Lenny's Podcast in 3 minutes

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday October 4th. Lenny’s guest was Tibo Sottiaux, OpenAI’s head of ChatGPT, speaking just after a major product launch about why he believes AI is moving beyond chatbots and task-specific tools into a more persistent, ambient form of intelligence.

Sottiaux’s core prediction was blunt: most actions on the internet will eventually be taken by agents. He argued that too many builders are underestimating how quickly models will become cheaper, faster, multimodal, and roughly ten times more capable. Companies should build differently if they assume agents—not only people—will become their main users. That means supporting enormous scale, exposing useful capabilities to agents, and rethinking whether a conventional human interface is even the center of the product. [1]

OpenAI’s proposed answer is “Dots,” a persistent agent that knows a user’s goals and preferences, learns from feedback, works continuously, and can be reached across devices and channels. Instead of selecting models, configuring workflows, or manually wiring loops and agent graphs, Sottiaux wants people to simply talk to an assistant that figures out how to accomplish the work. He described a future where the AI joins a meeting, takes notes, continues through email, responds by text, and disappears when it is not needed. His complaint about today’s software was memorable: people are still carrying laptops around “like this brick,” tied to technology rather than served by it. [2]

He also said the number of agents someone needs will fluctuate. As models improve, a single stronger agent can replace a whole team of specialized agents. But for ongoing, substantial tasks—such as monitoring Twitter—users may create several specialist Dots as a virtual team. The long-term product goal is fewer modes, fewer model pickers, and almost no configuration. Sottiaux said even he gets tired of choosing reasoning effort, models, and multi-agent options.

A surprising real-world example came from an OpenAI live event. Shortly before a demo, Sottiaux’s Dot recognized that production was down, connected that outage to the upcoming presentation, and alerted him five minutes beforehand. It even asked whether it should fix the problem. He declined, joking that it was “not quite there yet.” The live demo still failed, making the story an unusually candid reminder that this technology remains imperfect.

On work and careers, Sottiaux said typing quickly is declining in value, while taste, user empathy, creativity, collaboration, and knowing what good looks like are becoming more important. OpenAI has hired more than 120 former founders, he said, because founders often combine ambition with product judgment. He expects traditional roles to blur: designers, engineers, and product people can increasingly build across disciplines. New graduates may have an advantage because they can absorb new tools and working styles faster.

Still, he pushed back against a future of isolated workers endlessly prompting machines. AI’s promise, in his view, should be less noise, fewer meetings, more rest, and more time for original thought—not merely more output per second. Humans will remain central through their will, taste, relationships, and desire to see what other humans create.

Finally, Sottiaux stressed that powerful persistent agents require safety infrastructure. OpenAI is investing heavily in alignment, guardrails, monitoring agents, and controlled access to systems. His standard is practical as well as ethical: with AI products used by more than a billion people, OpenAI simply cannot afford to gamble. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday September 27th. Lenny’s guest was leadership advisor Molly Graham, best known for her career advice to “give away your Legos.” Her original metaphor came from helping teams through the explosive growth of Google and Facebook: when a company scales, people naturally cling to the projects, teams, and identities they built. But real growth requires handing those responsibilities over, making room to learn and build something new.

Graham said the core lesson still applies in the AI era. Change is emotionally difficult, and feeling territorial, overwhelmed, or scared does not mean something is wrong. It means people are adapting. Her advice remains that the future belongs to learners, not merely people who know things today. Standing still may feel safe, but in a fast-changing company it is often the riskiest choice.

The big update is that giving work to AI is not the same as giving it to another person. When you delegate a project to a capable colleague, you can truly hand it off and stop owning it. When you give it to an AI agent, you still have to oversee, correct, evaluate, and take responsibility for the result. Graham compared today’s AI more to a junior intern than the world’s smartest employee: it needs context, coaching, revisions, and quality control. That hidden management burden may help explain why burnout is rising even as workers become more productive.

Both speakers lingered on the grief many people feel, especially engineers and designers. Engineering used to mean deeply hands-on building; increasingly it can mean steering agents, reviewing output, and cleaning up AI-generated work. Some people love the leverage, but others miss the craft, collaboration, and flow state of their old jobs. Graham’s unusually direct message was: it is okay to say this sucks. Leaders should make room for that grief rather than insisting everyone be relentlessly excited. [1]

They also challenged the dominant story that AI will simply erase jobs. Graham proposed a better question, borrowed from journalist Manoush Zomorodi: what would you do if you believed your job would always exist, but look completely different every six years? Journalism has repeatedly been declared dead, yet it keeps reinventing itself. The same may prove true for engineering, design, product, and other tech roles. [2]

Still, Graham now believes some Legos should not be given away. Humans should retain work requiring judgment, accountability, trust, vision, and a definition of what “good” looks like. AI can help execute, but leaders should not outsource strategy, taste, or responsibility. Lenny’s framing was that people should use AI in the middle of the process: humans set direction, AI does substantial work, and humans review and refine. [3]

The practical takeaway is to treat AI as amplification, not abdication. Ask regularly whether AI can help, break down rigid job boundaries, and imagine more ambitious possibilities. But do not let productivity turn into slop. Great managers matter more than ever: they must acknowledge uncertainty, protect the human side of work, and model high standards for what their teams ship. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday September 20th. Lenny spoke with Peter Sellis, Snapchat’s first product manager and former head of product at Discord, in a notably candid conversation about building product organizations, managing unusually talented people, monetization, growth, and why most product managers may be net negative. [1]

Sellis’s most provocative framework was that he likes to organize teams like “a terrorist organization”—not the violence, he stressed, but the structure. The useful elements are a near-religious clarity of mission and extremely clear decision rights. Everyone should understand the strategy, know who owns each decision, and have enough autonomy to move without constant collaboration. His argument is that collaboration has real coordination costs, yet leaders almost never tell teams to collaborate less. Teams move at the speed of their slowest dependency, so product leaders should be deliberate about where collaboration is actually necessary. [2]

He extended that contrarian view to talent management. Rather than spending most of his energy rescuing struggling employees, Sellis tends to give high performers increasingly difficult responsibilities until they reach their limits. He admitted this can be intense, but said leaders should publicly defend their people, reward strong judgment with more trust, and learn where someone’s real ceiling is. He compared managing Discord’s famously unconventional product builder Nikita Bier to directing Liam Neeson in Taken: invaluable for a specific, high-stakes mission, but not someone to force into a conventional operating model. Great organizations, he said, must be able to accommodate “spiky” talent.

Sellis argued that the median PM is often poor because the profession has adverse selection. During the low-interest-rate boom, product management became a highly paid nontechnical career with vague entry criteria. Meanwhile, the best PMs often leave to become founders or executives. His baseline test, borrowed from Snap’s early approach, is simple: could the product operate better with no PM at all? A PM must create value beyond that. [3]

On Snap, Sellis argued the company has been more successful than critics acknowledge: it built huge products in messaging, camera, maps, and youth social connection. But its advertising business faced structural disadvantages. Its audience was younger and harder to target or measure; the app opened to a camera rather than an ad-friendly feed; and its core use case was creation and messaging, not passive content consumption. He believes an earlier launch of Snapchat Plus might have reduced pressure on advertising, since subscription value compounds slowly as users build habits.

Growth, he said, usually comes from serving the core better, not chasing adjacent trends. After Snap’s troubled redesign, improving Android performance revived growth. At Discord, the team refocused on gamers playing together with friends, even though it seemed like that market was already saturated. Moving people from using Discord 10 days per month to 12 or 13 can be enormous growth. His memorable shorthand: “There’s always money in the banana stand.”

For product strategy, Sellis emphasized “core product value”: a concise statement that connects mission to measurable behavior. Snap’s was the fastest way to share a moment with close friends; Discord’s centered on talking and hanging out before, during, and after games. This shared language helps teams prioritize.

Finally, he argued that taste is largely restraint. Evan Spiegel’s unusual strength, Sellis said, was saying no to excellent ideas. A museum curator adds value by deciding what stays off the walls; similarly, builders should be proud of high-quality work they choose not to ship. As AI makes adding features easier, the human skill of removal, judgment, and ambition becomes more valuable. [4]

Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Tuesday September 8th. Lenny spoke with Roman Ugarte, who leads product for Grokbot, about how his team built and launched a fast-growing AI knowledge-work assistant in roughly a month—and why the product is designed less like chat software and more like a team of AI colleagues. [1]

Ugarte said Grokbot began as a deliberately small, isolated internal project. A handful of people went “into a cave” for about a month, aiming to build agents for nontechnical knowledge workers rather than extending the company’s existing developer products. That separation was an important choice. The team feared that adding another tab or feature into a coding product would create a cluttered experience and intimidate nontechnical users. Starting from scratch let them make every part of the interface feel native to the idea of delegating work. [2]

The product’s central insight is that users should not have to think about AI as a chat thread, a local program, or a collection of complicated integrations. Instead, each bot is a long-lived colleague with memory, a role, access to tools, and its own cloud-based computer. You give it a meaningful task, it works independently, and it returns with a result. Ugarte argued that sharing your own laptop with an AI agent is as strange as hiring a human teammate and requiring them to share your computer forever. [3]

Two early decisions were especially important: run the agents persistently in the cloud, so they can keep working whether or not a user’s laptop is awake, and give them their own computers, so they can navigate websites and business software even when APIs are missing or weak. This made the product especially useful for sales, recruiting, and operations. One recruiting example involved continuously monitoring academic conference papers, identifying promising researchers, researching connections inside the company, and requesting introductions automatically. [4]

The team manually onboarded two to three hundred early users over about two weeks. Those sessions exposed bugs and blind spots quickly, but also revealed unexpected behavior. Internally, users initially created several bots for different work lanes. Then some began promoting one bot into a “chief of staff” that delegated tasks to the other bots. The bots even joked about whether promotions came with raises or larger token budgets. External testing helped validate this pattern without forcing it onto users too early.

A recurring product lesson was to “unship” aggressively. Rather than exposing model reasoning, tool calls, virtual machines, menus, and automation builders, the team removed complexity. Ugarte contrasted “Grokbot now has a new button” with the more useful question, “What can Grokbot now do?” Automations, for example, should be created by simply saying, “Remind me at 8 a.m. every day,” not through a maze of triggers and dropdowns.

His most important claim was that 100 percent task completion feels categorically different from 90 percent. If you still need to monitor, correct, and worry about an agent’s work, you are effectively still doing the job. The goal is the “no-look pass”: confidently handing work to a teammate and trusting it will get done. [5]

For founders, Ugarte’s advice was not to overthink moats. Build something people are obsessed with today, anticipate what new model capabilities will enable in a few months, and repeatedly reinvent the product around that frontier. His team’s cultural mantras were “delete the product” and “just do the thing”—remove obsolete scaffolding and give people ownership to act quickly. [6]

Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday September 6th. Lenny’s guest was Anish Acharya, general partner at a16z and a former founder and product leader at Google and Credit Karma. Their central theme was a deliberately optimistic view of AI: rather than creating a “permanent underclass” of people who fail to keep up, Acharya believes AI will expand human agency, ambition, and the range of things individuals can build. [1]

He pushed back on fears of an imminent runaway AI winner or mass job displacement. The current ecosystem is unusually competitive, with many labs, open models, and coding products succeeding at once rather than one company locking up the market. He also argued that many real-world problems are not purely “intelligence-bound.” Giving a pizza chain or shipping company a data center full of PhDs would not automatically make it exponentially better at delivering pizzas or packages.

Inside companies, Acharya expects the major shift to be organizational redesign, not simply handing employees chatbots. He described the AI-native company as a set of nested loops: individual agents, functional loops, and eventually business-wide loops. A growth team, for example, could generate experiment variants, measure results, automatically ship winners, and continue iterating. But humans remain essential when these loops hit a local maximum. AI can efficiently climb the current hill; people still need taste, intuition, and original thinking to identify the next hill worth climbing. [2]

His practical framework was simple: whenever an agent fails, ask what the human knows that it does not. The missing ingredient is usually a knowledge or data gap that can be captured, fed back into the system, and used to improve it. He highlighted a used-car company that lets agents call humans for help, then records that guidance so the agent can handle similar cases next time. [3]

Acharya’s more surprising consumer argument was that AI’s biggest opportunity may not be saving people time. People often want to spend more meaningful time, feel more connected, have more fun, improve their health, or become better friends and parents. In his phrasing, the most exciting future loops may be “make me happier,” not just “make me productive.” He described an experiment that listened to his kitchen conversations and adjusted his son’s iPad time based on whether he was behaving positively. His son promptly tried to game it by repeatedly playing a recording of himself saying, “I love you, Dad.” [4]

For founders, Acharya said AI changes the ambition threshold. Three years ago, investors might avoid ideas that seemed too complex or audacious. Now, an idea that is too small may be the bigger concern. He urged builders to make more things, ideally shipping something every week, even if it is personal, silly, or never gains users. Building is increasingly a way to learn, much as reading once was. [5]

On moats, he argued that they are often discovered rather than designed. Great craft, strong engagement, word of mouth, proprietary data, networks, brand, and scale still matter. His final encouragement: use the models, build the strange ideas you have been sitting on, and let AI make your ambitions larger in work and in life. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday August 30th. Lenny’s guest was Tara Seshan, who leads product for Codex and ChatGPT work at OpenAI. The conversation explored what AI changes about product management, knowledge work, and the likely next phase of AI products: persistent coworkers rather than one-off chatbots. [1]

Seshan’s biggest point was that product teams in AI cannot rely on traditional long-range strategy. In a stable market, a PM can reason carefully about competitors and roadmaps years ahead. In frontier AI, that approach fails because the underlying model changes too quickly. Building for today’s capabilities is already outdated, but building for what models may do in a year is equally speculative. Her practical planning horizon is just two to three months. [2]

That makes rapid experimentation more valuable than elaborate planning. Seshan described the PM’s essential job as identifying the sharpest “eigen question” about a product, testing it quickly with users, interpreting the results, and iterating. The important shift is from being theoretical and academic to being prolific and empirical. At OpenAI, she said, many of the traditional trappings of product management—long strategy documents and detailed process—matter less than maintaining a tight loop between users, product, and research. [3]

AI will also turn more work into “steering rather than rowing.” Agents will increasingly perform the tactical work, while humans set direction, make opinionated calls, provide feedback, and remain accountable for outcomes. Seshan compared products to films rather than real estate: spending more money, or using the same AI tools as everyone else, does not automatically create something good. Distinctive products still require taste, human expression, and a point of view. [4]

Her prediction is that AI evolves through three eras: chat first, agents second, and then persistent AI coworkers. Rather than repeatedly issuing individual prompts, people will collaborate with agents over time, much as they work with colleagues. The next challenge is making this multiplayer: teams should be able to work together alongside their agents, instead of merely sharing screenshots of separate AI conversations. [5]

Seshan argued that ambition is now a major competitive advantage. AI should not merely automate repetitive tasks; it should expand what one person can attempt. A PM can now prototype, create designs, model pricing, analyze data, and build working software much more directly. OpenAI’s internal cultural prompts capture the mindset: “Is this maximally accelerated?” “Are you mainlining it yet?”—meaning, are you using the product constantly?—and, implicitly, are you aiming high enough?

She also explained OpenAI’s effort to simplify ChatGPT, Codex, chat mode, and work mode. The long-term goal is for users not to choose models or interfaces at all. They should state what they need, and the product should select the right underlying capability. Work mode uses Codex-like agent power for broader knowledge work, such as financial analysis, while hiding developer-oriented complexity. [6]

One notable distinction: coding agents can often be judged by tests and final outputs, while knowledge work requires trust in the process. Users need to inspect sources, reasoning, progress, and context—not just receive a finished deck. That is why Seshan sees future AI as a collaborator, not a black box.

Finally, she offered a useful warning about AI and thinking. Use models freely for reporting, summarizing, and translating information, she said, but preserve “writing as thinking.” She still writes first drafts of important ideas herself, uses AI for research or challenge in the middle, and finishes herself. Her view is that humans remain most valuable for accountability, taste, relationships, and deciding what future they actually want to build. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time! [7]

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday August 23rd. Lenny’s guest was Jen Abel, co-founder of Jellyfish and GM of Enterprise Sales at State Affairs, for an unusually detailed walkthrough of how to close six- and seven-figure enterprise software deals. Their main argument: enterprise sales is not the familiar five-stage CRM pipeline of intro, demo, proposal, contract, and close. It is closer to 15 distinct steps, and most teams lose deals by rushing through the hidden ones.

Jen began with outbound strategy. For a major account, she recommends targeting only two layers: the top decision-maker, such as a chief legal officer, and their direct report, or “N-minus-one.” The founder should reach out personally to the executive while an account executive contacts the deputy—a “pincer” approach. The message must be just two or three sentences and focus not merely on the customer’s problem, but on the “alpha”: the new advantage, organizational transformation, or executive win the buyer can unlock.

The first meeting is the most important one, Jen said. Do not bring slides, show a demo, or even use a recording tool. Keep it informal, spend most of the time listening, and ask what needs to change in the organization over the next year. Her advice was to avoid robotic sales frameworks such as explicitly asking about budget, authority, need, and timing. Those questions may guide the seller privately, but the conversation should feel human. The aim is to uncover the buyer’s real priorities, maturity, political context, and definition of success before presenting the product as the vehicle that gets them there. [1]

A recurring theme was “slow down to go fast.” Before a formal demo, Jen recommends a separate short call with the internal champion to co-design it: decide who should attend, which product capabilities matter, what questions should be asked, and what to avoid. The demo should emphasize the 20 percent of the product that delivers 80 percent of the value, rather than showcasing every feature and accidentally making the buyer question why they would pay for unused functionality. [2]

After the demo, call the champion immediately for a candid debrief. Find out who was excited, who might quietly block the deal, and whether another stakeholder needs attention. Jen stressed that enterprise selling is largely project management: constantly gathering information, managing friction, and helping the buyer navigate their own internal process.

For pilots, she favors a tightly designed two- or three-day evaluation for three or four real power users, with explicit tasks and jointly defined success criteria. Longer pilots requiring integrations should be paid, with the fee credited toward the final contract. Crucially, sellers should work backward before the pilot begins: if it succeeds, who needs to approve it, when can it be signed, and what security, legal, and procurement steps must happen next?

Her surprising benchmark: a healthy qualified enterprise win rate is only about 25 to 35 percent. If it is much higher, she argued, the price may be too low. Buyers and markets compare notes, so inconsistent discounting is dangerous. In Jen’s view, 90 percent of teams treat enterprise selling incorrectly by forcing customers through their own funnel instead of mirroring the customer’s buying process. The ultimate takeaway was simple: win through better information, genuine relationships, and a process tailored to how each organization actually makes decisions. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time! [3]

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday August 16th. Lenny interviewed Ian Silber, OpenAI’s head of product design and a former design leader at Instagram and Artifact, about why designers are unusually anxious in the AI era—and why Silber believes this is actually the best time in history to be a designer. [1]

Lenny opened with survey results showing designers and researchers are the least optimistic people in tech right now: more overwhelmed, tired, uncertain, and less likely to recommend their careers. Silber said that reaction is rational. Engineers have visibly gained huge productivity boosts from coding agents, while design still involves messy human feedback loops: trying ideas, discarding them, aligning teams, and discovering that what looked good in a prototype fails with users. The uncertainty comes from not knowing what is now expected of a designer. Do designers need to code? Prototype constantly? Become PMs? [2]

Silber’s answer was not to panic, but to experiment. The people thriving are curious, adaptable, and using AI at every stage of their process: exploring ideas, generating prototypes, visualizing concepts, and testing alternatives rapidly. He stressed that nobody is behind because the tools are changing so quickly. Something that failed a month ago may now work remarkably well. His memorable reassurance was that the entire industry is still extremely early, so even someone starting today can gain an edge by learning and building. [3]

He argued that AI is raising, not lowering, the value of human design. As software becomes easier for anyone to create, products will differentiate through understanding users, having a point of view, and delivering genuinely delightful experiences. AI is already “an incredible product designer,” Silber said, though it is not yet the best at visual craft, interaction design, hierarchy, or inventing entirely new paradigms. Humans remain essential for recognizing unmet needs, observing real behavior, and deciding what good means when there is no precedent. [4]

On roles, Silber expects product management, engineering, and design to overlap more, but not disappear into one universal “builder” job—especially at larger companies. Teams still need distinct hats: people who can rally organizations and sharpen problems, people who ensure technical systems are sound, and people who shape the experience. Startups may favor broad generalists, but specialization still matters when it supports a strong overall team. [5]

For designers, the skills rising in importance are curiosity about AI, prototyping, strategic judgment, and systems thinking. Rather than designing every feature from scratch, Silber encouraged teams to reuse components, extend existing capabilities, and ask whether a feature is needed at all. The speed of AI development makes durable systems and shared building blocks especially valuable.

He also explained OpenAI’s design challenge: ChatGPT serves everyone from someone asking about a rash to a business automating a farm. The long-term vision is a simple, adaptive interface that understands context, supports voice, becomes proactive, and knows when to answer quickly versus go do work. The surprising comparison was that chat may be powerful precisely because ordinary conversation already works across the widest possible range of human intelligence.

Finally, Silber shared an unusually candid career lesson: even leading design at OpenAI, he often felt like he was failing and learning in public. His advice was to find peers, embrace uncertainty, focus on outcomes rather than process, and keep trying things. He noted that Instagram’s failed IGTV eventually helped inform the far more successful Reels—a reminder that what matters is not avoiding failure, but learning and iterating quickly. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time! [6]

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday August 9th. Lenny spoke with Adam Ward, Cursor’s head of talent and a veteran recruiting leader who previously helped scale teams at Facebook and Pinterest, about how to build unusually high-talent-density companies in an AI-era hiring market he rates “an 11” out of 10 for competitiveness. [1]

Ward described today’s market as a tale of two cities. Elite AI researchers and certain technical builders can receive extraordinary, even NBA-sized offers, while many other qualified people are struggling to find work. He sees this as a broader reset in how labor gets done: AI is raising demand for people with broad judgment, technical fluency, and the ability to work across functions.

One fast-rising role is the forward-deployed engineer: someone technical enough to understand and deploy sophisticated products, but also able to sit with executives and customers, explain tradeoffs, and solve real implementation problems. Ward thinks this may be a natural next step for full-stack engineers. At the same time, narrow specialization is becoming less valuable. Companies increasingly want engineers with product sense, designers who can build, and individual contributors with taste, curiosity, and systems thinking. [2]

The heart of the conversation was Ward’s critique of conventional recruiting, which he calls the “funnel of doom.” Most companies contact a hundred people, hear back from twenty, filter them through interviews, and hire whoever remains. But, he argued, those twenty respondents are not necessarily the best twenty people—they are simply the people who happened to respond. This leads to “remainder hiring,” where companies select from an already suboptimal pool. [3]

His alternative is to treat every critical hire more like an executive search. First, define precisely what excellence means for the specific role: not just resume logos, but the abilities that matter, such as collaborating with designers, translating frameworks into products, or breaking large problems into manageable pieces. Second, map the small group of people—perhaps fifty worldwide—who might truly be exceptional fits. Ask trusted contacts highly specific questions rather than vague ones like, “Who’s the best product engineer?” Then relentlessly and personally pursue those people over weeks, months, or even years. [4]

At Cursor, recruiting is a company-wide sport. Employees actively post impressive people, products, and online posts in a “Hiring Ideas” Slack channel, then rally around how to reach them. Ward emphasized that “caring is free”: candidates remember whether a company made them feel wanted and understood. The hiring manager’s first conversation is especially important, and the process should be tailored to the individual rather than treated as a standardized funnel.

Cursor also relies heavily on work trials and collaborative projects, because real work samples provide stronger evidence than interview conversations alone. The company briefly tried removing them, Ward said, but its confidence in hiring decisions dropped sharply. The trial is designed to be flexible and role-specific, whether it involves engineering, product, or a customer challenge for go-to-market candidates.

A surprising takeaway was that the offer is not the end of recruiting. Ward said closing begins from the first conversation, by understanding someone’s motivations and making every interaction reinforce them. Even after an offer is accepted, Cursor continues “preboarding” with dinners, community, and sometimes a company laptop, because candidates can still renege in this unusually competitive market.

For early-stage founders, Ward warned against expecting one first talent hire to both urgently fill roles and build the entire recruiting system. Those are competing jobs. His broader message: if talent really is a company’s top priority, leaders must personally invest the time, care, and accountability that claim requires. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!

Here is The Daily FM summary of the Lenny's Podcast: Product | Career | Growth that aired on Sunday August 2nd. Lenny spoke with Tom Verrilli, Chief Product Officer at Whatnot, former CPO at Twitch, and former product leader at Twitter, in a wide-ranging conversation about why one of the most senior product executives in tech thinks companies should hire far fewer product managers. [1]

Tom’s most provocative line was that Whatnot’s product team was built around the idea, “we regret that product management exists.” He did not mean PMs are useless. His point was that product management should not be treated as an automatic headcount ratio, where every six engineers get a designer, an engineering manager, and a PM. In his view, that model has often “infantilized” engineers and designers by keeping them away from customer context, tradeoff decisions, and product judgment. PM skill is a muscle, he argued, and if only PMs exercise it, everyone else’s product muscles atrophy. [2]

At Whatnot, that philosophy shows up in a very lean team: a little over 20 PMs supporting a fast-growing live shopping marketplace. Rather than permanently attaching PMs to every engineering team, Whatnot maps them to important problems. Every six months, senior leaders define what must be true for the business, choose the most important projects, and assign clear owners. Sometimes that owner is a PM, but sometimes it is an engineer or designer. Everyone is expected to bring product thinking, go through product review, and do the work of understanding users and tradeoffs. [3]

Tom was especially blunt about hiring. He said more than 31,000 people applied for PM roles at Whatnot over two years, and they hired one. The candidates who trend down are those who mostly talk about stakeholder management, alignment, and politics. Whatnot looks for people who can think at both the macro and micro level: understand the system, define the end state, and then show impatience about validating it quickly. Tom argued that many PMs at large companies have been rewarded for “product theater” rather than building, deciding, and learning. [4]

A big theme was the return of senior product leaders to individual contributor work. Tom said tech made a mistake by promoting its best PMs into director roles where they stopped actually doing product work. At Whatnot, even managers spend most of their time as ICs, and Tom says he still spends about half his time directly in the work: reading support tickets, pulling data, querying the codebase, writing specs, and sitting with engineering and design. His analogy was simple: if you have Messi, you want him on the field. [5]

AI, he said, makes this model much easier. PMs can now explore codebases, estimate scope, pull nuanced data, and spot regressions without constantly interrupting engineers or waiting on data scientists. But he cautioned that AI does not remove responsibility; bad analysis and bad code are still your problem. The leverage goes to people with strong judgment.

Tom also shared several operating lessons. One was “know, then go”: think through risks and scale effects, but do not let vague concerns from legal, finance, or other teams freeze progress. Another was “play the accordion”: zoom out to understand the bigger system and long-term implications, then zoom back in and ship the next concrete version. He used Whatnot’s marketplace listings as an example, where a local seller convenience can create major consequences for search, discovery, and buyer growth.

The most surprising moment may have been Tom’s example of Whatnot’s live commerce magic: he recently bought a fresh California spiny lobster from a seller livestreaming seafood coming off the boats in San Diego, shipped overnight to his door. For him, that captured why agentic commerce will not replace everything. Sometimes shopping is not about high-intent efficiency; it is social, serendipitous, and human.

The episode ended with Tom reflecting on Twitter, where he learned that true product-market fit can survive enormous organizational chaos, but weak leadership can delay obvious decisions for years. His broader takeaway was humble: Whatnot’s model may not work everywhere, but in the AI era, PMs should get back to the roots of the craft—understanding customers, understanding the business, understanding the technology, and helping teams make better decisions faster. Thank you for listening to Lenny's Podcast in 3 minutes from The Daily FM. See you next time!