What do you do when you've blown your annual budget by May?
A quick note before we get into it. Work is a Verb isn't an AI newsletter, and it isn't going to be one.
But every week, readers tell me some version of the same thing. They're worried. They've been laid off, or they're watching their team get smaller, or they're reading another headline about how the role they spent a decade building is going to be automated by Christmas. The fear is real, and the fearmongering coverage isn't helping. So once in a while, it's worth saying something honest about what the AI story actually is — and isn't.
This week, Uber gave us a useful one.
Uber's expensive lesson
Per Yahoo Finance, Uber spent $3.4 billion on R&D in 2025, a 9% jump over the year before. A meaningful chunk of that went to AI tools — Claude Code, Cursor, Codex — pushed across their engineering org, with internal leaderboards ranking engineers by their AI tool usage.
Then, a few months into 2026, their annual AI budget was gone.
They got what they planned for. Not what they paid for.
Here's where the story gets interesting, because it's not the one you'd expect.
Roughly 11% of their live backend code is now written by AI agents up sharply from near-zero a few months earlier.
But the costs ran way ahead of the benefits. Why? In part because Uber did exactly what management does whenever a new tool arrives: they built a leaderboard.
They told their engineers, in effect, we are tracking how much you use these tools. Uber engineers — who are very good at hitting whatever metric you put in front of them — used the tools. A lot. They climbed the board. They burned through tokens.
They did, by every measure on the dashboard, a great job.
Two weeks ago I called this dynamic the new seat time. Uber's leaderboard is the most expensive proof of it yet.
What this means if you're job hunting right now
Every week, readers tell me they're worried. Laid off, watching teams shrink, reading another headline about how AI will have replaced us all by Christmas.
The fear is real. Most of the coverage is not.
It's coming from people who don't understand how LLMs actually work, or from people who have a massive financial incentive to hype them.
(Because CEOs saying their product is good is news now, apparently.)
AI is hitting the job market. I'm not going to pretend it isn't, and I'm not going to pretend it doesn't hurt. But it isn't replacing jobs the way executives bet it would. They were wrong about the math. They were wrong about how cheap it would be. They were wrong about how effective it could be. The walkback is happening; it's just quieter than the original announcement was.
What Uber accidentally proved this quarter is that the job is not doing the thing. The job is knowing what's worth doing and executing it well.
That's always been the job.
