Key Takeaways
- Generative tools have altered product management more in two years than the previous twenty years combined, turning speculative spec writing into working software sprints.
- At Meta Superintelligence Labs, VP of Product Mona Sarantakos steered personal AI agent Muse by trading slide decks for 24-hour functional prototypes.
- Fast code generation makes cheap, low-effort software ("AI slop") easy to ship, destroying the advantage of pure shipping speed.
- Sarantakos shifted her top interview requirement from execution and roadmap tracking to product taste and consumer intuition.
Functional Prototypes Kill Speculative Spec Docs
Product managers spent twenty years writing thirty-page documents, managing Jira boards, and sitting in circular roadmap debates. Sarantakos, who has led product teams for more than two decades, points out that the entire craft shifted harder in the past twenty-four months than in the prior twenty years combined.
The reason is simple: you no longer need an engineering team to test an idea. Non-technical product managers can prompt working software into existence overnight.
“The fact that anyone can prototype anything, the fact that I can come up with an idea and I am not an engineer and I can build something out of it enough to be able to have a conversation,” (play from 1:02:40) Sarantakos told Lenny Rachitsky.
When building Muse, Meta's personal AI agent, Sarantakos cut the theoretical arguments that stall early-stage roadmaps. Instead of passing static slide decks back and forth, her team built functional prototypes and internal tools in twenty-four-hour loops. A live software prototype settles architectural and interface questions faster than ten planning meetings.
“I don't think we did this job because we want to write a bunch of documents,” (play from 1:04:44) Sarantakos noted. “I don't think we did this job because we want to keep a tracker, or have really painful circular conversations.” (play from 1:04:48)
If your team still spends three weeks writing requirement documents for an unreleased feature, you are moving at a pre-AI pace.
AI Slop Makes Product Taste the Only Defensible Skill
When the cost of writing code drops to zero, the volume of bad products explodes. Anyone with an API key can push a functional app to the web in an afternoon. Sarantakos calls this output "AI slop," and it is the fastest way to kill user retention.
“AI slop is very easy to put out into the world and you can get products out there very quickly,” (play from 1:06:24) Sarantakos explained. “But if they're not good and they slop, no one's going to use it. Establishing that bar and that taste is very important right now, and that's probably the number one thing I look for when I'm interviewing PMs or working with PMs.” (play from 1:06:33)
Execution used to be scarce. A product manager who could run a tight sprint, maintain a ticket queue, and deliver a release on time had a job for life. Today, those mechanical tasks are automated. Execution without discernment just produces more clutter.
What remains scarce is product instinct: knowing whether a response feels canned, whether an avatar like Muse's Jolly persona delights or annoys a user, and where an evaluation suite catches broken behavior. Sarantakos argues that product taste is now the single differentiator between software that lasts and disposable wrappers that fade in a month.
“Product taste now is the most important thing,” (play from 1:05:24) Sarantakos said. “Having that kind of high taste, high bar, product instinct.” (play from 1:05:28)
What to Do With This
Ban text-only PRDs for your next product feature. Give your product managers 24 hours to build a working prototype using modern coding agents or internal test setups, then run your design review on real software instead of a slide deck. When hiring, stop grading candidates on project trackers; give them a raw AI prototype and ask them to name the three design flaws that make it feel cheap.