90 minutes of unfiltered product advice from Snap and Discord’s product chief | Peter Sellis
In a Nutshell
Peter Sellis designs teams with clear ideological alignment and decision-making authority, allowing them to move fast without coordination overhead, and focuses resources on top performers rather than improving average ones. Most product managers are net negative because the profession attracts non-technical talent during ZERP years and great PMs exit to found companies, leaving a distribution where the median is below average. The best product leaders exercise restraint by saying no to good ideas, define "core product value" to guide decisions, and practice "long-term greedy" thinking where delivering advertiser ROI today drives retention and growth tomorrow.
These notes were generated by AI and may contain inaccuracies.
Peter Sellis explained that he designs teams like a terrorist organization. A good terrorist organization has two key elements: an ideological culture where everyone understands why they are doing what they are doing, and a very clear organizational structure of who can be trusted with which decision.
He emphasized the hidden cost of collaborating that executives rarely acknowledge. Collaboration comes with coordination costs that assure teams move as slow as the slowest node in the system. He tries to aggressively internalize the cost of collaboration.
At high growth companies, product managers should be making decisions that the founders would have made six months, a year, or two years ago. The combination of a clear ideological goal and organizational structure where people understand who the directly responsible individuals are allows teams to achieve results without over-collaborating.
Sellis described his approach of spending time writing the best people into the ground versus improving the average people. His philosophy is: if somebody is strong and crushing it, he hands them more and more rope. If he sees somebody drowning, he does the absolute opposite of trying to save them.
Sellis believes most product managers are bad and are often net negative to the company. The great product managers are exiting the system by becoming founders and executives. The profession is set up so the median product manager is probably pretty bad.
Snap is famous for not hiring PMs for a long time. The baseline is that the company should be able to operate the product without any product manager, so any PM hired must provide value over nobody.
Sellis noted that a new skill for PMs is pushing back on things AI is adding to products. He saw Evan personally say no to better ideas than he's seen almost any other consumer startup launch. Exercising the muscle of saying no and restraint is one of the few muscles of taste you can exercise. If a museum has all of their collection on the walls, the curator hasn't done anything.
Sellis clarified that he didn't start with the goal of building a terrorist organization, but when thinking about the elements he wanted from organizations, he noticed that terrorist organizations operate this way. The elements worth keeping are autonomy and trusting people further down the chain, plus having a very clear mission ideology so everyone's aligned and decisions are quick.
The head of product at high growth companies where founders are still in charge acts as a vessel for their vision. Sellis tries to take and absorb and internalize the founder's vision, then create small phrases and language that can be used and repeated as shortcuts or syntaxes of the whole strategy.
Sellis noted that building a cult is not that different from his approach, since you can't spell culture without cult. However, cults are more about the person leading them, whereas terrorist organizations are more about a higher power or ideology. He suggested there is a lot of alpha in reading about cults and taking lessons from the penultimate moment before they do the thing that makes them well known as a cult.
Sellis managed Nikita Beer at Discord after acquiring his startup. He worked with Nikita for over a year along with phenomenal engineers who came with him. Managing Nikita is a loose term—he described it as seeking to control or direct him rather than manage him.
He likened Nikita to Liam Neeson in the Taken movie series: if your daughter gets kidnapped, he has a very special set of skills that are hard to come by and useful in certain situations. However, when you need him to raise the daughter or buy groceries, all he knows how to do is kill everybody in the grocery store.
Sellis views it as a failure state if product organizations can't handle really spiky talent. Nikita took this to extremes—Sellis once tried to get a one-on-one with him and threatened to book him on Intro if he didn't answer DMs. He hit Nikita up on four different messaging apps before getting a hold of him.
He looks up to Phil Jackson's management style, noting his ability to mesh talent of very different types that literally does not get along otherwise to create winning teams. He has done this both as a player and coach in two different places, combining eastern and western philosophies.
Sellis thinks of product leaders more like bands putting out albums. You can love a band but sometimes they have a dud album. The question is whether you stay a fan, ride it through, or just listen to the old stuff.
Nikita Beer said he doesn't believe product management is real, which connects to Sellis's view that most product managers are bad and net negative. He provided a mathematical explanation: when you're at the top end of any skill that is normally distributed, you're at the tail end. For product management, the ones who are actually product managers are on the right hand side of the distribution, so it becomes a power distribution where the median is far below the average.
This is emphasized in product management for two reasons. During the ZERP era, it became the most lucrative career in tech for nontechnical people. The definition is very subjective, and once you're in, you're compensated as if on a technical ladder. If you're very good, the skills developed are perfect for executive roles or starting your own company, so great product managers exit the system. Anyone who's in does anything possible to stay in, weighting the distribution.
Sellis noted that people who say PMs suck are usually the case because they've just not worked with a great PM. A second order effect of this distribution is that the whole profession looks bad when people work with mediocre PMs.
He thinks it's getting better because when interest rates go down, opportunity costs are lower and people do things like hire extra people. He's been focused on developing a system for evaluating product managers—not just the person but the person and the place they are at that time—to make sure the team is net positive.
Great engineers don't complain about product managers much because they know how to bend PMs to their will and don't put up with bad ones. Ironically, really bad engineers also don't complain because they can hide behind PM decisions. It's the ones in the middle who complain.
Sellis tried to be as objective as possible about Snap, acknowledging it may be his life's work with some chip on his shoulder. He identified three things that hold back Snap from an ads perspective.
First, the user base is extremely young. Snap appealed to Gen Z at an excessive rate early on, and now even Gen Alpha. There is a step change in advertising CPMs between 17 and 18 and between 18 and 21. If you're under 17, rules around tracking, measuring, and targeting ads are extremely constraining. This makes it harder to drive ROI and build performance products around this group.
Second, almost every display ad product that isn't search-based opens to the screen where you
Snapchat's core design makes advertising extremely difficult because it opens directly to the camera rather than a feed where ads can be placed. The product is fundamentally creation-oriented and visual rather than consumption-oriented, making it hostile to traditional advertising formats. There aren't established ad products that work well for messaging apps, and even the most successful Asian chat apps aren't exceptionally profitable from a dollars-per-time-spent perspective. Snapchat attempted to build a multi-billion dollar ads business on a lower-value audience, a difficult ad format, and a service that was more utilitarian than consumption-focused.
Despite advertising challenges, Snapchat achieved remarkable consumer product success with 70-100% penetration among people under 21 in almost every western country. It's the number one messaging app among teenagers, their default maps app (used more than Apple or Google Maps by this demographic), their most popular astrology app, and their most common camera. The company went public before Twitter, Reddit, and Pinterest despite starting later, and reached four billion dollars in ads revenue faster than these competitors.
Snapchat Plus subscription product faces brutal math because time works against subscription products - cohorts take significant time to mature and penetration grows slowly but linearly. The subscription model works differently than content products like Spotify or Netflix; users need to become acclimated to daily value before seeing subscription worth. Launching the subscription product four years earlier might have reduced pressure on the ads business to scale quickly, potentially allowing more freedom on the content consumption side to compete with TikTok.
The most underrated aspect of systems thinking is having a mathematical background to describe systems in mathematical terms - mapping stocks and flows, understanding probability distributions, and being able to mathematically communicate how a system works. The five whys method (asking "why" five times) is an effective practical approach to systems thinking that came from Toyota. Creating a spreadsheet model with all variables, inputs, outputs, and levers can be clarifying even if the model never leaves a shared drive.
Instagram aggressively copied Snapchat features, sometimes copying even non-working features because they assumed everything Snapchat touched turned to gold. Instagram's version of Stories became superior to Snapchat's original because it better suits the product and benefits from Instagram's strong design team. Instagram's map features and other copied elements will take time to develop fully.
Evan Spiegel demonstrated exceptional product taste by personally saying no to more high-quality ideas than almost any other consumer startup has launched. The best product leaders function like museum curators, where the current product represents only a small slice of the overall collection, and the curator decides what belongs in the product at any given time and place. Taste can be evaluated in interviews by asking about high-quality ideas that were prototyped but never shipped - products held back because the world wasn't ready or the product wasn't ready.
AI is particularly bad at removing things from designs and products, always adding rather than cutting elements. The skill of understanding what to remove becomes increasingly valuable as AI becomes a bigger part of building and designing processes. Lowering the threshold for creativity enables both existing creators to do more and allows people previously restricted from entering the arena to participate, putting pressure on creating content that deserves human attention.
Ads remain the product-market fit for the internet, and subscription products are overrated compared to ads from a consumer perspective. OpenAI is speedrunning all the ads infrastructure components that Meta developed from 2010-2020. The key mathematical mechanism is a modified VCG (Vickrey-Clarke-Groves) auction where organic value to the user factors into advertiser bids, requiring advertisers to pay more to reach people who may not want to be reached or aren't as relevant.
In chat environments like ChatGPT, the auction mechanism should maximize trust rather than organic value. When users ask for purchase recommendations, the system must balance objective answers against advertiser influence, with trust being the key variable. Amazon search demonstrates the risk of too many ads before organic results, potentially undermining user trust in the system. The critical challenge for OpenAI is mathematically instantiating the auction mechanism to maximize trust over the long term.
Ad formats at OpenAI represent a societal-level job in advertising, especially considering unknown variables like voice interfaces. OpenAI's decision to jump directly to an ad-free experience for the top tier was disappointing because ads could be additive to the actual chat experience. The format question remains open - what advertising looks like in chat environments, particularly if voice becomes the primary interface.
Writing best people into the ground versus improving average performers represents a counterintuitive approach to talent management that goes against common startup and leadership wisdom.
Peter describes his approach to managing high-performing team members by deliberately increasing their workload and decision-making authority until they begin making mistakes. Rather than rescuing struggling employees, he lets them fail independently, viewing this as a mechanism to identify true capability boundaries. He maintains a leadership principle of publicly defending team members while never criticizing them in front of others.
The approach stems from observing that confident individuals perform better when given more responsibility, creating a self-reinforcing cycle. He acknowledges this can be frustrating for team members and lacks mechanisms for helping people improve, but emphasizes that lack of attention doesn't necessarily indicate failure - some hires are trusted to execute independently.
Peter advocates challenging conventional wisdom in team building, particularly around collaboration being inherently positive. He suggests that if a principle cannot be reasonably disagreed with, it may not be substantive enough to guide decision-making.
Both at Snap and Discord, Peter learned that network-effect businesses naturally want to grow, making it more effective to focus on existing users rather than acquiring new ones. The key metric is daily active use, as most consumer products should optimize for daily engagement patterns that align with natural human rhythms.
The math is straightforward: acquiring 30 monthly users equals one daily active user, making retention improvements among existing users far more efficient than new user acquisition. At Snap, focusing on Android performance after the 2018 redesign led to multi-year growth. At Discord, focusing on intentional multiplayer gaming - despite assumptions of 100% penetration - drove the fastest growth since the pandemic by improving voice quality and reducing friction points that caused users to choose alternatives.
Peter references the phrase "there's always money in the banana stand" to describe the value of perfecting core user experiences. Gamers represent Discord's most critical, detail-oriented user base, and addressing their feedback created exceptional product quality that benefited all users.
Two approaches guide expansion decisions: identifying core product strengths (Snap's camera, Discord's voice) and finding new cohorts where these apply, or observing unexpected user behaviors that indicate demand for adjacent use cases. The latter represents users "voting with their feet" by working around limitations.
A more innovative approach involves backcasting from company philosophy rather than present constraints - envisioning ideal future solutions without current technological limitations, which led to concepts like glasses for moment-sharing at Snap.
Most companies fail to find successful expansion beyond their original product market fit. Airbnb exemplifies this - despite attempts at new initiatives, growth continues to come from refining the core marketplace experience of supply and demand matching.
The risk of focusing solely on core optimization is vulnerability to 10x better alternatives that can render existing products irrelevant overnight. SpaceX provides a model for avoiding this through deliberate self-disruption - recently announcing the sunsetting of their Falcon 9 rocket system, the world's most effective launch vehicle, to focus resources on the heavier-lift Starship needed for future Mars missions.
This represents a culture that willingly builds competitors to their own successful products, avoiding the innovator's dilemma by prioritizing mission over current advantages.
Both SpaceX and Tesla publish 10-year plans publicly, creating ambitious roadmaps that serve as challenges to competitors. Tesla even released their electric vehicle patents, demonstrating confidence that their execution capability exceeds others' ability to copy their stated direction.
AI tools are lowering barriers to building, requiring corresponding increases in ambition levels. Most humans aren't naturally wired for ambitious thinking that challenges the status quo, making this a skill that must be deliberately developed. The goal is achieving acceleration rather than just velocity - creating the impulse needed to overcome static friction that maintains current states.
Product managers increasingly need to elevate team ambition by asking what 1000x versions of ideas would look like, though this requires credibility and cannot be achieved through superficial suggestions to "do more" or "go faster."
Peter discusses several recurring pieces of advice and frameworks that have become known as "Celicisms" among teams he's worked with. One key concept is the idea of being long-term greedy in advertising. This approach recognizes that long-term revenue in advertising is causally related to the ROI delivered to advertisers today. The framework emphasizes focusing on driving advertiser ROI, even when it means accepting lower short-term revenue, because this approach leads to retention and increasing advertiser budgets over time.
The concept provided ideological cover for long-term thinking during difficult periods, particularly when working at public companies facing missed numbers or potential layoffs. The phrase "long-term greedy" clarified that the approach wasn't charity but rather strategic investment in future growth.
At both Snap and Discord, Peter implemented the concept of core product value, which combines mission with business growth levers. This framework requires defining both conceptually and through metrics what drives daily user retention.
For Snap, the core product value was defined as "the fastest way to share a moment with the people you care about," which broke down into measurable components: time to load for "fast," opening to camera for "share," and focus on best friends for "people you care about." This framework helped prioritize product decisions by evaluating whether initiatives directly supported this core value.
For Discord, the core product value was "the best way to talk and hang out with your friends before, during, and after playing games," with similar breakdowns for "talk," "hang out," and the timing components around gaming sessions.
These frameworks included proprietary internal metrics definitions that influenced growth strategies and onboarding processes. New hires received dedicated onboarding presentations about core product value that were delivered consistently for seven years at Snap.
Peter identifies three pairs of seemingly contradictory qualities that characterize great product managers:
The first oxymoron involves being confident bordering on arrogance while maintaining serious humility. This combination allows PMs to have strong conviction about product direction while giving appropriate credit to designers, engineers, and other team members who make critical contributions.
The second oxymoron requires being extremely organized and process-driven while remaining comfortable with ambiguity to the point of almost despising authority. This means maintaining checklists and processes while being willing to discard and redo them when circumstances change.
The third oxymoron combines long-term thinking with a bias for action that can be borderline annoying to others. Great PMs maintain clear visions of how they want systems to work months or years ahead while setting an urgent pace that influences entire teams and companies.
Light Spicy Take
Peter's least controversial contrarian view is that the most important career decision a product manager makes is choosing their life partner. He argues that a supportive partner provides growth effects and intellectual sounding boards that exceed the benefits of consuming professional development content.
Medium Spicy Take
The medium spicy take relates to taste and collecting, particularly observing that people with genuine taste know when to stop one step before trends become mainstream. Peter suggests that what someone has built and held but not shipped to the world serves as a test for this quality, and that knowing when to abandon something before it becomes "cringe" is a marker of sophisticated taste.
Heavy Spicy Take
Peter's spiciest take argues that creating truly great consumer products for certain user cohorts requires product managers and designers with specific inalienable traits that cannot be learned or acquired. These include gender, race, ethnicity, and socioeconomic background. He clarifies this isn't about matching demographics directly (like hiring young black men for products targeting young black men) but rather that certain backgrounds and upbringings make people better suited to understand specific parts of Maslow's hierarchy of needs that consumer products address.
Recommended Books
- "How to Get Filthy Rich in Rising Asia"
- "When Genius Failed" (about the rise and fall of Long-Term Capital Management in 1998)
- "Einstein's Dreams" by Alan Lightman (recommended specifically for consumer product managers)
Gaming Recommendation
Instead of movies or TV, Peter recommends "Clair Obscur: Expedition 33," last year's game of the year from an indie French studio. He describes it as having "Louvre-level beauty" in its story, music, and sound design, representing games as an underappreciated art form.
AI Product Recommendation
Peter highlights Hex as the AI product that most significantly changed his workflow. He notes it was one of the few multiplayer AI experiences that worked effectively, enabling real-time collaboration with data during meetings, though he acknowledges ongoing issues with latency and inference time.
Life Motto
Peter shares a quote from John Steinbeck's "East of Eden" about the passage of time: "duration has no signposts upon which to drape duration" and "from nothing to nothing is no time at all." This philosophy motivates him to be generous with his time and pursue experiences because periods that pass slowly in a positive way are those filled with rich experiences.
Peter's first product role was at Ustream, one of the first live streaming companies alongside Livestream and Justin.tv during the golden age of live streaming enabled by increased bandwidth. While Ustream ultimately failed by not focusing on gaming like Justin.tv (which became Twitch), Peter worked with notable mentors including Josh Elman as product advisor in 2009 and Matt Schlick as the first product manager who later sold Moldbook to Meta.
The company gained significant popularity through the Shiba Inu puppy cam, a 24/7 stream of Shiba Inu puppies that people used as background audio while working. Users reported the stream was "soothing" and some claimed it "saved my life," demonstrating early examples of ambient, comforting live content that created deep user connections.
Peter Sellis has an extensive background in consumer products, including work at a company that incubated Dollar Shave Club and Liquid Death.
The host expresses appreciation for the generous time spent on the episode, noting it's one of the longer episodes due to the depth of content covered. Sellis acknowledges the conversation as an opportunity to externally express product thinking that hasn't been widely shared before.
Sellis credits the host with having done more than any person to shape the product career field and improve it over the years, expressing gratitude for the time and for everyone who listened.
When asked how listeners can be useful, Sellis explains that people who listen through the podcast likely gained insight into aspects of product thinking that differ from typical product leaders who came up in Silicon Valley.
Sellis welcomes listeners to reach out if something resonates with them, if they disagree with something and want to correct it, if they have "beef" with any points, or if something fascinated them. He specifically states he loves hearing when people disagree with his views and loves hearing about such disagreements.
Listeners are directed to Peter Sellis's social media channels if they want to engage in discussion or disagreement.
The episode concludes with mutual thanks between the host and Sellis, followed by standard podcast closing remarks encouraging subscriptions on Apple Podcasts, Spotify, or other podcast apps, and requesting ratings and reviews to help other listeners find the show. The website lennispodcast.com is referenced for past episodes and additional information.
Keep Lenny's Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
