to make this loud. You showed me the path. A routing table and create-memory method ARE the backbone of everything I just described. I took what you shared, adapted it to my setup, and it worked. The cleanup does not happen without you. Thank you, genuinely. You saved me weeks, a real headache, and a couple of gray hairs I can't afford with the houseful I've got at home.

--------------------------------------------------------------------------------------------------------------------------------------- My name is Malcolm, I'm a recent Sound Engineering graduate out of Bristol U.K and in the last few months I've become obsessed with leveraging AI to make cool things. Ever since installing OpenClaw on an always-on laptop to the side of my desk and getting it to run a daily cron-job, I've been hooked! That was at the start of May, but I think I was jumping the gun a bit with this next part; looking for quick cash and convinced by AI gurus, I founded a humble AI-for-small-businesses and consulting company called 'Joy Automations' towards the end of the month to implement AI for small businesses in my area in Hampshire, England. Rightfully or not, I trusted that my few weeks of playing with OpenClaw would qualify me to earn top-dollar for implementing AI into local businesses, and I set out to speak to some about the services my new company could offer. I recieved mixed reactions and thinking from business owners about AI integrations, which is not unexpected for my beginner business: what was worrying me more was the complete inadequacy, incompetence, and forgetfulness that my personal agent would often exhibit when I'd ask it simple things. How was I to advice for local businesses if I couldn't even get my own agent to be reliable? That's when I found Jake's videos and ICM-related ideas, initially on social media. I didn't understand much of it at first, but I figured I'd pay attention since he's clearly set apart from many other AI creators on Instagram with his message, priorities, and style. Not trying to sell me something was a big part of that too. I have since watched his YouTube videos: Long story short, I've realized that it's worth it to slow down on the business side of things for now and get my foundations in order; to build a strudy agentic foundations for all of the work I do going forward, with ICM and durable context at the forefront!

In a previous post I mentioned a tool I built that helped reduce my co-worker's workload by a day. Here's the other part of the story and the reality of building tools for other people. The initial wow factor was immediate, it showed to my co-workers and my boss that this type of workflow improvement was possible with the help of AI. But then we had to actually put it into production. What the demo bought me was buy-in, the hard part was executing. On first actual production run with my co-worker I realized I had misunderstood parts of the workflow they described to me, and so we immediately hit friction. It didn't operate how they assumed it would and we ran into bugs because he wasn't following my thinking when executing the tool. I learned a hard lesson: Building the Tool is the Easy Part -- Building a Tool for someone else is Hard So we had to abandon the first run and I sat down with my co-worker to really drill down into what they do exactly, step by step, and once I really got the way they wanted to work then the tools naturally fell into place. They've since been using them on their own and have said it's even better than their wildest dreams. I also built a set of tools for another co-worker on another project where I learned the same lesson. It was a set of debugging tools that was designed around my thought process of how I would go about solving the given problem. And it was impressive because it made our project possible whereas we would have sunk excessive man hours on grunt work. But I didn't take into consideration the tech hurdle for my co-worker, and it took me an hour and a half just to walk through it, at the end they said "this is great but I wish it did this." It was a novel idea I hadn't thought of because I thought it wasn't possible. But I asked Claude anyways and to my surprise we came up with a custom fit solution using parts of the scripting API that were too dense for me to learn in the past. And what we ended up coming up with was even better and fit how my co-worker wanted to work.

My dad, when you asked how he was, said one word back. "Strugglin." Most people say "good" or "fine." It's the armor we put on before we even think about it. My dad didn't. He'd tell you the truth, and then he'd watch what the truth did to the room. Some people froze — they had no idea what to do with an honest answer. Some leaned in and asked how they could help. He wasn't testing anyone. But it told you something about a person, depending on how they took it. He went to the same diner a lot. Some of the waitresses couldn't stand him for it, thought he was just a cranky old man. Others treated him like another grandfather, and he treated them like a granddaughter. He'd go to their high school games, get them graduation cards, and help fix their cars. Same man, same word, two completely different reactions. The word "Strugglin" sorted the room. It took me a long time to understand that "strugglin" wasn't him being negative. He was a Christian, and he'd read enough of Romans 5 to know what struggle was actually producing — character. He wasn't going to pretend the struggle away just to make a conversation more comfortable. That was his posture. Honest, every time, and let it land where it landed. I've been thinking about this a lot today, since it's Father's Day, and I'm thinking about where we are all spending our time. We're all building systems right now. ICMs, second brains, orchestrators, full app builds. I've been reading a lot of the posts and comments in here, and the more I think about it, the more I realize the thing everyone's actually reaching for isn't a system at all. It's a posture. A way of meeting the world that's theirs and nobody else's. The system is just where we try to put it. @Ruben Aguirre wrote two posts in here this week that got at this better than I could. He owned a hard one — scoped way more work than he got paid for and ended up with a furious client. And one of the lines that stuck with me was his: "AI without my brain on top of it is the problem." Not AI. AI without him on top of it.