This is the fear every builder carries around right now, and in some ways it’s earned but in other ways, it just means you were building at the wrong abstraction. In most fields, once you get past the entry level, there is no such thing as "best". There’s no data sheet for it. Ask ten senior engineers to name the best code and you get ten answers, and every one of them will be right by their own lights. A guy named Gerald Weinberg wrote a whole book about this in 1971, The Psychology of Computer Programming. Programming was never really a technical activity. It’s a human one. What counts as good code gets decided by people, arguing, reading each other’s work, carrying their scars from the last project into the next one. The machine ran either version fine. The “best” part happened entirely inside someone’s head. And that’s true up and down every high level field I’ve worked in or near. The best compliance review, the best sales call, the best way to structure a deal. None of it sits in a manual because none of it is a fact. It’s a position, formed by one person, over years, out of everything that went wrong in front of them. Which is exactly why a model can’t converge on it. You can train on every line of public code ever written and you still won’t get Weinberg’s room full of engineers to agree, because there was never an answer to find. There were only people, each with a working theory of good that they’d bet their name on. So when a member in this community sits down and writes their package, they aren’t transcribing the industry standard, although many may do that in the first pass. They’re publishing their opinion of best, the one they built by doing the job, and a company that licenses it is buying that opinion on purpose just like you would a salary. That’s what hiring always was, if you look at it straight. You never paid a person for the facts they knew. You paid for their particular theory of how the work should go, and now that theory has a format it can live in.

I kept trying to figure out the right folders, files, agents, and structure first. Now I think the better starting point is: What is something I already do repeatedly that I wish worked the same way every time? Once I understand the actual work, what goes in, what decisions happen, and what "done" looks like, the structure gets much easier. New rule for myself: Don't start by designing an ICM. Start with a real piece of work that already hurts. What was the first workflow where ICM finally clicked for you?

Company / role: Head of Organizational Development at a home builder that's scaling. What changed my thinking: The idea that a company has to be queryable before AI can do much for it. I recognized early that our most valuable data sits in our people's heads, and it only really comes out in meetings, and if the system can't see it, it can't use it. The token maxing idea landed too, that the real win is more usage from a lean team, not more headcount. What this means for us: I tried the obvious version first. I recorded meetings and pushed them into a second-brain setup, and it didn't work out very well, a pile of transcripts isn't a company you can ask a question. So I changed the approach, I started treating our information as one continuous gathering system, connected a bunch of our sources to Claude so I can ask about the company at any time, and built dashboards off that. The shift is from storing what people said to making the whole place legible. What needs to become more visible: The knowledge that only shows up in meetings, and how our people are actually using AI day to day. Right now both depend on the right person being in the room to mention it. One change we can make: I'm going to run a weekly competition for staff to share a way they're using AI, with a reward for the best one. It does two things at once, it pushes real usage instead of more headcount, and it drags the clever methods out of people's heads into a place everyone can see and copy. My biggest takeaway: If the company is queryable and everyone's actually building with AI, capturing what people know stops being a filing problem and becomes a habit. I'm still not sure how much of the meeting-brain data we can reach this way, or whether a weekly prize keeps people going once the novelty wears off.

What I currently sell: nothing yet, which is sort of the point. I run ops at a homebuilder and I've been building AI stuff internally for a couple of years. In my head what I'd sell is the build, skills, workflows, agent setups, the folder architecture. All the stuff I'm proud of and none of it is what anyone would actually pay for. The client problem: somebody senior spends hours every month assembling a report by hand out of three or four systems that don't talk to each other. It's not hard work, it's just tedious and it has to be right, so it lands on the person whose time is worth the most. That's the shape of it everywhere I look, and most people have stopped noticing it because it's just how the month goes. The outcome I could sell instead: our cost-to-complete report used to take about two hours per house close. It's fifteen minutes now. We closed roughly 47 houses in the ten months to May, so that's around 98 hours a year back, and it scales with growth. Nobody needs to know how it works to care about that. How I'd structure it: a small paid pilot on one report, priced against what they'd pay a vendor or what it costs to have someone do it by hand. Then a retainer if it holds up. I'm not chasing a share of revenue on the first one, because hours saved is clean and attributable and revenue isn't. If ten other things moved that quarter, the client knows it and so do I. What would change: right now if I did this hourly, they'd be watching the clock and I'd be justifying it. If the number is the thing we agreed on, we're both looking at the same number instead of at each other. It also changes what I'm allowed to say no to, because anything that doesn't move that metric is out of scope and we both already know it. Why this makes sense for me: I've been leading with mechanism because I used to be a science teacher and explaining how something works is my default. That's exactly wrong when somebody is deciding whether to hire you. They want the problem gone, they don't want the lesson. I've got the proof already...

Apple has announced the release of its newest next generation AI models using the M5 and M6 chips. I’m hoping Christmas comes early this year. Are you planning on buying any of Apple‘s new products?