@Jake Van Clief I am building Onyx Commercial Roofing's system for maintaining our SOPs via github like you show in training. Here is what I have been able to come up with https://github.com/adib-gmap/OCR-SOP-structure-and-processes.git. Note: This is not the actual SOP's this is the Frame work built to Manage the SOP enviroment in a productive and scalable manner for our team. I have been writting sops for so long and managing them is hearding cats. Think of the SOP's as a individual soldier. I am trying to create guide for the O3 on how to manage a Company of soldier so your Company is dress right dress, not just in formation standing around, but also when conducting operations, the guide is how the Company moves as a whole. What rules the E8 is responsible for enforcing, if he wants to change them, how so? Same for E6's, E5's ect. I have been able to identify some gap, for each gap below I put in a placeholder rule. Is it right, what would you do instead? Changes and approvals 1. Major vs minor: MAJOR = someone has to do something differently; MINOR = more detail, no new obligation; PATCH = typos. Does a major change force retraining? 2. Who approves: the team owner approves MINOR; team owner + business owner approve MAJOR; never the author. Right people? 3. Safety / warranty SOPs (OSHA fall protection, manufacturer install specs): I added a qualified-reviewer sign-off. Needed, or overkill? Keeping it alive 4. Review cadence: safety/warranty every 6 months, ops 12, admin 24, plus an immediate review after an incident or repeated deviations. How do you set the S11 maintenance schedule: time, usage, or only when gaps show up? 5. Field deviations: the crew calls the PM first if it's safety or warranty, otherwise logs it within 24 hours. At the weekly meeting each one gets an outcome: one-off, coach the person, or change the SOP. Is that how you'd feed the S11 loop?

Found a piece of my own client work this week that had been sitting stalled for thirteen days. Fully drafted. Ready to go. It was stuck because one handoff step never happened, and nothing caught it until I stumbled on it by accident. Around the same time, I'd been asking clients a blunt question: what are you actually paying me for? AI can build the website, write the ads, draft the campaign now. Three separate conversations, no coaching between them, same answer: assurance. They're not marketers. They don't know what's happening day to day. They want to know someone trustworthy is watching it. So the plan was to automate proof. AI reads the data, drops five plain bullets into the agenda instead of a generic monthly PDF. Here's what we did. Here's why. Then I looked at why nobody caught the stalled one. I never had a name for "stuck because someone outside the loop hasn't gotten back to me" as its own thing, separate from normal waiting. Both just read as "waiting." If you can't tell the two apart, nothing will flag the difference until you trip over it by accident. If the product is "I'll make sure it's handled right," the proof of that can't be an AI summary nobody double-checks. An unchecked summary just hides the same invisible-work problem one layer deeper. When part of your process depends on a real person outside your system getting back to you, how do you tell "still waiting, that's normal" from "quietly stuck," before you stumble into it by accident? Who's actually watching for that, if you're honest? I had been disciplined about not building a UI for my ICM because I thought it was a waste of time. But now I feel like I need a dashboard for KPIs and statuses so I know what's actually going on under the hood. Anyone else running into this? And by ICM I mean my Operating System or Second Brain.

https://youtu.be/9dYcwOkpCE8?si=xR-v5G1P27L7KPIU Fun - dropped yesterday. Went from specific inflexible agents to using files and folders.

I have something that sorts my incoming work before I see it, and a while back I let it start learning from what I did afterwards. Replied, acted on, ignored, deleted. It adjusts what it puts in front of me next time. It works, roughly. But I keep hitting the same wall: I am a bad teacher and I have no way to tell it so. A week where I am underwater looks exactly like a week where nothing mattered. A Sunday cleanup where I clear a pile of things in four minutes teaches it that the whole pile was unimportant, when really I was just tired of looking at it. And the things I avoid because they make me uncomfortable are often the ones I should have opened first, so it learns to hide precisely the wrong stuff. I have patched the obvious cases. Anything cleared in a big batch counts for less. Age is taken into account before an ignore counts as a judgement. But those are guesses about my own behaviour, and I have no way to check whether the guesses are any good. What I keep landing on is that the signal is cheap and abundant and mostly meaningless, and the signal I actually want, did this turn out to matter, arrives weeks later if at all, and never in a form anything can read. So: anyone here letting a system learn from your behaviour rather than from your instructions? How do you stop it learning your bad weeks? Or have you tried it, decided it was not worth it, and gone back to writing the rules yourself?

Made an MCP that connects EdubaWare to your favorite models !