I named this newsletter The Increment because I believe in getting just a little better every day. No grand reinvention. No "we're transforming everything" keynote moment. Just steady, honest progress, one small improvement after another.

I still believe that. But I left out half the idea, and it's the half I think about almost every day now.

Here's the missing half: incrementing only works if you're incrementing the right thing. You can spend years getting better at something and still be improving the wrong hill.

There's a story I love about how the snowmobile got invented. The early attempts took a horse-drawn sled and bolted an engine onto it. Faster sled. It still got stuck in the same drifts, still couldn't climb the same slopes — it was just quicker at not doing those things. It wasn't until someone stopped asking "how do we speed up the sled" and started asking "what are we actually trying to do" — move a person over snow, regardless of what's under them — that you get tank treads and skis. A genuinely different machine, not a faster version of the old one.

That's first principles thinking. Not another idea about moving faster. A periodic check that you're still climbing the right hill before you spend one more increment climbing it.

The distinction that matters is form versus function. The form is what a thing looks like and how it currently operates — the sled, the engine, the process, the tool. The function is the actual job underneath it — move a person over snow, resolve a customer's issue, get a decision made. Most improvement effort, almost by default, goes into the form: make the existing thing faster, smoother, shinier. First principles thinking is the discipline of occasionally stepping back and asking whether the form still serves the function at all — or whether it's time to rebuild around the function directly.

I bring this up because I think most AI rollouts happening right now are the faster sled.

"Give everyone a chatbot" keeps the same process, the same handoffs, the same approvals — just quicker. It feels like progress, because it genuinely is faster. But if the underlying process was never the right one, you haven't built a snowmobile. You've built a faster sled, and you'll get stuck in exactly the same place — just sooner, and with a bigger invoice.

I've spent more than ten years in and around enterprise process work, and I watch this fork happen constantly: a team gets a genuinely capable new tool, and the entire project becomes "how fast can we bolt this onto what we already do," never "does what we already do still make sense, now that this exists." The tool isn't the problem — it's usually remarkable. The problem is nobody re-asked the function question before pointing it at the old form.

Now let me put on my other hat, because I spend my days advising leaders on exactly this.

I'm not telling you to blow up every process and rebuild from scratch. Most days, incrementing the form is exactly the right move — that's the whole spirit of this newsletter, and "first-principle everything, always" is its own kind of exhausting theater nobody has time for.

But there's a specific, recognizable moment to stop incrementing the form and re-derive the function instead. In the fullest version of this framework there are three triggers worth watching for: the improvements you're making have hit diminishing returns, a cost or constraint that used to gate you has genuinely collapsed, or you find you can no longer even state what function the thing is supposed to serve. I'll walk through all three in detail this Thursday only LinkedIn at Linkedin.com/in/isaacchiles, with real examples of each.

For today, the one that matters most is the middle one: when a cost that used to gate you just collapsed. That's precisely what's happening with AI right now. A huge amount of what used to be expensive or slow just isn't anymore. When a constraint like that disappears, the old form was built around limits that no longer exist — and you can polish that old form for years without ever noticing that the real opportunity was to rebuild it around the actual job it's supposed to do.

So here's the full idea, half of which I left out of the name: The Increment was never just about small steps. It's about small steps on the right hill. Increment the form — but first-principle the function, especially right now, while the ground is shifting under a lot of our old assumptions about what's actually possible.

What's one thing at your company that's been incremented for years — made better, faster, smoother — that nobody's asked lately whether it's still solving the right problem?

— Isaac

P.S. — This is the first of five editions applying old operating-science ideas — constraints, flow, the human bottleneck — to how companies actually get value from AI. If the "right hill" idea landed, the next one (Edition 4, Theory of Constraints) is about why most AI rollouts improve everything except the one thing that matters.