Summary
Tom Verrilli, CPO of Whatnot, argues that product management has been over-scaled into a bureaucratic ratio game and needs to return to a small cadre of senior operators who do actual IC work rather than manage pods. He walks through how Whatnot runs on roughly 20 PMs for a rapidly growing marketplace, why he rejects the "hire great people and get out of their way" mantra in favor of leaders staying deep in ground truth, and how AI has collapsed the cost of data analysis, codebase comprehension, and cross-functional coordination. Take him seriously and you stop hiring pods by ratio and start assigning your most experienced operators directly to the problems that matter.
1. Product management is a trade built by reps, and pod-ratio hiring atrophies the muscles of everyone around it
The only legitimate argument for PM as a specialist function, Verrilli says, is that "it's a trade, not a qualification," a muscle built by reps, but the "HR ratio" pattern of adding a PM, designer, and EM for every six engineers turns that logic against itself. Hiring a PM for notifications infrastructure "infantilizes the engineers and the designers who are perfectly capable of making good decisions but just never had to." Whatnot instead maps its ~22 PMs to problems, not teams, so engineering teams can go stretches without an assigned PM even while shipping heavily. Rep the muscle where it matters, not where headcount ratios say it should.
2. Whatnot hired one PM from 31,832 applicants because most candidates specialized in politics, not customers or code
Verrilli's brutal stat, one hire out of nearly 32,000 applications over two years, is a filter against what he calls product theater: PMs whose "specialty wasn't technical or customer-oriented, it was politics." Trending down in his interviews are candidates who lean on driving alignment, stakeholder management, and "I didn't keep the CEO updated" failure stories. Trending up are candidates who show both macro systems thinking and micro impatience ("here's how I'd validate that very quickly"), plus specificity about things they personally decided rather than babysat at Fang-scale companies. If your interview story is about alignment, you're already losing.
3. Promoting your A-players out of the work was the original sin, and the fix is VPs and CPOs doing 50 to 90% IC work
The industry took its best PMs and promoted them into directors who "don't be hands-on anymore, your goal is just to coach," creating yo-yo PRD reviews and career-first politics. At Whatnot the ~5 PM managers spend 90%+ of their time on IC work and Verrilli himself spends about 50%; he even shipped some production code (quietly cleaned up by real engineers). The leverage argument: one senior PM with 10 to 15 years of honed instincts sees more of the board than three junior PMs and can be paid VP money across three heads instead of paying a full org chart. He points to CTOs of Workday, Instagram, Box, and Super.com becoming ICs at Anthropic as the direction of travel. Why wouldn't you want Messi playing instead of coaching the academy.
4. AI collapsed the cost of data science, codebase comprehension, and live user debugging into a single feedback loop
Hex threads now let a PM pull cohort analysis that "would take a week or two with an Amazon L7 data scientist in 2017," and Verrilli says he's spent less time talking to data scientists in the last year than ever while spending 10x more time in data. Talking directly to Claude about the codebase replaces the classic "how hard would this be" tax on engineers. Most striking, watching a user struggle on a live product while simultaneously querying the codebase to diagnose whether it's a bug or a comprehension gap creates a real-time three-way feedback loop between user, code, and observer. The feedback loop is on steroids.
5. Averages lie, and Verrilli's most repeated failure is killing features used by 3% of users who depend on them 100%
Asked for a failure pattern across his career, Verrilli points to relying on population averages without checking the individual use cases underneath. A feature used by only 3% of users looks deprecatable until you notice it's 100% of what a specific cohort does, and killing it is "kind of like a Westfield mall just turning off the power in the lead-up to Christmas" for those sellers' businesses. He echoes Bezos: when data and anecdote disagree, trust the anecdote. Averages are a seduction, not a signal.
6. "Know then go" beats endless alignment because thinking through failure modes in your head is nearly free
Verrilli's antidote to alignment paralysis is a mental discipline: before writing a PRD or shipping code, mentally simulate what happens if usage is 1000x expected, what breaks at scale, what legal or finance might object to, then move anyway. In product review he asks "what do we do if the experiment is green, what do we do if red," and if the PM can't answer, the strategy isn't real. Growth-stage companies aren't hunting 5% stat wins, they're hunting things that move the whole business, which requires playing the full tree in your head first. Think through everything, solve the important few, ship.
7. Play the accordion: pull all the way out to strategy, push all the way back into a shippable V1, repeat
Verrilli's mental model against both spaghetti-iteration and strategy-doc paralysis is a piano accordion: you must fully expand ("what are we trying to get done") to bring in air, but no music is made until you compress into a specific shipped V1. The value is created in the compression, but you must re-expand after every ship to update your model. His concrete example: Whatnot didn't need listings because sellers could hold up AirPods on stream, but zooming out revealed search-arriving buyers need them, then zooming back out revealed forcing three-minute listings on every seller kills seller throughput. Never stop moving between the zoom levels.
8. "Hire great people and get out of their way" is wrong; the CEO clearing his day to sit in the trenches is right
Verrilli explicitly rejects the fashionable devolve-everything mantra, arguing it produced verify-then-trust cultures where leaders lose ground truth and can't make good macro calls. His model is Grant, Whatnot's founder-CEO, who in reviews will say "I don't think this is right, I'm going to clear the rest of my day, let's sit and figure it out," then pull tickets, code, and data line by line with the team. This kills the "listen for yes" review dynamic where PMs just farm green lights for engineering credibility, replacing it with joint truth-seeking. Top-down works when leadership is actually in the weeds, and micromanagement is only a slur when leaders are managing from above without knowing what's true.
9. Twitter's real lesson was that leadership weakness masquerades as complexity, and product-market fit forgives almost everything
Verrilli endured nine heads of product in two years at Twitter, an era he compares to a therapist asking about your childhood. The takeaway that stuck: everyone knew Twitter had to lift the 140-character limit (Japanese users tweeted 6x more often because kanji packed more meaning per character), but leadership stalled for years through working groups until someone finally shipped it after he left and "nobody died." Editing tweets took another two and a half years of the same excuse. Balancing that, Twitter's product-market fit was so strong that Elon changed the name, brand, URL, headcount, and team and the network effects still held. Most things aren't complex, leadership is just weak.
10. Agentic commerce won't eat retail because most shopping is low-intent, and live commerce is the first format to combine internet scale with mall-style social discovery
Verrilli welcomes agents for programmatic buys like light bulbs and air filters and for high-intent taxing searches like "black shoes delivered by Thursday for a wedding," but points out e-commerce has never exceeded 20% of US retail spend in 30 years (25% in the UK) against a $7.5 trillion industry. Most shopping is a low-intent wander through the mall where a shoe store owner's taste and curation are the product. Live commerce economics break the CPM logic of entertainment: a Whatnot stream with 50 viewers is like a physical shoe store where 50 people are permanently browsing, so "you'd never close." His accidental purchase of a live California spiny lobster from EFishCo, auctioned as boats unloaded in San Diego and shipped overnight on ice, is the proof no agent was going to schedule for him. Different customer needs, not a winner-take-all fight.
Ask this episode