A lot of talented people are optimizing hard for a signal that just stopped working. They're polishing the resume, collecting the logos, sharpening the framework answers. All of it is reasonable, and all of it is being graded on a rubric that companies are quietly throwing out.
The hiring market is splitting in real time, and the crack runs straight through the middle of careers that looked safe a year ago. Nikhyl Singhal calls this the most chaotic period in the history of product management, and puts a number on it that should stop you: companies will shed 30,000 people and rehire 8,000, all AI-first. That is not a hiring slowdown. It is a 73% net reduction with a different population walking back in the door. Most of the survivors are new hires, chosen for something the old process never tested; only a few are incumbents who reinvented in time.
The clearest tell is what companies now put candidates through. When Sierra scrapped traditional coding and algorithm rounds and rebuilt the whole format around three phases, plan a product, build it solo for two hours with any AI tools you want, then defend the result in a review, they were testing something the old process never reached. They replaced the coding phone screen with system design, because, as they put it, vibe-coding an app is easy now; the hard and relevant problem is getting it into production at scale. Aakash Gupta sees the same shift from coaching hundreds of PM candidates: the 45-minute vibe-coding round now shows up at companies including Google, Figma, and Perplexity, and generic frameworks and rehearsed behavioral answers no longer pass.
Read those two changes together and the new filter is obvious. The model executes now, so the job rewards judgment under a tool: can you scope a messy problem, drive a product decision, and tell which of the outputs in front of you is the right one. That is exactly the capability AI does not supply and can't yet be trained to fake.
Here is the uncomfortable part. The thing companies now screen for is diverging from the thing people spent a decade making themselves good at. Roughly a third of product roles I see posted are AI-flavored now, and a small fraction of the people in them have ever shipped an agent end to end. The gap between what made you valuable and what gets you hired is the fork, and most people can't see it because they're standing on one side of it.
I watched an early version of this inside my own team. Everyone had the same tools and the same access. Within weeks, a handful of the ten had pulled away from the rest, far enough that there was no going back. What set them apart was the judgment they brought to the tools; the hours they logged in them barely mattered.
So the question worth sitting with, whether you're hiring or being hired: are you measured on what you can produce, or on what you can decide? The market now pays for the second, and the first no longer buys you time.