Last week I spoke to a room full of USC students.
They weren't excited. They were scared.
Undergrads about to graduate. Grad students who'd already paid for the credential. Every question was some version of the same one:
Did the thing I spent four years preparing for just disappear?
Here's the reality: partly, yes.
It's also the best news they'll get all year.
The Ladder Lost Its Bottom Rung
Let's not pretend the fear is irrational.
Stanford's Digital Economy Lab dug into ADP payroll data and found that since late 2022, workers aged 22–25 in the most AI-exposed jobs saw employment drop 13% compared with everyone else. Experienced workers in those same jobs? Stable. Some even grew.
Recent grads are now unemployed at a higher rate than the workforce overall. That's an inversion, and the New York Fed has been flagging it for over a year.
The entry-level job was always an apprenticeship with a job title on it. Do the grunt work. Learn the system. Earn the right to do the interesting stuff.
AI ate the grunt work.
The top of the ladder is still there. The first few rungs aren't.
The Field Just Got Reset
Nobody tells the room full of anxious 22-year-olds this part:
When the ladder breaks, everyone already on it loses their head start too.
The 15-year veteran's edge was knowing the system. The process. The templates. The way things are done here. The tools are rewriting exactly that, every quarter.
Nobody has ten years of experience building with agents. Nobody. Domain judgment still matters. But when it comes to how to use the tools, the VP and the senior are standing on the same starting line.
And the people winning this reset aren't the ones you'd predict.
They don't have the cleanest resumes. They're the ones who never fit the template in the first place.
The neurodivergent thinkers. The ones who ask "why do we do it this way?" one time too many in a meeting. The ones who never took in the corporate playbook because it never made sense to them.
That's not a coincidence. Julie Logan at Cass Business School found that about 35% of US entrepreneurs identified as dyslexic, roughly triple the rate in the general population. People who couldn't follow the standard path have always been overrepresented among the people who build new ones.
AI took that from an edge case to the main event. The tools reward people who don't know what they're "supposed" to do with them. No playbook to unlearn. No box to think outside of, because they were never in it.
The Resume Is Dead. Bring Receipts.
Here's where hiring is going. At 1752, it's already there.
The interview question that matters isn't "walk me through your experience" anymore.
It's: Show me what you built.
Show me the tool. Show me how you used it. Show me where it broke and what you did next.
A resume tells me what someone let you do. A build tells me what you can do.
The gap between permission and capability is the gap between the A players and the B players this cycle.
And the cost of building has collapsed. In YC's Winter 2025 batch, a quarter of the startups had codebases that were about 95% AI-generated. That isn't corner-cutting. That's the new normal.
You can build a working MVP in a weekend now. A weekend.
So "I didn't have the resources" doesn't work as an excuse anymore. And "I had an idea once" doesn't impress anyone.
Ideas were always free. Execution was the expensive part, and its price just fell through the floor.
Be Lazy. Strategically.
Here's the instinct I hire for. It's also how my own team runs.
Every time we hit a problem, the first question isn't "who's going to do this?"
It's: How do we never have to do this again?
How do we get lazy about it, in the productive way? Where's the leverage? Which automation gives us the most value for the least effort?
That's the muscle. Take something complex and make it automated and simple. Remove work instead of adding it.
Say a candidate walks in having automated their own job search. Or their club's sponsorship pipeline. Or their parents' small-business invoicing. That person has already shown me the only skill I'm hiring for.
The candidate with a 3.9 and a polished cover letter has shown me they can follow instructions.
The AI can follow instructions now.
"But I Can't Code."
Neither can I.
I still build.
We built the entire UI for 1752.ai. We streamlined our pipelines. We built our own analytics so we can see what's working and what isn't.
I didn't write a line of code for any of it.
Most people hear "go build something" and assume it's advice for the CS majors. For the kids who've been shipping side projects since high school.
It isn't.
Let's be honest: coding was never the skill. It was the toll booth. For thirty years, it stood between people with ideas and people who could actually make things. If you couldn't pay the toll, by learning to code or by hiring someone who had, you didn't get on the road.
AI tore the toll booth down.
The skills that matter now are the ones that were always underneath:
Knowing which problem is worth solving. Pain > novelty.
Describing it clearly. If you can explain the workflow to a smart intern, you can explain it to the model.
Being stubborn enough to figure out why it broke. Because it will break. Constantly.
Here's the fair objection. Building something you don't understand is the same outsourcing your brain trap we wrote about this summer. Ship it blind, and the first bug owns you.
True.
But understanding the problem > understanding the syntax. You don't need to read every line of code. You need to know what "working" looks like, and to keep testing until you get there. That's not engineering. That's judgment, and it's the thing no model can hand you.
Every time the tool fails you, you find another edge: what it can do, what it can't, and which parts of your future job it's about to rewrite.
Right now, that's the expertise that counts. You won't get it from a lecture. Or a certificate. Or this newsletter.
You get it by shipping.
What To Do This Week
Students: Pick one problem you actually have. Build something that solves it, this weekend. Not a perfect thing. A working thing. Then put it in front of three people and watch them use it.
People inside big companies: Find the most tedious recurring task on your team and automate it. Don't wait for a pilot program. Bring the result to Monday's meeting. You'll be the most valuable person in the room.
Founders and hiring managers: Stop screening for pedigree. Add one question to every interview: "Show me something you've built with these tools." Watch how fast the pipeline sorts itself.
The Bottom Line
The students in that room were scared because the old path disappeared.
They're right. It did.
But the old path was always someone else's path. Rungs designed by people who'd never met you.
Uncertainty doesn't mean there's no opportunity. It's what opportunity looks like before anyone's claimed it.
The field just reset. For the first time in a generation, a 22-year-old with a laptop and a free weekend can out-ship a team of ten.
So stop waiting for the ladder.
Go build something.
Sources: Brynjolfsson, Chandar & Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", Stanford Digital Economy Lab / SIEPR (2025); Federal Reserve Bank of New York, "The Labor Market for Recent College Graduates"; Julie Logan, Cass Business School, City University London, study of dyslexia among entrepreneurs (2009); Y Combinator's Jared Friedman on the W25 batch (March 2025).