A mobile developer on my team recently took on a Node 12 to latest backend migration. He shrugged at it. Eighteen months ago, that same migration would have been a multi-week spike led by a senior backend engineer, with a kickoff meeting, a risk doc, and a rollback plan circulated three days in advance. This time: one mobile dev, a few days, done.

He didn't quietly become a backend genius overnight. Something else changed.

The temptation is to call this an AI productivity story. Yet another anecdote about how Cursor or Claude made someone 10x. But that misses what's actually happening. The migration didn't get easier because the AI wrote the code. It got easier because of the wall the mobile dev would have respected eighteen months ago. “I don't know Node/that's not my area/someone else should do this” : got short enough to step over.

I've been sitting with this pattern for months. I was going to write a piece about how AI created a K-shaped distribution among developers: the top of the curve gets 10x, the bottom of the curve gets 20-30%, and the gap between them is widening. That's true. But it's just the surface of the thing. The deeper pattern is that those climbing the top arm of the K aren't necessarily the most senior, credentialed, or technical.

They share a different trait. And it's not skill.

Three composites from the last six months, none of which fit the old story about who gets ahead:

A product manager I worked with in the US. His last startup failed to gain traction, and he transitioned into PM work after the shutdown. He now consults for multiple companies in parallel. The thing that changed isn't that he learned to code; it's that he stopped waiting for engineers to translate his ideas into something he could test. He gets his hands dirty. He ships POCs in a weekend that would have taken his previous team a sprint. None of those POCs are production-grade. None of them need to be. They exist to compress the feedback loop with the client from three weeks to three hours.

A non-technical founder in my extended network ships a new small app almost every month. Most of them go nowhere. A few get traction. He builds with Claude and a stack of low-code tools, and the limiting factor for him isn't building anymore; it's distribution. (More on that in Part 3.)

And the mobile dev. Not the most senior person on the team. No backend specialization. What he has is a willingness to take action across a boundary he used to respect.

None of these three share a tech stack, a title, or a years-of-experience number. What they share is a single trait: the willingness to act on incomplete knowledge across domains they don't formally own. That's the top arm of the K. That's what AI is actually rewarding.

The framing matters because the implications are different.

If AI rewards skill, then the answer is training: invest more in your senior engineers and level up your team; the gap will close eventually. Comforting story. Wrong story.

If AI rewards agency, the picture is harder. Agency is partly teachable, but a lot of it is temperament. It involves the willingness to start before you feel ready, to act without permission, to be wrong in public, and to adjust. Those aren't things you bootcamp someone into in six months.

What I actually see separating the arms of the K isn't four technical capabilities. It's four behaviors that look more like dispositions:

Willingness to act on incomplete knowledge. The top arm starts. The bottom arm waits to feel ready. AI made the cost of starting almost zero, but only for people who were willing to start without the comfort of mastery.

Critical thinking across boundaries. A frontend developer at SoluteLabs now isn't just expected to execute Figma designs. They're expected to flag when a flow doesn't make sense, when the product logic is off, or when something should be questioned upstream. A backend dev is expected to think about the user's path through the product, not just the cleanliness of their API contract. The boundaries between roles have softened, and the people who thrive are the ones who were already inclined to look past them.

Product instinct. AI removed the bottleneck on the how. What to build is now the differentiator. Devs who can think from a product perspective, who can tell when something is worth building, and who can spot a bad spec before they implement it, produce far more value than devs who execute precisely on whatever ticket lands in their queue.

Comfort being a generalist. Pure frontend specialists, pure backend specialists, and pure mobile specialists: these roles still exist at the extremes (infra, compilers, ML systems), but in the broad middle of software work, range now beats depth. Every dev at SoluteLabs has effectively become a full-stack developer, not because we made it a policy but because the work itself demanded it.

Notice that none of these are about IQ, credentials, or even raw coding ability. They're about temperament and posture toward the unknown.

If the multiplier is agency rather than skill, then training your way out of the gap doesn't work the way you'd hope.

Skill compounds with practice. You can seat a junior next to a senior for a year and watch them get better. Agency is harder. Some of it is teachable; you can create environments where people are given permission to act, where failure isn't punished, and where cross-boundary moves are rewarded. Those things help.

But a chunk of agency is who someone already is when they walk in the door. And that means the question isn't just how do we level up our team. It's did we hire for the right thing in the first place?

Most of us didn't. In the interview process, most companies still run tests for the tasks that AI now performs: algorithmic puzzles, syntax recall, and bounded problem-solving with clean answers. They don't test for the willingness to start without a clear path. They don't test for cross-boundary thinking. They don't test for product instinct.

We screened for the previous job. The job changed. Now we have to figure out what to do about it.

That's the next issue: What it actually looks like to manage a team through this shift, including the parts nobody wants to write about.

Experiencing similar traits around you? Hit reply.
I want to hear what you're building and what it is that you’re seeing.

Karan Shah
Founder & CEO, SoluteLabs
Building AI-native products before it became cool.