Nikhyl Singhal has had over a thousand career conversations with product leaders. He maps how the PM role changes at each stage of a company and why AI kills the parts most PMs hate.
Product management is really two words, and companies bold-faced the wrong one. The last five years grew huge teams of 'product MANAGERS' whose job was moving and packaging information for some other decider, and AI now does that better and cheaper. What survives and gets paid is the 'PRODUCT builder': someone with the judgment to decide what to build, whether it works, and whether it fits the system. So layoffs and record hiring are happening at once, for different classes of person.
What a PM actually is
In every tech company some people build stuff and some sell stuff, and the PM is the glue in between, connecting what you're trying to build with how to build it. Old-school enterprise (IBM, Microsoft) ran on a project manager writing a rigid PRD to hand to engineers. Consumer companies had no such playbook: the founder was the product person, and Apple famously ships with just a designer and an engineer, no PMs.
The four-stage S-curve
Singhal's core framework: a company's growth curve creates four totally different PM jobs. Stage 1 is finding product-market fit through rapid experimentation, and PMs don't exist here, only founders. Stage 2, once you feel the 'sucking sound' of real demand, you stop experimenting and add a quieter PM function for predictability, process, and getting multiple teams aligned. Stage 3 is hypergrowth, where big PM teams scale the hit product and expand into adjacent lines. Stage 4 is big-tech late stage, where you fight the innovator's dilemma and force zero-to-one bets again.
Solve real problems, iterate fast
From Google Hangouts failing: it solved an inside-the-building problem (seven code bases, one unified app) that customers didn't actually have, while WhatsApp won with a reliable text-only play in India and layered voice and video on later. Three lessons: solve a real customer problem not an imagined one, stick with things that look weak early, and win on iteration speed. Chrome shipped every six weeks vs Firefox's quarter and IE's year, so it won.
AI kills the info-mover
Three years ago, at zero interest rates, everyone had six job offers but hated their jobs, because product was responsibility without authority: packaging information for someone else to decide. AI removes exactly that drudgery, so leaders now report more joy because they can whip out Claude Code and build directly instead of depending on an engineer, designer, or boss. A morning agent summarizes every customer chat and sales call in prioritized order, which used to sound like science fiction.
Layoffs and record hiring together
Big tech faces 30 to 70 percent layoffs this year alongside skyrocketing salaries; the top 1 percent of PM pay has more than doubled in 18 months and Singhal helped negotiate four eight-figure product-leader contracts. There are more open PM roles now than ever. Companies over-hired PMs during free-money years to organize, not build, and are pulling back to where they were five years ago. The person most at risk is the mid-30s middle manager who only knows how to move information.
How to stay hireable
Brand is at an all-time low; Anthropic and OpenAI don't care what logo you carried, they can tell in an interview how modern you are with the tools. Three things matter: be current and hands-on with AI tools and have an opinion on what to build; build a durable network (he still talks to 25 undergrad classmates who brought him luck); and hold a systems-programming mindset, because when you can build anything by expression, the whole problem becomes should you build it, does it fit, and is it working.
- Product management is 'product' plus 'manager', and the last five years wrongly emphasized 'manager'; AI is now collapsing the manager half.
- PMs don't exist during the product-market-fit search; that stage belongs to founders doing rapid experimentation.
- Product-market fit is a 'sucking sound', a natural pull, and the moment it hits you must stop experimenting and start building consistency.
- The winning products at Google always started ugly; what mattered was speed of iteration, not how good the launch looked.
- Hangouts failed because it solved an inside-the-building problem customers didn't have; WhatsApp won with a reliable text-only core.
- Careers are chapters in a book, not periods in a hockey game: 50 working years at 2 to 3 years per job means 15 to 18 jobs, so optimize the sequence.
- Always join an environment growing slightly faster than you; when you get comfortable and stop learning, it's time to leave.
In their words
“Product-market fit means that you've got a sucking sound. You've built something, all of a sudden there's a natural pull.”
“Your career is not like periods in a hockey game. It's like chapters in a book.”
“There's no more management. It's all about building. It's all about being hands-on.”
Terms to know
- PRD
- Product requirements document, the rigid spec a project manager once wrote and handed to engineers in old enterprise software.
- Product-market fit
- When enough people naturally pull on your product that you can keep going; Singhal calls it a 'sucking sound'.
- S-curve of growth
- The company growth arc whose four stages (fit, scale, hypergrowth, late-stage) each demand a different kind of PM.
- Forward deployed engineer
- A technical person who embeds with a customer to solve their problem, then pulls those learnings back into the core product.
- Info-mover
- Singhal's term for a PM whose job is packaging and moving information for someone else to decide, the role AI now replaces.
- Skip
- Singhal's company, a talent-agency and community for top 'product builders', with ~125 heads of product from firms like Anthropic and OpenAI.
Nikhyl Singhal at Stanford CS 153: Frontier Systems
New to this? Come build with us.
Reading is good. Building with people is better. Our drop-ins are free and open to total beginners.