"Drive It Like It's Stolen": Why Agency Beats Skill Now
When a model can hand you the skill on demand, the scarce thing is the person who just goes and does it. The case for hiring, and becoming, the one with agency.
Table of contents
The phrase I can't get out of my head is "drive it like it's stolen." It's how Notion's head of product describes the way his team ships: urgency, ownership, zero waiting around for permission.
It sounds like a culture poster. It's actually a bet about what's scarce — and I think it's the right bet.
The thing that quietly flipped
For most of software's history, skill was the bottleneck and agency was cheap. Wanting to build something was easy. Being able to was the hard part. So we hired for skill, promoted for skill, and treated initiative as a nice garnish on top of competence.
AI flipped which side is scarce. Skill is now something you can summon. Stuck on a language you don't know, an API you've never touched, a domain you've never worked in? The model closes that gap in minutes. The half-life of "I don't know how" got very, very short.
Agency didn't get cheaper. You cannot prompt your way into giving a damn. Initiative, ownership, the instinct to just start, the nerve to ship before someone says you're allowed — none of it comes from a model. It's the part that stayed scarce while everything around it got abundant.
Why agency beats skill now: models supply skill on demand, but they can't supply the will to act. When expertise is rentable and initiative isn't, the scarce input is the person who starts.
This isn't a new idea. We just stopped listening to it.
Paul Graham has been banging this drum for years. His test for a founder wasn't intelligence or even domain knowledge — it was being "relentlessly resourceful." Not the person who knows the answer. The person who finds a way whether or not they know the answer.
For a long time that was startup-specific advice, because startups were the rare place where resourcefulness mattered more than a polished skill set. AI just made the whole economy a little more startup-shaped. When skill is a commodity, resourcefulness is the moat — everywhere, not only in a garage in Mountain View.
There's an uncomfortable corollary lurking here, which is that most of us were schooled for the opposite. Sit still, follow instructions, produce the expected output, get the grade. That's compliance, and compliance was a fine proxy for value when execution was the bottleneck. It's a terrible proxy now.
Why agency compounds and skill plateaus
Here's the mechanism, because "agency good" is just a slogan without it.
When trying something is expensive, hesitation is rational. You plan, you ask, you wait — because a failed attempt costs real money. When trying drops to near zero, hesitation becomes the only real cost. The person with agency runs ten experiments while the careful one is still scoping the first. Nine fail and cost nothing. The tenth teaches them something the careful person will never learn, because they never ran it.
That's a compounding loop. Every cheap attempt sharpens judgment, and sharper judgment aims the next attempt better. Skill plateaus — the model meets you at the same level no matter who's asking. Agency keeps climbing, because it's the one input the model can't supply on your behalf.
Want to feel this directly? Pick the small tool you keep wishing existed and ship it this week, letting an agent cover the parts you don't know. The knowing is the cheap part now. The starting is the whole thing.
What it means for who you hire
Stop over-indexing on the résumé that lists the most skills. Those skills are increasingly rentable, and you're paying a premium for a thing the market is busy commoditizing.
Hire for evidence of agency instead. The side project they shipped because they wanted it to exist. The broken thing they fixed that nobody asked them to. The story where they didn't wait for the green light. That's the signal that survives this era, because it's the one trait you can't install after the hire.
The hard part: a brilliant person with no initiative is now a worse hire than a scrappy one who ships. We spent decades optimizing for the reverse. That habit is going to be expensive to keep.What it means for you
If you've been hoarding skill and rationing action, the market just inverted on you.
So lower your own bar for starting. Stop waiting until you "know enough" — the knowing is on tap now. This is the same reason the bottleneck moved upstream: when making is free, the person who decides to make wins.
Drive it like it's stolen. The only thing that's genuinely scarce anymore is the will to drive at all.
Related posts
How to Ship Your Own Internal Tool as a PM (No Eng Ticket)
You keep wishing a small internal tool existed. You can build it yourself this week with an agent covering the parts you don't know. Here's the path from idea to deployed.
What Happens After Coding Is Solved?
If writing the code becomes free, the whole job moves upstream. The bottleneck isn't typing anymore. It's knowing what to build and proving it came out right.