Hold It Lightly

Yesterday I rebuilt an iOS app from scratch in Swift, in a single day, with a PM agent orchestrating copy decisions in the middle of the session. I’m a designer. Six months ago I hadn’t written SwiftUI. Six months ago “PM agent orchestrating mid-session” wasn’t a phrase I’d said out loud.¶

That’s the actual job description now.¶

Most “AI-native” posts I’ve read this year are about workflow stacks. Prompt frameworks. Tool-combo decks. Copilot routines. Every conference talk a new playbook. I get the instinct. “I have a system” is comfortable for senior people who built careers on having systems. I’m not sure it’s where the work lives anymore.¶

The half-life of any of those playbooks feels like weeks now. The thing I locked in last quarter isn’t what I should be doing this quarter. The model changed. The tooling changed. The agent capability changed. The shape of what one person can ship in a day changed. By the time a workflow is stable enough to write down, it’s already behind.¶

I came up through a school of senior leadership that said: develop process, codify it, scale it through your team. I learned it from people I respect, and it worked because the underlying capability was stable for years at a time. It moves every couple of weeks now. Lock something in and you’ll find yourself six versions behind by the time the team adopts it.¶

So being AI-native, to me, in 2026, isn’t a posture. It isn’t a stack. It isn’t a workflow diagram on a slide.¶

It’s one thing: a willingness to drop what you knew worked yesterday.¶

That sounds simple. It’s the hardest move senior people have to make, because we’re senior precisely because our process used to work. The muscle memory of “this is how I do it” is what I keep catching myself in. Not abandoning it. Loosening. You hold what you know, but you hold it lightly enough that next week’s tool can replace it without it feeling like loss.¶

What I’m trying to hold onto, week to week:¶

The process you wrote down is already wrong. Not “going to be wrong.” Already. The model that came out this week didn’t read your doc. Treat the doc as a snapshot, not a constitution.¶

Cross domains constantly. The fastest way to stay AI-native is to keep doing things you don’t know how to do. Designer ships an iOS app. Engineer writes the brand strategy. PM cuts the demo film. The tools have collapsed the cost of crossing. The only thing left is the willingness.¶

Be embarrassed by last quarter’s work. If you look at what you shipped 90 days ago and feel proud, you probably didn’t move. Pride at quarter-end is a lagging indicator of stagnation.¶

The leaders I learn from most are the ones who, today, wouldn’t defend the process they used last quarter. The ones I worry about are the ones whose deck still has the workflow diagram on slide three.¶

Adapt more. Lock in less.¶