Here's the problem, close to word for word from a client (they help with healthcare claim denials, so they handle very sensitive data): "We need to use AI smarter at our company. But we have real security concerns about linking Claude and ChatGPT to our Microsoft Suite. Without linking it, it feels like we can't really do much. And we pretty much cannot let it read or write any file on our computers, for security reasons." So two questions for anyone who's been near this: 1. What would you actually call this problem? 2. Who's the right kind of person to diagnose it AND solve it, and roughly how would they solve it?

Okay, so. Yesterday in my last post I said the next post would explain where I've been. Grab your coffee. Or the whiskey. This one earns it. I'm usually the loud one in here. The 1 AM experiments, the folder systems, the posts that run longer than some of your contracts. Then about a month ago I went quiet, and quiet is not my natural state. My natural state is "please, someone make him stop." June 24 was the day the wheels came off. That morning I sat down and let our project manager go. If you've never had to do it, no amount of business books prepares you for the part where there's a real person on the other side of the screen and you're the one changing their whole year. One of those conversations you rehearse in the shower for a week and it still comes out sideways. We handled it with as much care as we could, offered the strong recommendation he honestly earned, and it still sat on my chest the rest of the day. Except the day wasn't done. A few hours later, the strongest CMO on my team gave me a 30-day notice. Same. Day. Two seats gone before dinner. Every founder in here knows the math that starts running in your head at that moment, and none of it is about marketing. ******************************************* Then it got personal. ******************************************* My wife was two weeks past her due date with our newest baby. If you've lived a two-weeks-past-due house, you know the whole family walks around holding its breath. When labor finally started, the plan was a peaceful home birth. What we got was a chaotic ambulance ride and four days in the NICU. Not because the baby needed it. Because the medical system is run by insurance companies and litigation attorneys, and "healthy baby" apparently isn't a discharge criteria anymore. Four days of asking strangers for permission to take home our own healthy kid. When we brought up leaving against medical advice, the not-so-subtle response from the doctors included the words "child protective services." We're not talking about a sick baby.

So Opus 5 has been around for a day now. What are your thoughts? Has it lived up to expectations? What are your praise and gripes?

Working with the EU on anything AI/systems is a headache in itself, so many rules to follow without even realizing that your VPS has to be hosted in the EU. Crazy stuff. Short story time, I had a project with a Medium sized company in Sweden, we were building a Facebook prospecting system. Now the thing is that they wanted a feature (quick note: I had already built the system with GDPR in mind, using Hetzner and all) that involved scraping individual profiles, which isn't really allowed. Now the question is, with AI giving you a workaround for almost anything, how far would you go before your morals stepped in. For me it's simple, if it even comes close to crossing the line legally or morally, it's a big no. Huge shoutout to @Tiffany Coyle for educating me on GDPR!

🎟️ PRIZE: FREE SEAT IN THE LYCEUM 🎟️ Pick your cohort. Technical, Business, or Creator. Your call. ---- 🇬🇧 We're back. Good morning from London. 👋 Thanks for the patience last week. Jake and I needed a few days to breathe before London Tech Week kicked off, and you all responded with nothing but support. We don't take that for granted. Now let's get back to building. ---- 📋 THE CHALLENGE Build a folder-based AI operator that handles ONE operational workflow end-to-end. You pick the workflow. This week's deliverable is one operator folder that someone could drop into a Claude project and use to handle a real business workflow without babysitting. ---- 🎯 PICK YOUR WORKFLOW The workflow is yours. Pick something specific. Pick something you'd actually use. A few sparks to get you thinking: - 🎫 Customer support triage (which tier handles this ticket?) - ✅ Content review and approval - 📨 Lead intake and qualification - 💸 Refund request handler - 🤝 Partnership pitch evaluator - 🎙️ Podcast guest pitch sorter - 💼 Freelance project intake - 📄 Resume screen for one specific role - 📅 Meeting request triage (book, decline, delegate) The more specific, the better. "Customer support" is too broad. "Refund request triage for an ecommerce store doing under 200 orders per month" is right. 📎 If you want a fully written client brief as a reference, the attached PDF walks through one example. Don't build the example. Use it as a template for how to think about scoping your own operator. ---- 🗂️ THE METHODOLOGY If this is your first comp, welcome. Here's what you need to know: This week (and every week) you're learning interpretable context methodology. Folders as architecture. Each file does one job well. Your operator is a folder with five things: - 📄 identity.md (who the operator is and what workflow they own) - 📐 rules.md (the decision logic: criteria, edge cases, escalation rules) - 💬 examples.md (decisions in action, including at least one edge case) - 📚 reference/ (checklists, templates, rubrics) - 📖 README.md (how to use it)