Angine de Poitrine are a duo from Saguenay, Quebec. They play microtonal math rock on a custom-built double-necked guitar-bass hybrid that took a local luthier 150 hours to make. The frets are spaced at quarter-tone intervals. The guitarist can barely see through his mask, so the fret markers are oversized and phosphorescent. Their songs are largely instrumental, built on odd meters and drawn from Indian ragas, Turkish psychedelia, and Gamelan. They’ve been playing together since they were thirteen.

In February, KEXP uploaded a video of the band performing at the Trans Musicales festival in France. It now has over 5 million views. Their debut vinyl, self-released in 2024 in a run of 200 copies, started selling for $300-$600 on Discogs. Their entire 2026 tour is sold out.

One of the top comments on the YouTube video reads: “This is the only way we can win the battle against AI.”

The flood

Here’s what was happening in music while Angine de Poitrine were building a grassroots following on the Quebec festival circuit.

By January 2026, Deezer was receiving 60,000 fully AI-generated tracks every day — roughly 39% of all uploads. Deezer’s own study found that 97% of listeners couldn’t tell the difference between AI and human-made music in a blind test. Suno, the largest AI music generation platform, generates more new tracks every two weeks than the entirety of Spotify’s catalogue. Its CEO Mikey Shulman went on a VC podcast and said, with a straight face, that making music “is not really enjoyable” because it takes too much practice. He later said he wished he’d chosen different words.

Doug Shapiro has written persuasively about what happens when content becomes infinite: attention stays fixed, so the value of any individual piece of content converges toward zero. The only things that retain value are filters, curation, and trust. I’ve written about this same dynamic in gaming: 19,000 games launched on Steam in 2025, and the median sold 500 copies. Too much content chasing existing audiences through established distribution channels.

Music just got there faster. AI-generated tracks account for 39% of Deezer’s uploads but only 0.5% of actual streams. 85% of their streams are fraudulent. The content is technically competent and functionally worthless.

Unpromptable

Try typing “microtonal math rock performed by anonymous duo in papier-mâché monkey masks playing a custom double-necked quarter-tone guitar-bass hybrid influenced by Gamelan and Turkish psychedelia” into Suno. Even if it could parse the instruction, what came out would be a statistical average of training data. It would sound like something. It wouldn’t sound like this.

AI music tools are interpolation machines. They produce the weighted average of their training data. They’re brilliant at the middle of the distribution and structurally incapable, I’d argue, of generating true outliers. Everything about Angine de Poitrine exists in the gaps - between notes, between genres, and between what algorithms know how to classify. The music sounds wrong in a way that makes your brain reach for it.

I wrote a while back about how the best products tell people to fuck off.

My argument: in a world of infinite choice, being “pretty good” is a death sentence. You have to be specific enough to create identity. The people who love Angine de Poitrine don’t just enjoy the music, they become evangelists. This is the Marmite principle applied to the age of generative AI: the less your work resembles anything a model could produce, the more it spreads.

Illegibility as distribution

The lesson here isn’t “be weird for the sake of it.” Angine de Poitrine have been playing together for twenty years. The instrument was custom-built over 150 hours. The microtonal vocabulary comes from years spent studying non-Western musical traditions. The weirdness is earned.

The real lesson is about illegibility. The early-stage investments I am most excited about are in companies that can’t be easily pattern-matched - that don’t look like obvious winners until they are. There will be “obvious” winners too, but they are likely to be priced that way.

The same logic now applies to creative work, and increasingly to products. In a world where AI can produce competent versions of anything that’s been done before, the margin is in things that haven’t been done before.

Suno’s CEO says taste is the only thing that matters and skill is going to matter less. I think the opposite is happening. Skill that produces genuinely novel output - not just technically proficient but truly original - has never been more valuable, because it’s the one thing the machines can’t fake.

60,000 AI-generated tracks hit Deezer every day, and a duo in monkey masks just sold out a world tour. I know which side of that trade I’m on.

If you’re building something at the edges — something that couldn’t have been prompted into existence — I’d love to hear from you. That’s what we invest in at F4 Fund.

Frank Rotman wrote a brilliant piece in 2022 applying this metaphor to venture capital. His argument: structural changes in the VC ecosystem (more capital, more startups, more LP demand) had destabilized the old equilibrium, and only firms that moved quickly to a new “stable point” would survive. Everyone else would spiral.

The games industry has its own three body problem. Three forces are pulling on studios simultaneously, and they are pulling in directions that are very hard to reconcile.

Force 1: The price ceiling

Gamers do not want to pay more for games.

This is not a vague sentiment. It is one of the most durable consumer preferences in entertainment. The standard price for a new game sat at $60 for roughly two decades. When publishers began testing $70 in 2020, it was met with grumbling. When Nintendo launched Mario Kart World at $80 in 2025, the reaction was outright hostility. Microsoft tried the same thing with The Outer Worlds 2 and walked it back within weeks.

Eight separate price increases hit consoles, accessories, and subscriptions in the first half of 2025 alone. And it’s not working. Gen Z spending on games dropped nearly 25% year-over-year, driven by inflation and a tightening job market. The average Steam user played just four new games in 2024 and 2025. Mid-priced titles like Clair Obscur and Split Fiction gained traction at least partly because they weren’t asking for $80.

Force 2: AI hostility

Players hate AI. Many developers increasingly hate AI too. And the numbers are stark.

A Quantic Foundry survey of nearly 1,800 gamers found that 85% hold negative attitudes toward the use of generative AI in games. 63% chose the most negative response option available. The researchers noted this level of rejection is rare in years of survey work — worse even than the reaction to blockchain gaming, which is saying something.

It’s not just players. The 2026 GDC State of the Industry report surveyed 2,300 game professionals and found that 52% now view generative AI as harmful to the industry, up from 18% just two years prior. Only 7% called it a positive force. Visual artists led the opposition at 64%, followed by designers and narrative writers at 63%.

At GDC 2026, Moritz Baier-Lentz of Lightspeed (whose portfolio includes stakes in both Anthropic and Epic Games) told a panel he was “shocked and sad” that the industry was “demonizing” what he called a “marvelous new technology.” The room did not agree. The same survey found that 28% of developers had lost their jobs in the past two years. When the people selling the technology express surprise that the people being displaced by it are unhappy, you have a communication problem at minimum and a political problem at maximum.

Studios that have used AI visibly have learned this the hard way. Even Larian faced backlash for AI-assisted concept art and Embark replaced AI-generated voice acting with human performers in Arc Raiders after launch. The lesson is clear: players and developers are watching, and AI is a reputational risk.

Force 3: Costs are still going up

None of this would matter much if games were getting cheaper to make. They are not.

Average AAA development budgets have increased roughly 8x since 2000. And the curve is accelerating, not flattening. Each generation of hardware brings diminishing returns in perceived visual quality but disproportionately higher costs to achieve them. Joost van Dreunen has been tracking this and makes a persuasive case that the trend is approaching a sustainability crisis.

Even outside AAA, costs are rising. Mobile games are cheaper to build relative to console/PC, but user acquisition costs have become so punishing that the marketing budget can dwarf development by an order of magnitude. Scopely spent over $1 billion marketing Monopoly Go — and still turned a profit, but that’s a deeply weird business model.

The trap

Here is what makes this a three body problem and not just a tough market.

The obvious solution to Force 3 (rising costs) is AI. Generative AI can meaningfully reduce production costs in asset creation, testing, localization, prototyping, and increasingly in code and design iteration. Studios that adopt it aggressively could, in theory, make games faster and cheaper.

But Force 2 says you can’t do this. Or rather, you can, but your audience and your workforce will punish you for it. An 85% disapproval rate is not a speed bump. It’s a wall.

And Force 1 says you can’t solve the problem from the revenue side either. You can’t make games more expensive to compensate for rising costs, because your customers won’t pay.

So: costs keep going up. Revenue per unit won’t go up. And the most obvious tool for reducing costs is politically radioactive.

That’s the instability. Every direction you push creates a reaction from the other two forces. The old equilibrium — make expensive games with large human teams, sell them for $60, rely on microtransactions for margin — is groaning under the weight.

Finding the Lagrange points

Rotman’s insight was that in an unstable system, a few positions of relative stability exist — what physicists call Lagrange points. These are positions where the gravitational forces balance out, allowing a small object to maintain its orbit without expending much energy.

I’d argue there are four stable positions emerging in games. Not all of them will work for every studio, and I could be wrong about some of them. But here’s my best guess.

Stable Point 1: Go small, stay human

Avoid the cost spiral entirely. Small teams, lower fidelity, no AI controversy, price at $20-30 where consumer resistance evaporates. Two of the top 10 most played games on Steam at any given time are low-fidelity games made by small teams (Everwind and Slay the Spire 2 at time of writing). This is not a consolation prize. This is where some of the best returns in the industry are coming from.

I’ve written about this before as minimum viable fidelity. The idea is simple: figure out the lowest level of visual polish you can get away with to attract your target audience, and invest your savings in gameplay depth and iteration speed.

This position is stable because it renders all three forces irrelevant. You’re not spending $200M, so you don’t need AI to cut costs. You’re not charging $80, so price resistance doesn’t apply. And you’re not using AI in your production, so nobody’s angry at you.

The catch: you need a genuinely great game to break through at this level, and even that is not enough. There is no marketing budget to buy attention. Your game has to be its own distribution. In 2025, over 19,000 games launched on Steam. The median sold 500 copies.

Stable Point 2: AI in the pipes, not on the screen

Use AI aggressively in the production pipeline — code generation, testing, asset iteration, localization, internal tools — but keep it invisible to consumers. The studio equivalent of using AI for back-office work while keeping the customer-facing layer human.

This is probably where most large studios will end up, even if they won’t say so publicly. 36% of game companies surveyed already use generative AI tools in some capacity, but usage skews heavily toward management and business professionals rather than the artists and designers whose work players actually see.

The political calculus is straightforward: players don’t care how you QA tested your game or generated your internal documentation. They care whether the art looks like it was made by a person who gave a damn. Use AI where it’s invisible, and invest the savings in the parts players do see.

The risk is that this is an unstable secret. The moment a leak reveals the extent of AI use behind the scenes, you face the same backlash as studios that used it visibly. Ask any studio that’s been caught.

Stable Point 3: AI-native, with permission

Build for the subset of gamers who actively want AI-driven experiences. Procedural worlds, personalized narratives, emergent systems, AI dungeon masters. This is Rotman’s “non-consensus alpha” equivalent — you’re betting against the 85% disapproval number and finding the 15% who are curious or excited.

There’s a real audience here. The Quantic Foundry data showed that gamers motivated by power progression (leveling up, gear upgrades) were the most positive toward AI. And a vocal minority of players — especially older gamers — are genuinely excited about AI-enabled experiences like infinite story branching or deeply personalized game worlds.

The bet is that 85% hostility today doesn’t mean 85% hostility in three years. Consumer sentiment toward new technologies tends to follow a pattern: initial rejection, gradual normalization as quality improves, eventual acceptance. The same gamers who hated free-to-play in 2008 now spend freely on battle passes.

This may not happen. The anti-AI sentiment in gaming feels is deeper rooted than, say the resistance to F2P - it’s wrapped up in concerns about job displacement, artistic integrity, and the broader cultural backlash against AI that’s happening across many industries. Maybe it softens. Maybe it calcifies. I genuinely don’t know.

Stable Point 4: Platform and UGC

Offload content creation costs to players. Roblox, Fortnite Creative, Minecraft, and increasingly any game with robust modding tools.

In this model, AI becomes a creator tool rather than a replacement for creators, and that changes the politics entirely. A player using AI to build their dream Roblox experience is not threatening anyone’s job. They’re extending the life of a platform. The studio’s cost structure shifts from “make all the content” to “build the tools and the marketplace,” which is a much more scalable and defensible business.

This is the position that best resolves all three forces simultaneously. Development costs are shared across a creator ecosystem. Prices are flexible (free-to-play platform, paid creator content, take rates). And AI is welcomed because it empowers players rather than replacing developers.

The constraint: this only works for certain kinds of games. You can’t build a UGC platform out of a narrative RPG. And the platform positions are increasingly locked up by incumbents with massive network effects.

Unstable orbits

If you’re running a game studio and you can’t see yourself clearly in one of these positions, Rotman’s framework suggests you’re in trouble. The death spiral for games studios looks a lot like the one he describes for VC firms: miss targets, struggle to raise capital (or generate revenue), lose talent, ship worse products, miss targets again.

The studios most at risk are mid-sized teams making mid-budget games that need to charge full price and can’t easily reduce their cost base without AI but can’t use AI without blowback. That’s a lot of studios.

What I don’t know

I want to be honest about the limits of this framework. The three body problem is a useful metaphor, but metaphors simplify. The real system has more than three forces — platform economics, regulatory risk, cultural shifts in what “gaming” means — and the stable points I’ve described might turn out to be less stable than I think.

The biggest uncertainty is Force 2. Will gamer hostility toward AI soften as the technology improves, the way hostility toward free-to-play and DLC softened? Or is this different — more durable, more values-driven, more permanent? If it softens meaningfully, Stable Point 3 becomes much larger and Stable Point 2 becomes much easier to maintain. If it doesn’t, the cost pressure on studios without a clear path to profitability gets worse every year.

I don’t have an answer. But I think the framework is useful for at least asking the right questions.

The new rhythm of my workday goes something like this: set Claude Code working on a feature spec, flip to a founder call, come back and review what it’s done, give it notes and set it writing tests, jump to an LP call, return to find the tests ready for review. Rinse, repeat. Meanwhile, Boardy is pinging my inbox and my WhatsApp with founder intros—some good ones, annoyingly—and Howie (an F4 Fund portfolio company) has quietly scheduled three meetings while I wasn’t looking.

It’s a different cadence than I’m used to. More context-switching, more plates spinning, more checking in on work that’s been happening while I was elsewhere. It demands a kind of multiplexed attention that would have felt scattered a year ago. Now it just feels like the job. And somehow, despite the constant juggling, I get a lot more done.

This is just... my life now? I don’t remember signing up for it. One day I was a normal person who used software; now I’m in what feels like a series of ongoing relationships with entities that remember things about me, have opinions, and occasionally make small talk.

I'm not a developer. But the pattern I'm describing—interleaving human attention with autonomous AI work—is becoming universal. And the most advanced version of it, the bleeding edge where people are figuring out what this all means, is happening in software development right now.

The orchestration era

On New Year’s Day, Steve Yegge dropped Gas Town: a framework for running 20-30 Claude Code instances simultaneously. Yegge describes it as “an industrialized coding factory manned by superintelligent chimpanzees” that “can wreck your shit in an instant” if you’re not an experienced chimp-wrangler.

Gas Town casts you, feckless human, as The Overseer in charge of an army of agents with seven distinct roles like The Mayor (your concierge and chief-of-staff) Polecats (ephemeral workers that swarm tasks), Refinery (manages the merge queue), and Witness (watches over polecats and unsticks them when they drift). Work flows through “convoys” that start, execute, and land without intervention.

What matters isn't speed—it's that the system keeps running without you. “Gas Town can work all night,” Yegge writes, “if you feed it enough work.” You design features, file implementation plans, sling work to your agents, then check back later. The system keeps running.

This pattern is going to go broader than software development. I think 2026 will be the year of Claude Code, and Nikunj Kothari adeptly frames the reason why: “Claude Code should be thought of as Claude Computer. Alien intelligence that has human tools—browser, filesystem, terminal commands—with generalized ability to do ANY task. Terminal UX is all that’s stopping folks from utilizing this all the time.”

When better interfaces arrive, orchestration will go mainstream: spinning up agents, directing their work, monitoring progress, intervening when they drift, reintegrating outputs. Parallel attention across autonomous systems.

What gamers already know

This skill set already exists. It’s been honed for decades by people who play certain kinds of video games.

Last year, the Financial Times published a piece about Factorio’s grip on Silicon Valley (subscription gated - here’s a snarkier but worse PC Gamer article about it).

Shopify CEO Tobi Lütke lets employees expense their copies. “Factorio is the one video game that everyone at Shopify can expense,” he told the Invest Like The Best podcast, “because it’s just bound to be good for Shopify if people play Factorio for a little while. We are building global supply chains, and Factorio makes a game out of that kind of thinking.”

But the better analogy for agent orchestration might be StarCraft. Lütke shouted it out in the same interview as “a good teaching tool for tech workers.”

In real-time strategy games, you’re constantly managing multiple simultaneous processes: building your economy, training units, scouting enemy positions, micromanaging battles, researching upgrades. Pro players maintain 300+ actions per minute, but raw speed isn’t the point. The skill is attention allocation—knowing when to check on your expansion, when to shift focus to your army, when a fight needs intervention versus when you trust your units to handle it.

This maps almost directly to multi-agent work:

Parallel attention management. RTS players maintain awareness of multiple processes while focusing on the most critical. You’re monitoring everything but selectively intervening. That’s exactly what you do when running several Claude instances—or, in Yegge’s telling, when you’re the Overseer managing Polecats, Crew, and the Refinery.

Systems that run without you. Factorio factories and StarCraft economies keep producing while you attend to other things. Gas Town convoys “start up, complete, and land without intervention.” The design goal is the same: build something that works autonomously, then check back.

Context-switching as the game. In deep work, context-switching is a productivity killer. In RTS, it is the game. Rapid shifts between micro and macro, offense and defense, multiple fronts—this is the rhythm of agent orchestration.

The UX gap

Gas Town works, but it’s built on tmux. Yegge is upfront about this: “You will have to learn a bit of tmux. Or, you can wait until someone writes a better UI for Gas Town.”

This is terrible UX for a novel interaction paradigm. Managing autonomous agents through terminal commands is like playing StarCraft by typing coordinates.

Game designers have spent decades solving exactly this problem: how do you let humans direct complex, multi-agent systems in real-time? How do you surface the right information at the right moment? How do you make something intricate feel intuitive?

The command interfaces in RTS games—control groups, minimaps, production queues, alert systems—are the result of thirty years of iteration on parallel attention management. StarCraft 2’s UI is one of the most refined solutions to human-agent coordination that exists.

Subscribe now

Someone is going to build the StarCraft UI for AI agents. An interface where you spin up workers, assign them to control groups, see their status on a dashboard, get alerts when they need attention, coordinate their work through something better than a shared markdown file.

The people most likely to build it well are game designers.

What this means

If you’re building AI tooling: the talent pool you need might not be where you’re looking. Game designers and RTS veterans understand parallel attention, autonomous systems, and interfaces that make complexity legible. They’ve been training for this.

If you’re a founder in this space, consider that games are more than entertainment - they’re decades of UX R&D on exactly the problems AI orchestration needs to solve.

The discourse about whether games need AI is backwards. AI needs what game designers know.

The reaction from much of the creative community has been predictable outrage. The Writers Guild called it a sanctioning of “theft.” Critics see it as another step toward the devaluation of human creativity.

I think they’re missing the point.

Two years ago, I wrote about how Five Nights at Freddy’s had pioneered a new model of IP development I called the “fanverse”—a permissive, community-driven approach that treated fan creation as an asset rather than a threat.

FNAF’s embrace of fan games, fan fiction, and derivative content didn’t dilute the brand. It supercharged it, culminating in the highest-grossing horror movie of 2023.

Disney just took that playbook and industrialized it.

The attention war is already lost

Here’s the uncomfortable truth that traditional media still hasn’t fully internalized: they cannot win the war for attention. It’s a zero-sum game, and they’re at a massive structural disadvantage.

As Doug Shapiro has pointed out, linear media is fighting with one hand tied behind its back. Streaming services, social platforms, and games are all competing for the same finite resource—human attention—and traditional media’s weapons (scheduled programming, theatrical windows, cable bundles) are increasingly obsolete.

GenAI is about to make this much, much worse. As the cost of content creation collapses toward zero, the volume of content will approach infinity. Every creator with a laptop becomes a studio. Every fan becomes a filmmaker. The supply of “content” is about to explode in ways that will make the current glut look quaint.

In this environment, putting out a movie every five years and hoping people remember to care is a losing strategy.

The battle Disney can win

But there’s another game entirely—one where traditional media actually has an advantage.

Deep fan engagement.

The value of Disney’s IP isn’t just the movies themselves. It’s the emotional connection that millions of people have with these characters. That connection is the moat. The question is how to deepen it, how to make it stickier, how to ensure that when someone thinks about heroic stories they think of Marvel, when they think about magic they think of Disney princesses.

The key is enabling fans to engage with IP across multiple touchpoints, formats, and modalities—continuously, not episodically. And the most capital-efficient way to do that? Outsource the work to the fans themselves.

This is exactly what the Disney-OpenAI deal enables. Instead of Disney spending hundreds of millions to produce content that fans consume passively, fans will spend their own time and energy creating content that deepens their emotional investment in Disney’s characters. Every Sora-generated video of Stitch doing something ridiculous is another touchpoint, another moment of engagement, another thread tying that fan more tightly to Disney’s universe.

This is not Netflix

The most common critique I’ve seen is that this deal is equivalent to Disney licensing content to Netflix—a strategic blunder that helped build a competitor. But this analogy is superficial at best.

Netflix was a substitute distribution layer. It used exclusive rights to Disney’s product to directly compete for attention and subscription dollars. Netflix took Disney’s content and used it to build Netflix’s audience, Netflix’s brand, Netflix’s business. When Netflix got strong enough, it started making its own content and became a direct competitor. Netflix eroded Disney’s moat.

OpenAI is something different. It’s a capability layer that enables derivative products. When fans create Sora videos featuring Disney characters, they’re not building OpenAI’s brand—they’re reinforcing Disney’s. The emotional connection, the nostalgia, the fandom: all of that accrues to Disney. OpenAI is just the tool.

If anything, this deal strengthens Disney’s moat by dramatically expanding the surface area of fan engagement while keeping the IP firmly in Disney’s control.

The risks are real

This doesn’t mean the deal is without risk. A few questions worth watching (h/t again - go subscribe to ):

How effective will the guardrails be? Disney and OpenAI have committed to “robust controls” to prevent harmful content. But we all know how that tends to go. Somewhere, someone is going to generate something deeply inappropriate with these characters, and it’s going to end up on social media. The question is how often, and how Disney responds.

What happens when fan content hits monetizable platforms? The deal covers Sora-generated content, but what happens when that content gets posted to YouTube or TikTok? Who owns the monetization rights? How do you prevent commercial exploitation? These are thorny questions that the announcement doesn’t fully address.

What does this mean for human creators? The WGA’s concerns aren’t baseless. If fans can generate decent-looking Star Wars content at home, does that reduce the perceived value of professionally-produced content? I suspect the answer is more nuanced than the critics fear—professional content and fan content serve different needs—but it’s a legitimate question.

Hollywood’s risk aversion problem

What strikes me most about this deal is how unusual it is for Disney—or any major studio—to move this aggressively on new technology.

Hollywood’s default response to technological change has always been resistance. The industry fought VCRs, fought DVDs, fought streaming, fought everything. And they were wrong every time. The technologies they fought ended up creating enormous new revenue streams, but because they fought instead of led, much of that value was captured by others.

Disney deserves credit for bucking this trend. Instead of waiting for the inevitable—fans using AI tools to create Disney content regardless of permission—they’ve chosen to get ahead of it. They’re setting the terms, building in guardrails, and positioning themselves to benefit from the shift rather than be victimized by it.

Is it a perfect deal? Probably not. But in a world where the alternative is watching your IP get remixed without your permission or participation, this seems like the smarter play.

The fanverse is coming for everyone. Disney just decided to own theirs.

If you’re building at the intersection of AI and entertainment, we’d love to hear from you.

But as I’ve talked to founders over the past few weeks, I’ve realized the matrix isn’t just a framework for processing your feelings about AI. It’s also a useful guide for thinking about building and investing in AI products.

Here’s the matrix again, for reference:

After a lot of conversations, I’ve come to believe there are two viable paths through this framework—and one anti-pattern that founders should avoid at all costs.

Path One: Relief to Excitement

The first path—and probably the more accessible one for most early-stage startups—is to start in the Zone of Relief and expand into the Zone of Excitement.

The Zone of Relief is where you’re solving a problem that people are capable of handling themselves, but really don’t want to. It’s work that doesn’t engage anyone’s uniquely human capacities. Nobody became a math teacher because they dreamed of spending their evenings marking worksheets. Nobody became a venture capitalist to play calendar Tetris.

Relief products have a few advantages as a wedge. They solve a real pain point that people will pay to eliminate. They integrate into existing workflows. They build trust through consistent, reliable execution of unglamorous tasks. And critically, they generate data.

That data is the bridge to the Zone of Excitement.

Consider an AI grading product for math teachers. The wedge is obvious: teachers can grade papers, but they’d rather not spend their evenings doing it. A product that handles this reliably creates immediate value.

But here’s where it gets interesting. Every graded assignment produces data—not just about what grade a student earned, but about what specific concepts they’re struggling with, what types of errors they make, and how their understanding is evolving over time.

A teacher couldn’t practically use this information to create individualized follow-up assignments for each student. There are only so many hours in the day. But an AI product can. It can take the output of grading and generate personalized homework that’s specifically designed to address each student’s gaps.

This is a Zone of Excitement expansion. Teachers want to provide personalized instruction—it’s closer to why many of them entered the profession in the first place—but they can’t do it at scale. The AI makes it possible.

The strategic logic is powerful: you’ve used a Relief product to build distribution, engagement, and data, then parlayed that into an Excitement product that creates deeper value and stickier relationships. The grading gets you in the door; the personalization keeps you there.

Path Two: Straight to Excitement

The second path is to skip Relief entirely and go straight for the Zone of Excitement. This is higher risk, but potentially faster and more explosive.

Suno is the clearest example. They didn’t start by handling some tedious adjacent task and expand from there. They just went straight to the thing: type a prompt, get a song. Something you wanted to do but couldn’t.

The numbers are staggering. Suno just raised $250 million at a $2.45 billion valuation. They’re generating $200 million in annual revenue. Users create 7 million songs per day—an entire Spotify catalog’s worth of music every two weeks.

Claude Code fits here too. It doesn’t handle the tedious parts of coding while I do the interesting bits. It unlocks work I simply couldn’t do before. The distance between “I wonder if I could build a tool that does X” and “I have a working prototype” has collapsed from months to hours.

The challenge with this path is that it’s harder to execute, and likely more capital intensive. You can’t start with a narrow use case and expand; you need to deliver something transformative from day one. The product has to be good enough that people immediately grasp what’s newly possible.

You also have fewer opportunities to build trust incrementally. With a Relief product, you can prove yourself on small stakes tasks before asking users to depend on you for bigger things. With an Excitement product, you’re asking for that trust upfront.

That said, when it works, it really works. Suno didn’t need to convince people that AI music generation was valuable—they just made it possible and people showed up in droves.

The Anti-Pattern: Dread as Provocation

There’s a third path that I’d advise founders to avoid entirely: using the Zone of Existential Dread as a deliberate provocation to get attention.

The canonical example is Cluely, the startup whose tagline at launch was literally “cheat on everything.” Their initial product was designed to help you cheat on interviews by feeding you answers in real-time—work that not only could you do yourself, but that specifically requires you to do it yourself for the whole exercise to have any meaning.

The Zone of Dread isn’t just about AI doing something you could do. It’s about AI doing something that violates fundamental social contracts. The entire point of a job interview is to create an accurate picture of a candidate’s capabilities. A product that short-circuits this isn’t just disrupting a workflow—it’s actively eroding trust in social institutions.

I’m not naive about the world. People cheat on interviews. Companies lie in job postings. The labor market is full of asymmetric information and principal-agent problems. But there’s a difference between acknowledging that reality and building a business that accelerates the erosion of whatever trust remains.

To their credit (or perhaps discredit), Cluely understood exactly what they were doing. The provocation was the point. Rage bait goes viral. A founder getting suspended from Columbia for using his own tool to fake interviews is a story people share. Within weeks of launch, they had 70,000 signups and $5.3 million in seed funding. A few months later, Andreessen Horowitz led a $15 million Series A.

Here’s the problem: they had “rage market fit,” not product-market fit.

Rage market fit is when people engage with your product because it makes them angry or fascinated by its audacity, not because it genuinely solves their problems. It’s a sugar high that feels like growth but doesn’t convert to lasting value.

The trajectory tells the story. After the initial virality faded and retention proved elusive, Cluely has tried to pivot toward something resembling a Zone of Relief product—an AI meeting assistant, a note-taker. But they’re doing it from a position of weakness. They’ve already told the world exactly who they are.

The deeper issue is trust. Zone of Relief products work precisely because users trust them to handle important-but-tedious tasks reliably. You earn that trust through consistent, unglamorous execution. You don’t earn it by announcing to the world that your company exists to help people deceive each other.

Implications for Founders and Investors

If you’re building an AI company, I’d encourage you to think carefully about where you’re starting and where you’re trying to go.

If you’re pursuing Path One (Relief → Excitement):

• Pick a wedge that’s genuinely painful and ubiquitous in your target market

• Design for data collection from day one—think about what information your Relief product will generate that could power an Excitement expansion

• Build trust through reliable execution before attempting the expansion

• Make sure the Excitement product is genuinely something users want to do but couldn’t, not just a feature addition

If you’re pursuing Path Two (Straight to Excitement):

• Be honest with yourself about whether you’re actually enabling something new or just making an existing task slightly easier

• Plan for the trust-building challenge—you don’t have the luxury of proving yourself on low-stakes tasks first

• Be prepared for a longer, harder road to initial traction

• Make sure the magic is immediately apparent

If you’re tempted by the anti-pattern:

• Don’t.

• Seriously, don’t.

• The attention is intoxicating but the business fundamentals are terrible

• You’re optimizing for a viral moment, not a company

From an investor’s perspective, I’ve become increasingly focused on the path from wedge to expansion. A Relief product that has no obvious route to Excitement is probably a feature, not a company. An Excitement product that can’t articulate why it’s genuinely enabling something new is probably riding a hype cycle.

And anything that starts in the Zone of Dread? That’s a pass. Life’s too short to invest in companies that make the world worse.

The robots are coming regardless. The question is whether we build them to extend human capability or to undermine human trust. I know which side I’m on.

This is not a metaphor. The new rhythm of my workday goes something like this: set Claude Code working on a feature spec, flip to a founder call, come back and review what it’s done, give it notes and set it writing tests, jump to an LP call, return to find the tests ready for review. Rinse, repeat. Meanwhile, Boardy is pinging my inbox and my WhatsApp with founder intros—some good ones, annoyingly—and Howie (an F4 Fund portfolio company) has quietly scheduled three meetings while I wasn’t looking.

It’s a different cadence than I’m used to. More context-switching, more plates spinning, more checking in on work that’s been happening while I was elsewhere. It demands a kind of multiplexed attention that would have felt scattered a year ago. Now it just feels like the job. And somehow, despite the constant juggling, I get a lot more done.

This is just... my life now? I don’t remember signing up for it. One day I was a normal person who used software; now I’m in what feels like a series of ongoing relationships with entities that remember things about me, have opinions, and occasionally make small talk.

Here’s the thing, though: I don’t feel the same way about all of them.

The matrix

When I actually sat down and thought about why some of these agents make me feel optimistic about the future and others give me low-grade existential angst, I realized there’s a pattern. It comes down to two questions:

1. Can I do this work myself?

2. Do I want to?

Plot any task on these two axes and you get four zones, each with its own emotional valence:

Let me walk you through them.

Zone of Relief

This is where Howie lives. Howie is my AI scheduling assistant, and it does something I am fully capable of doing but find soul-crushingly tedious: the back-and-forth calendar dance of finding a time that works for everyone.

Can I do this? Yes. Have I done this? For years. Do I want to do it? Absolutely not. Every minute I spend saying “Does 3pm PT work, or would Thursday be better?” is a minute I’m not doing literally anything else.

When Howie handles this for me, I feel nothing but gratitude. I did not become a venture capitalist to play calendar Tetris. The machine is welcome to it.

Zone of Excitement

Claude Code sits here. I am, charitably, a “somewhat technical person” - which is to say I understand some technology reasonably well and have lots of ideas for things I’d like to build. What I could not previously do (at least, not alone) is actually build them.

Claude Code changes this. The distance between “I wonder if I could make a tool that does X” and “I have a working prototype” has collapsed from months to hours. It’s not doing work I could do; it’s unlocking work I couldn’t.

This is where AI feels like a genuine expansion of human capability—a bicycle for the mind, in Jobs’ old framing. More reach, more leverage, more of the possible.

Zone of Indifference

I don’t have a great example here, because by definition I don’t think about this zone much. It’s work I can’t do and don’t want to do anyway. Maybe there’s an AI somewhere optimizing semiconductor fab layouts. Good for it. I have no particular feelings about this one way or the other.

Zone of Existential Dread

And then there’s Boardy.

Longtime readers will know I’ve been mildly obsessed with Boardy, an AI that talks to you about your work and interests and then introduces you to people it thinks you should know, for a while now. I first wrote about it earlier this year when it still felt like a curiosity—an uncanny but intriguing experiment in AI-native networking.

Since then, Boardy has honed in on a use case that hits closer to home. It now spends most of its time talking to founders and investors and introducing them to each other. It’s launching Boardy Ventures. It has a scout program.

Do you see where this is going?

A nontrivial part of my job is talking to founders, understanding and evaluating what they’re building, figuring out what they need, and connecting them with people who can help. I like to think I’m reasonably good at this. I would also like to think there’s something ineffable about the human judgment involved—pattern recognition honed over years, a sense for chemistry that can’t be reduced to cosine similarity.

This may be bollocks.

Boardy can talk to a thousand founders at once. It never sleeps. It never forgets a detail. The thing I do that I want to do, that I think is core to who I am professionally—Boardy is doing it, at scale, right now.

This is the zone of existential dread.

And yet.

When a founder calls me at midnight because their cofounder just quit and they don’t know what to do, Boardy could technically take that call. But would the founder want to talk to it? I don’t think so. Not yet, anyway. There’s something about crisis moments—about the texture of human presence, the weight of shared stakes—that an AI can’t provide. At least not in a way that feels real.

More fundamentally, I’m not sure Boardy has a model of the world in the way that matters. Ilya Sutskever, one of the architects of the current AI paradigm, said recently that we’ve moved from the “age of scaling” to the “age of research”—that simply making models bigger won’t get us where we need to go. The core problem, he argues, is generalization: “These models somehow just generalize dramatically worse than people.”

He calls the phenomenon “jaggedness.” A model can pass the hardest exams in the world and then get stuck in a loop where it introduces a bug, you point it out, it apologizes profusely and fixes it by reintroducing the previous bug, and you cycle between the two forever. Something strange is going on under the hood.

For matching founders with investors based on stated interests and background similarity, Boardy is probably already better than me. But for the judgment calls that actually matter—should this founder raise right now, is this investor actually a good fit for how this person operates, is there something off here that I can’t quite articulate—I’m not convinced the current architecture can get there. The model lacks the lived experience, the emotional intuition, the sense of how the world actually works that humans accumulate over decades of navigating it.

Maybe that changes. Probably it changes, eventually. But for now, I take some comfort in the gap.

Where you sit depends on where you stand

The agentic dread matrix isn’t really about the agents. It’s about you.

How you feel about AI is tightly correlated with how much of your livelihood overlaps with each zone. If most of your job is work you can do but don’t want to, AI feels like liberation. If it’s work you can’t do but wish you could, AI feels like a superpower.

But if you’ve built an identity around work you can do and want to do—if you’ve spent years getting good at something you genuinely care about—watching an AI casually replicate it produces a very specific kind of vertigo.

I don’t think the answer is to pretend the zone of dread doesn’t exist, or to convince yourself that your particular brand of human judgment is somehow immune. It isn’t. Mine isn’t.

A more useful question: What aspects of this work is AI genuinely unlikely to do well, at least for the foreseeable future? And how do I collaborate with it to extend the frontier of what’s possible, rather than defending territory that’s already lost?

I have some ideas about this—more in a later post. For now, I’ll just say: the robots are here, they’re in my calendar and my inbox and my code reviews, and I expect they will soon be everywhere else.

The zone of dread exists, and it’s my job to find a way out. My answer is to build new tools that amplify what I can do. If the machines are going to multiply, so am I.

The first kind is the hard that comes from building something new. You’re figuring out the product. You’re learning your customers. You’re refining your positioning. It’s hard because you’re doing something that hasn’t been done before, and the path isn’t clear yet. But beneath the difficulty, you can feel something pulling you forward. Customers are leaning in. Early users are coming back. The market is giving you signals that you’re onto something.

The second kind of hard is different. You’re working just as intensely, maybe more so. You’re shipping features, iterating on feedback, optimizing your funnel. You’re executing well by every reasonable metric. But it feels like pushing a boulder uphill. Every win is a grind. Every customer is a fight. You tell yourself this is just what building a company feels like, and you double down on execution.

The trap is this: if you’re good enough, you can make real progress even in the second scenario. And that progress can keep you blind to the fundamental problem.

The difference you can feel

I remember the last time I launched a game that had real product-market fit. From the moment it went live, something was different. Users weren’t just signing up—they were pulling us forward. The retention curves looked right. The referrals happened organically. It wasn’t easy, but it felt easy. Like we were running downhill instead of climbing.

I’ve also been on the other side. Working incredibly hard, making incremental progress, telling myself that this is what it takes. And it is what it takes—if you don’t have strong market pull, the only way to grow is to outwork everyone else. You can do this for years. You can build a real business this way, even.

But you’ll never know if you could have built something 10x bigger if you’d been willing to step back and ask whether you were grinding in the right direction.

The relentless founder’s blind spot

The strongest founders I know are relentless. They don’t give up. They don’t get discouraged. They see obstacles as problems to solve, not reasons to quit.

This is exactly what makes them successful. It’s also what makes them vulnerable to a specific kind of failure.

When you’re relentless, you can convince yourself that any problem is solvable through better execution. Conversion rate too low? Optimize the funnel. Customers churning? Improve the product. No inbound interest? Build better content, run better campaigns, automate your outreach.

All of these things might be the right answer. Or they might be sophisticated ways to avoid confronting a harder truth: you’re executing brilliantly in a market that doesn’t want what you’re offering badly enough.

Signal vs. noise

Here’s what makes this particularly insidious: good execution in a tough market produces results. You’ll get some customers. You’ll generate some revenue. You’ll hit some milestones. These feel like validation.

But there’s a difference between “we closed five customers this quarter through sheer force of will” and “we closed five customers and had to turn away ten others because we’re not ready yet.”

The first is a warning sign. The second is product-market fit.

When I see founders in brutally competitive spaces—industries with infinite noise, where everyone’s inbox is already flooded, where the tenth-best solution might work fine—I worry. Not because they can’t win, but because winning requires perfect execution just to achieve mediocrity. Meanwhile, a slightly different approach in an adjacent space might deliver 10x the results with the same effort.

How to know if you’re in the trap

Ask yourself these questions honestly:

Are you getting wins because you’re working hard, or because the market is pulling your product forward? If every customer feels like a conquest rather than a conversion, pay attention.

Would a less capable founder fail immediately in your market, or slowly? If the answer is “slowly,” that might mean the market has some baseline demand. If it’s “immediately,” you might be the only thing keeping this alive.

Are you seeing surprising success, or exactly the level of success you’d expect from your effort? Product-market fit often shows up as results that feel disproportionate to input. If everything tracks linearly to your hustle, something might be off.

When you talk to potential customers, do they lean in or nod politely? The difference between “this is interesting” and “holy shit, when can I have this?” is the difference between a feature and a must-have.

What to do about it

This isn’t about giving up at the first sign of difficulty. Building anything worthwhile is hard. But there’s a difference between the hard work of creation and the hard work of pushing water uphill.

If you’re in the trap, you have options:

Validate before you pivot. Talk to people who know your space cold. Get real backchannel feedback from operators who’ve been there. Sometimes what feels like a grind is actually just the early stage of something real, and you need perspective to tell the difference.

Look for adjacent spaces with less noise. If you’re building outbound automation in a world where everyone else is also building outbound automation, maybe there’s a fundamentally different approach to the same problem. Maybe the insight isn’t “better outbound” but “what comes after outbound?”

Give yourself permission to explore. You don’t have to abandon ship to test other waters. Strong founders can run multiple experiments. See what gets traction. See what feels different.

Trust the feeling. When you have real product-market fit, you know. Customers pull you forward. Growth feels inevitable rather than manufactured. If you’ve never felt that, you might not recognize it. If you have felt it before and don’t feel it now, pay attention to that signal.

The hardest part about being relentless is knowing when to stop being relentless about the current path and become relentless about finding a better one.

The courage to question

I don’t want to discourage anyone from pushing through hard things. Some of the best companies were built by founders who refused to quit when everyone else would have.

But I’ve also seen brilliant founders spend years executing perfectly in the wrong direction. Not because they weren’t smart enough or didn’t work hard enough, but because their greatest strength—their unwillingness to give up—prevented them from asking whether they should be fighting a different battle.

The question isn’t whether you can succeed through pure force of will. If you’re good enough, you probably can.

The question is whether you should have to.

Western studios are losing their edge

Five years ago, every major Asian publisher was writing checks to Western studios. American and European developers were the hot commodity - perceived as more experienced, more sophisticated, capable of creating “AAA quality” content that could compete globally.

That era is over.

Now, Asian publishers view the West primarily as a market for their domestic content, not as a source of development talent. And honestly? They’re right to. Most Western studios simply aren’t cost competitive anymore, while it’s become increasingly clear that Asian developers can create content with massive global appeal.

The numbers are shocking. I heard about one C-level executive at a venture-backed US studio pulling down $3M in annual salary. Not equity - cash. Meanwhile, I saw a demo from a team of veteran Japanese developers who had worked at major AAA studios, and they need a total of $4M to reach Early Access. The entire budget for their game - which looks better than most $50M productions - is less than what one Western executive is making in salary.

This isn’t sustainable. And the market is figuring it out.

The one exception: AI-native studios

There’s exactly one area where Western studios still command premium interest from Asian investors: AI-native game development.

The US is still seen as the leader here, primarily because of proximity to OpenAI, and the other frontier model companies. I had multiple conversations with investors and publishers who are genuinely curious about studios at the bleeding edge of exploring AI-native gameplay - not just using AI as a production tool, but building games where AI is fundamental to the core experience.

This makes sense. The infrastructure, talent, and experimentation happening around AI is still concentrated in the US. But it’s a narrow window of advantage, and it won’t last forever.

Anime has won

The global takeover of anime is accelerating faster than most people in the West realize. The US writers’ strike created massive content holes that international IP rushed to fill. Global platforms like Netflix have brought anime into living rooms everywhere, and it’s found huge traction, especially with younger audiences.

But here’s what’s really important: anime isn’t just something people watch anymore - it’s foundational IP spinning off into live action adaptations, gaming, merchandise, everything. The entire entertainment stack is being rebuilt around these properties.

If you’re a Western studio or investor and you’re not thinking about how anime-style content and IP fits into your strategy, you’re already behind.

AI adoption without the angst

One of the most striking contrasts I noticed: prominent anime creatives are using AI production tools enthusiastically and openly, without the existential fear and political baggage I see in the US. There’s much less worry about audience rejection or being “canceled” for using AI.

My theory on why this is: AI is less politically and culturally charged in Japan because the billionaires aren’t as visible and inequality isn’t as stark. When you don’t have a Zuckerberg or an Altman or a Musk dominating the cultural conversation and personifying inequality, the technology itself doesn’t become as loaded with political meaning.

It’s just a tool. And people are using it to make things.

Microsoft’s handheld gambit

Microsoft had giant billboards all over TGS showing their new ROG-branded Xbox handheld with one simple message: “THIS IS AN XBOX.”

It’s a clear statement of intent - the Xbox platform is no longer defined by a box under your TV. Cloud, mobile, PC, handheld - it’s all Xbox now.

The device itself looks compelling. But here’s the question that matters: in a world where Sony dominates Japan, Nintendo owns portable gaming, and Steam Deck has won over PC gamers, who exactly is this for?

And more importantly, at a premium price point, will the bet that “Xbox everywhere” is more valuable than “Xbox as a place” actually pay off?

When your platform is every screen, you control none of them. That’s either the future of gaming, or it’s a strategic retreat dressed up as innovation. We’ll find out which.

The bottom line: Power in the games industry is shifting East, cost structures in the West are broken, and AI is being adopted much faster in markets where it hasn’t become a culture war issue. The studios and investors who understand these shifts will thrive. Everyone else is in for a rough few years.

If you’re working on something that reflects these new realities, I’d love to hear from you.

Elden Ring sold 20 million copies by telling most gamers they weren't good enough to play it.

No difficulty slider. No hand-holding tutorials. Few concessions to "accessibility." Just a brutally difficult game that actively punishes you for not paying attention. The kind of game that makes you want to throw your controller through the screen.

And yet it became 2022's biggest hit.

This shouldn't work. Every product manager's instinct, every MBA framework, every venture partner's pattern recognition says you should reduce friction, broaden your funnel, make things easier. Sand down the rough edges. Maximize your TAM.

But when you try to build something for everyone, you build something for no one.

The art of the trade-off

When you finally beat a boss after 50 straight deaths, something magical happens. The frustration of those 49 failures doesn't just disappear - it transforms into one of the most powerful feelings in gaming: earned mastery. That moment of triumph feels so good precisely because you paid for it in blood.

This is the trade-off at the heart of great product design. By deliberately excluding players who want a casual experience, FromSoftware can deliver peak emotional moments that are impossible in more accessible games. You can't have the high without the low.

A game with a difficulty slider can never deliver that feeling. Not because it's technically inferior, but because when victory is guaranteed, it's meaningless.

The Marmite principle

There's a British spread called Marmite that’s made from yeast extract. It overwhelms your tastebuds with salt and umami and I love it. My wife finds it revolting.

Their tagline isn't "Pretty good on toast." It's "Love it or hate it."

They literally advertise that half their potential customers will find their product disgusting. And they've been doing it successfully for over a century.

This is product-market fit at its purest: build something so specific that the people who get it can't imagine living without it, even if that means everyone else thinks you're insane.

Build for believers, not browsers

The best founders I know are all making brutal trade-offs that create enemies.

Linear built issue tracking for people who care more about velocity than customization. No custom fields. No complex workflow configuration. Just ruthlessly fast, keyboard-first, and deliberately constrained. The Jira power users who need seventeen custom statuses? Hate it. Meanwhile, Linear's users ship faster than everyone else because the tool refuses to let them overthink their process.

Owner tells restaurant owners they can't customize their own websites. Want to move that button? Too bad. Want a different font? Nope. They've tested these layouts on thousands of restaurants and know what converts.

Superposition (one of our portfolio companies) is building AI-powered recruiting software that encodes founder Edmund Cuthbert's specific, battle-tested approach to hiring. After seven years as a recruiter, he knows what works. Traditional recruiters who want infinite customization might rage against it. Founders who just want to hire great engineers fast become evangelists.

Notice the pattern? Each of these products makes an explicit trade: we will serve this specific group extraordinarily well, and we accept that means actively repelling everyone else.

Why polarization pays

In 2024, over 19,000 games launched on Steam. The year before? 14,000. The year before that? 12,000.

The same tsunami is hitting every category. There are 200 project management tools. 500 note-taking apps. A thousand AI writing assistants.

When there's infinite choice, being "pretty good" is a death sentence.

The median game on Steam sells 500 copies. The median SaaS product never breaks $10K MRR. They die not because they're bad, but because they're forgettable. They tried so hard not to offend anyone that they failed to delight anyone.

Meanwhile, products with sharp edges - products that actively repel certain users -create something more valuable than broad appeal: they create identity.

When you beat Malenia after dying 100 times, you don't just feel satisfaction. You feel transformed. You ARE an Elden Ring player now. And identities are viral - we can't help but signal them to others.

FromSoft doesn't make games for everyone. They make games for people who get pleasure from mastery that is forged through suffering. That's maybe 5% of the total market. But that 5% will buy every single game they make, at full price, on day one, tell everyone they know about it, and spend hundreds of hours creating content about it.

5% of people who absolutely love you is worth more than 50% who think you're fine.

The founder's playbook for polarization

Here's how to do this:

1. Define your trade-offs explicitly: Stop asking "how can we serve more users?" Start asking "what are we willing to sacrifice to serve our core users better?" Write these trade-offs down. Make them public. Let them repel the wrong people.

2. Make your product a filter: Your product should actively reject people who aren't right for it. Raya requires an application and approval from a committee. Most people get rejected. This exclusivity is part of the value proposition.

3. Turn your constraints into your marketing: When someone complains that your product is "too hard" or "too opinionated" or "too expensive," screenshot it. Your true fans will rally to defend you, creating more authentic energy than any ad campaign.

4. Measure depth, not width: Stop looking at your total user count. Start measuring: How many people would be genuinely devastated if your product disappeared tomorrow? If that number isn't growing, you're trying to please too many people.

5. Double down on what makes you weird: Whatever makes certain people uncomfortable about your product - that's probably your biggest asset. Don't sand it down. Sharpen it.

The courage to exclude

Every founder's instinct is to be loved. We want everyone to use our product. We want universal validation.

But in a world of infinite abundance, the scarcest resource isn't users or revenue or even attention. It's giving a shit.

People only care about products that stand for something. And you can't stand for something without standing against something else.

So make your trade-offs. Pick your tribe. Build something that your competitors' users will hate, because it means your users will defend it with their lives.

The question isn't whether your product is good enough. The question is whether it's specific enough.

Because being hated by the right people is a sign you're onto something great.

Are you building something worth hating? We should talk.

Some dude with a Roman statue avatar posts "CLIMATE CHANGE IS A HOAX." Ten thousand retweets.

The platforms know what they're doing. Engagement equals revenue, outrage drives engagement, therefore outrage drives revenue. They dress it up with words like "meaningful social interactions" but we know what really moves the needle.

There’s a version of the future where AI makes this worse - a diabolically optimized slop cannon firing at the human face, forever - but I think there are some reasons to be optimistic.

Not by detecting misinformation. By being a better listener than any of us.

I've been testing new AI products for the fund. Started skeptical, especially after the ChatGPT wrapper gold rush of 2023. But something's different about this wave.

Take Boardy. Twenty minutes on the phone with an AI about what I actually do, what kind of people give me energy, what conversations I want to have. I found myself being honest - telling it what I believe and why.

The AI didn't judge. It asked follow-ups. When I contradicted myself, it gently probed instead of calling me out. A week later it introduced me to someone, and it was an interesting conversation. Next time it reaches out, I’ll take the call.

Known throws out the swipe in favor of a conversation with an AI matchmaker. As soon as I started the call, it dove right into asking me about where I grew up and how that shaped me as a person - the antithesis of what shows up on a Tinder bio.

Superposition (disclosure: F4 Fund portfolio company) is doing this for technical recruiting. Instead of those insane job requisitions that demand 10 years experience in frameworks that have existed for 3, founders just talk about their company. The AI figures out what kind of person would thrive there and goes out to find them.

What makes this work? Infinite algorithmic patience.

Humans are terrible listeners. We wait for our turn to talk, pattern match to our own experience, finish people's sentences. My wife has pointed this out roughly one thousand times. She's right every time.

AI doesn't get tired. Doesn't need to be interesting. Has no ego to defend.

This all sounds great, but listening is the easy part.

What matters is what comes after - the mysterious alchemy that separates great human networkers from the people who just introduce everyone they know to everyone else they know.

I know a headhunter who's placed half the C-suite in gaming. She doesn't just listen to what companies say they want. She reads between the lines and picks up on the stuff they don't know to ask for.

Can AI do this? Can it actually develop taste and judgment? The jury is still out.

A great matchmaker once told me she ignores what men say they want in a partner and listens to how they describe their exes. The woman they describe with the most energy - even if it's negative energy - tells you their actual type. That's the kind of pattern recognition that matters.

These new systems are betting they can find those patterns. Not just "you both like hiking" but "you both get animated discussing the same kinds of problems" or "you have complementary blind spots" or whatever subtle signal actually predicts connection.

Traditional platforms are attention routers - moving eyeballs around, maximizing surface area. These companies are trying to build understanding engines. Whether they succeed depends on if AI can develop something resembling taste.

Will AI become as good as the best human matchmakers? Probably not. But it might not need to be. It just needs to be better than swiping through photos or broadcasting into the void. That's a pretty low bar.

The question isn't whether machines can listen better than we do - they already can. It's whether they can develop the judgment to know what to do with what they hear.

If they can, we might finally have an alternative to the slop cannon. If they can't, at least we'll have had a few decent conversations along the way.

But one detail from today’s announcement of their acquisition by Grammarly jumped out at me: after a decade of building and raising over $100M (most recently at a near-unicorn valuation), Superhuman had "tens of thousands" of users.

There’s a story here about venture math, distribution, and timing that every founder should understand.

Luxury product meets venture math

Superhuman has always been something of a luxury product, charging ~$30/month for something most people get for free. Even with tens of thousands of users, that's a decent business - $35M in annual revenue at time of acquisition, although perhaps only marginally profitable with ~200 mostly US-based employees.

Here's the problem: that's just not the scale of growth that gets mega-fund VCs excited, especially at a time when the fastest growing AI companies are measuring “time to $100M ARR” in months, not years.

Superhuman is one of many startups that raised at rich valuations during ZIRP, which in hindsight may have reduced their optionality. But there’s also perhaps something to be said about product vs distribution.

Distribution always wins

The real story here is the user count disparity: Grammarly's tens of millions versus Superhuman's tens of thousands.

Distribution trumps product. This has always been true, but it's going to become even more true in the age of AI, where features can be duplicated with alarming speed. A great AI writing assistant? Every email client will have that within 12 months. Keyboard shortcuts and workflow optimization? Table stakes.

What can't be easily duplicated is access to millions of users who trust your brand and use your product daily. Grammarly has that. Superhuman, for all its product excellence, remained a high margin but niche product.

AI-native innovator’s dilemma

Superhuman also faces the classic innovator's dilemma as AI reshapes email. They've built an incredible product, but it's still fundamentally an email client rooted in pre-AI UX paradigms.

Look at their design choices: many of Superhuman's keyboard shortcuts deliberately mirror Gmail's. That was smart when competing with Gmail, but it's baggage when competing with AI-native clients like Net that can start fresh.

Superhuman’s value to Grammarly

Grammarly already has desktop and mobile writing apps - they're not just a browser plugin anymore. But email is where working professionals actually live, spending multiple hours of their workday reading, writing, and managing communications.

Grammarly wants to be the AI-native productivity suite that helps you communicate across every medium. Email is a huge part of that puzzle, and Superhuman gives them a premium entry point.

The more customer touchpoints you own, the more context you can gather. The more context you have, the better your AI becomes. And the better your AI becomes, the stronger your personalization moat.

What this means for founders

A few takeaways:

Venture math is unforgiving. If you take big checks at big valuations, you're committing to building a business that can support those numbers. There's no shame in the boutique business model, but be honest about what you're signing up for.

Distribution beats everything. In a world where AI can replicate features quickly, your sustainable advantage comes from access to users, not cleverness of implementation. Plan your distribution strategy before you plan your product roadmap.

Context is the new moat. In an AI world, sustainable advantage comes from proprietary data loops and user touchpoints, not feature differentiation.

Superhuman's exit is well-timed. They built a good business and found the right acquirer at the right moment. But it's also a reminder that in venture-backed startups, being good enough isn't enough. You need to be massively scalable, or you need to find someone who is.

The Mediator

Trust is the New Oil

A few weeks ago, I floated a half-baked idea during a presentation: “in the age of AI, trust is the new oil.” The more I bake it, the more I believe this phrase captures the possibility that AI will re-order the media economy as we know it. In many ways, it also points to a hopeful future, even for traditional media…

Read more

10 months ago · 12 likes · Doug Shapiro

We're witnessing a fundamental shift from "whoever captures the most attention wins" to "whoever builds the most trust wins."

Here's what most people miss: attention is finite and zero-sum. Your time scrolling TikTok can't also be spent reading newsletters. But trust works differently. It's infinite and collaborative. Me trusting one creator doesn't reduce the trust I have available for another.

The implications are significant. Human curation becomes premium infrastructure. Long-term creator relationships start outvaluing viral moments. Community-driven IP becomes the strongest possible moat. Companies with direct customer relationships beat those dependent on algorithmic distribution every time.

The winners will be the companies and creators who've spent years building genuine relationships. Think about the "boring" newsletter writer with 10K subscribers who get genuinely excited about every email. Or the game studio that actually listens to community feedback and iterates based on it.

The attention economy isn't going anywhere, but sharecropping on the attention platforms is going to become harder and less valuable with every passing year.

In a world of infinite content, finite trust becomes the ultimate currency.

You should listen to the whole thing, but here's a summary:

1) The fundamentals were broken:

• Game studios raising $50-300M without clear competitive advantages

• Burn rates of $3-4M monthly ($40-50M annually) PRE-LAUNCH

• No capital reserved for go-to-market or post-launch iteration

• Founders optimizing for perfect games, not viable businesses

2. Recent failures tell the story:

• Mountaintop: $86M raised, shipped one season, shuttered

• Singularity 6: $50M+ raised, acquired by Daybreak for pennies on the dollar

• Elodie: Raised at $90M post, then did a round at $9M post (selling 44% for just $4M)

• That's No Moon: $110M raised, followed by a desperate $10M extension

• Build a Rocket Boy: $300M+ raised, nothing shipped yet

3) The "vertical slice seed round" model sets companies up for failure in a normal capital environment:

• Series A/B funding typically requires market validation

• Games need massive capital pre-launch ("the naked Series B problem")

• Too many game founders think like developers, not company builders

4) The cold, hard capital efficiency reality:

• Successful companies return capital to VCs. Period.

• The "Riot diaspora" fallacy: People from Riot/Blizzard aren't automatically successful

• Discord pivoted from a failed game with dry powder and succeeded

• Indie successes often do more with $1-5M than VC-backed studios do with $50M+

5) What actually works in games venture:

• Durable competitive advantages (distribution, technology shift, network effects)

• Founders who understand capital as a strategic weapon, not just fuel for production

• Products with clear answers to "why would someone play this?"

• Teams that can articulate their business, not just their game mechanics

Let's be honest: We're witnessing a cataclysmic reset in games venture. The companies in the "deadpool" - those at risk of imminent failure - include some of the most heavily funded studios of the past five years.

The bar for raising venture in games is now astronomical. It should be. When studios burn through $100M+ with nothing to show for it, something is fundamentally broken in how we're building these companies.

A final observation: venture capital works best when it exploits a technology shift. Most of these companies were funded in the absence of one (mobile was the last), which left them competing in a red ocean with no tailwind.

Let’s start with the obvious: compared with building a company, this is an easy job. As much as the loudest among us love to valorize the profession, VCs are not in the arena–we are cheering from the sidelines.

An easy job, then, but a hard one to do well. By definition, only a minority make it to the top quartile in which the asset class starts to justify its duration and illiquidity, and it’s only in the top decile where reputations are really made.

Does founder experience make you a better investor? The data are mixed, but I think there are a few traps that former founders are much more likely to fall into. This is an essay about those traps, in the hope that writing about them will make me less likely to blunder into them in future.

Falling in love with the idea, not the founder

Occasionally, someone pitches you an idea that you just fall in love with. Maybe it addresses a longstanding problem you wish someone would address, or it perfectly fits a thesis you’ve been developing.

Regardless, this is a trap. Every time I’ve fallen more deeply in love with what a founder is building than the founder who is building it, it has generally worked out worse than cases where the opposite was true–when I’ve felt that the idea needed work, but the founder was absolutely top-tier.

That this should be the case is not surprising. Ideas are vital, but cheap and easily changed, while success rests so much on the quality and velocity of founder decision making and execution that of course that’s what matters.

I think ex-founders are especially susceptible to making this mistake because the ability to get irrationally excited about an idea is so completely ingrained that it’s hard to shake. Founders need to do this. It’s what allows you to sell a ludicrously improbable proposition to investors, to employees, to yourself.

This superpower is a liability as an investor, because it predisposes you to imagine what the best entrepreneurial version of yourself would do if you were running the company rather than pay close attention to what the founder is actually doing.

My operating framework is now much simpler: focus on finding the strongest founders who are what my friend Charles Hudson calls “in the neighborhood of a good idea”, and support them as best I can.

Bias to action

Startups are a high APM game. Action generates heat and light: you do things, and those things generate information and outcomes that help you figure out what to do next. The best founders are extremely high agency and move unreasonably fast, because they understand that doing so buys you more times to run the OODA loop before you’re out of runway.

These instincts are incredibly valuable for a founder. For investors, a bias to inaction serves us better most of the time.1 Founders have to make a lot of decisions, but most of those decisions are relatively low stakes and reversible. You can roll back a bad feature, or part ways with a hire that’s not working out.

Investing, by contrast, is a fundamentally low APM activity. The job more closely resembles what Jeff Bezos describes as the role of a senior executive: you get paid to make a small number of high-quality decisions. What’s worse, you make the most consequential and irreversible decision (whether to invest) first!

More time usually means more information about the founder, the company and its general trajectory. Bad VCs are notorious for substituting social proof for their own judgment, but that’s not always what’s going on–the truth is, you can often learn more by waiting than by moving quickly.

Since I learned to channel my bias to action into the work of building our firm rather than our investing, the quality of our decision making has gotten better.

Moving from benevolent despot to coach

As a founder, everything is your fault. The corollary to this is that you can and should do and say whatever is necessary to fix what is broken, and quickly. In startups, there is no time to get buy in from key stakeholders. Sometimes you have to make the decision, dictate the choice and live with the consequences.

I’ve never served on the board of a company I didn’t start, so perhaps things are different for VCs who hold board seats, but from my vantage point at our little fund, we are nowhere close to being in charge of anything once we’ve written a check.

This is a hard dynamic to adapt to after a couple of decades doing everything your way. While some founders thrive on direct feedback, very few like being told what to do (nor should they!). It’s their company and they have more context than you.

I have found being a coach instead of the boss to be a difficult transition. The only thing that seems to have helped is getting much better at asking the questions that lead to what you hope is the right conclusion.

Ultimately though, investors are involved, but founders are committed. The best VCs I know recognize this fundamental asymmetry and approach their role with appropriate humility. Two years in, I'm still unlearning founder habits and rebuilding these muscles. If the sidelines have taught me anything, it's that knowing what play to call is far easier than having the wisdom to keep quiet.

“Technology changes, but people don’t.”
― David von Drehle, The Book of Charlie

One way to think about the history of technology is as a series of new solutions to the same old human problems, needs and desires.

While technology does change us in ways both good and bad, our fundamental needs—for food, shelter, love, belonging and self-actualization—have remained constant for thousands of years.

What does change are the ways technology improves our ability to meet those needs, and the companies that figure them out at scale sometimes end up creating billions or trillions of dollars of value.

How does this process of distilling new technology into breakout products happen? I find it helpful to think in terms of primitives.

Breakthrough products are built from new primitives

Every major technology shift introduces new primitives: foundational capabilities that serve as the building blocks for new products.

Take mobile. The introduction of the iPhone and its imitators unleashed four major primitives:

• Multi-touch screens, enabling intuitive, gesture-driven interactions.

• Ubiquitous connectivity, ensuring continuous internet access wherever you are.

• Location awareness (GPS), making precise real-time positioning universally available.

• High-quality cameras in everyone’s pocket.

The most transformative products of the mobile era weren't desktop applications adapted to smaller screens—they were novel experiences that could only have been built using these new primitives.

Consider three iconic mobile products:

• Tinder used touch (swipe gestures), location awareness, and camera integration to reinvent dating. Before mobile, online dating meant browsing static databases on desktop computers. The swipe-based UI and geolocation created an entirely new model of casual, immediate, and localized interactions.

• Uber used ubiquitous connectivity and precise GPS to build a real-time transportation marketplace that couldn't have worked on desktops or laptops. Instantly summoning a ride, tracking its arrival in real-time, and frictionlessly paying upon arrival transformed personal transportation.

• Pokémon GO combined GPS, camera-based augmented reality, and ubiquitous connectivity to pioneer mainstream location-based gaming, creating a global cultural phenomenon.

Each of these examples relied on multiple primitives working together to create something genuinely new.

What are AI’s primitives?

But what exactly are the primitives of AI?

Unlike mobile, whose primitives are easy to intuit from its hardware, AI feels more abstract—it’s not a device so much as an emergent yet inscrutable machine god.

Still, let’s give it a shot:

• Content Plasticity: AI enables effortless transformation of content between different modalities—text, image, video, voice, or code—at near-zero marginal cost. This dramatically reduces the barriers to content creation, remixing, and adaptation.

• Agentic Abstraction: AI systems can abstract away complex multi-step processes behind simple, high-level user intentions. Users specify their goals, and the AI handles the messy execution details transparently.

• Conversational UI: Just as touchscreens became the default interaction model for mobile, natural language (spoken or typed) is emerging as the universal interface for AI-native apps. This unlocks products that are simultaneously simpler yet more powerful—an extremely rare combination.

• Deep Personalization: Traditional personalization mainly determined what content was shown to users - which tiles Netflix shows you, what songs and playlists Spotify suggests to you. AI enables generative personalization—customizing not just the content, but also how software interacts, communicates, and adapts its interface based on deep understanding of individual preferences, behaviors, and context. For the first time, we have software that is capable of displaying (or at least faking) personality and empathy.

AI-native products

I suspect the most transformative AI-native consumer products are yet to be built. We're still early, and entrepreneurs are just starting to experiment with these primitives to discover new categories.

One product I think about a lot is Boardy. Boardy aims to make valuable connections for you, much like a trusted friend or colleague. It does this by talking to you on the phone about your interests, your work and what kind of people you’d like to meet.

Talking to Boardy is an uncanny but surprisingly enjoyable experience. Despite the occasional telltale lengthy pause (something he tells you to anticipate up front in a canny bit of UX design), I found myself quickly opening up to him (I say him because he sounds like a friendly Aussie).

Here’s our last conversation.

Will Boardy deliver on its promise of a deeply personalized social network? I’ll find out when he reaches out with my first intro, but I’m intrigued and most importantly, interacting with it feels like something genuinely new and native to the AI era.

Conclusion

I suspect the founders that create the next category-defining companies will be those who find creative ways to combine these emerging primitives to address timeless human needs. For all our technological advancement, we still seek connection, understanding, creative expression, and mastery over our environment.

For startups, this represents an extraordinary opportunity, because technological platform shifts can weaken moats and reshuffle competitive advantages. While some incumbents like Facebook successfully navigated the transition from desktop to mobile, many once-dominant companies failed to adapt. And entirely new categories emerged that enabled companies like Uber to become giants.

The most exciting question isn't which primitives will define the AI era, but rather: which previously impossible solutions to age-old human problems will they finally make possible? And who will be bold enough to build them?

If you are, I’d love to hear from you.

Thanks to Ryan Rigney for applying his keen editorial eye to a draft of this essay.

That world is gone. As an Xennial—analog childhood, digital adulthood—I've watched our attention spans and—let’s be honest—my own fragment in real time.

These days, I catch myself doomscrolling Twitter while watching YouTube videos and responding to Slack messages. I'm not sure if this is evolution or decay. Maybe it's both.

What I am sure of is that fighting it is futile. When people talk about "fixing" modern attention spans, they sound like Renaissance scholars lamenting the printing press. They're not wrong about what we're losing.

Multitasking is a myth - research tells us it’s really just rapid task switching and it mostly just stresses us out and makes us dumber.

So, what happens next?

The science of attention suggests that perhaps all hope is not lost. In 1984, before we all carried dopamine slot machines in our pockets, psychologist Christopher Wickens proposed that our brains aren't single processors—they're parallel systems. Visual processing happens in one channel, auditory in another. We've always been capable of handling multiple streams of information. We just haven't had tools built for it.

Consider Memenome, which converts academic papers into TikTok-style brainrot videos with gameplay footage and AI narration. Your first reaction might be horror—mine was. But something interesting is happening here beyond just surrender to shortened attention spans.

Is this essay too long? This talking fish will give you the gist in under 60 seconds.

What's changed isn't just our attention—it's our ability to work with it. Generative AI makes it possible to transform content automatically, to personalize it, to make it flow through these dual channels without armies of video editors and voice actors.

Now we can turn PDFs into podcasts, into talking fish, into video games.

This isn't without cost. Something is lost when we optimize everything for rapid consumption. Deep focus isn't just about information processing—it's about giving ideas room to breathe, about following unexpected connections, about the kind of thinking that happens in the spaces between thoughts.

But the opportunities are too big to ignore. The same principles that make memenome work could reshape technical documentation, corporate training, product manuals—anywhere knowledge needs to move efficiently. The future won't look like a quieter version of the past. It'll look different, and that's okay.

The question is how we build tools that work with these new patterns while preserving what matters about learning and thinking.

If you are building in this area, we’d like to talk to you.

Listen: Apple, Spotify, YouTube

In this episode:

• Why multi-stage firms' move into seed is permanent this time—and what it means for founders and GPs

• Building a systematic approach to founder evaluation after 400+ investments

• How Precursor built a different kind of venture firm with a team of 13

• The counterintuitive approach to AI investing: focus on what works today

Some takeaways:

1. The symbiotic relationship between seed and multi-stage firms is ending after 20-25 years. Multi-stage firms are building permanent seed practices, driven by the desire to access great founders earlier and supported by dedicated capital pools.

2. Previous startup success is less important than previous startup experience when evaluating zero-to-one founders. Being successful in big tech companies often doesn't translate to early-stage success because the skills that drive promotion (stakeholder management, managing up) aren't relevant.

3. For pre-seed investments under $1M, Precursor uses a 75/25 founder/idea weighting - but founders must be "in the neighborhood of a good idea" because there isn't enough runway to pivot to something better.

4. When evaluating founders, problem-solving ability and specific "spikes" matter more than well-roundedness. The hardest traits to coach are diagnosing/solving problems quickly, moving with urgency, and attracting talent when it's irrational to join.

5. Hot companies in the first year of a fund often provide false signals - in Precursor's experience, none have ended up being the companies that truly matter for fund returns, and some have gone to zero.

6. References from managers are often poor predictors of entrepreneurial success because you're judging someone's ability to do a job they've never done before. The key is evaluating their potential to handle 8-10 entirely new responsibilities.

7. Precursor has evolved to become a default seller of ~20% in Series B rounds when pricing delivers meaningful returns and the sale won't disrupt the founder - a strategy developed to balance DPI pressure with maintaining upside.

8. For AI investments, the focus should be on capabilities that exist today rather than betting on future improvements. Many promising ideas are 12-24 months too early and won't survive long enough for the necessary innovation to happen.

9. Life balance requires choosing three priorities from among being a partner, parent, running your business, maintaining friendships, and pursuing interests. Most people can only do three well, and understanding this helps make intentional choices.

Full episode guide:

(00:00) Introduction of Charles Hudson and Precursor Ventures
(01:53) Why traditional seed investing's 20-25 year era is ending
(06:46) Pipeline vs returns motivation for multi-stage firms
(09:40) Early signaling risk and founder-investor fit
(12:18) Finding alpha in non-obvious founders
(15:04) LP perspectives on seed investing
(19:49) Different types of LPs and their decision-making processes
(21:49) Evaluating founders: Pattern recognition from 400+ investments
(26:55) Why previous startup experience matters more than success
(35:56) Charles's founder scoring system and framework
(42:02) Using personality assessments and references
(44:14) Building Precursor's 13-person team structure
(47:42) Why Charles has a five-person productivity team
(48:04) Unique investment team structure and autonomy
(50:46) Being a generalist investor vs specialist
(55:54) Reserves and selling strategy
(01:11:11) AI investing thesis: Focus on current capabilities
(01:19:05) Work-life balance and the three priorities framework
(01:25:16) Final thoughts

Where to find Charles:

• LinkedIn | Substack | Twitter | Precursor Ventures

Listen now on Spotify, Apple and YouTube.

Key takeaways:

• The entertainment industry's leadership is too disconnected from gaming culture, with many executives not understanding the scale and influence of the gaming industry

• The success of Five Nights at Freddy's ($80M opening weekend despite day-and-date streaming) demonstrates the power of engaged gaming communities

• IP holders from gaming should never grant perpetual rights to movie/TV studios and should focus on maintaining control of their IP

• Streaming-only movies struggle to become cultural phenomena compared to theatrical releases, as audiences don't perceive them as "real movies"

• The 2023 Hollywood strikes likely hurt writers and actors more than helped them, leading to reduced content production and allowing studios to reassess their business models

• Multi-tasking is becoming the norm for media consumption, suggesting a bright future for handheld gaming and content that can be consumed while doing other activities

• Asian-developed games, particularly from South Korea and China, represent a growing opportunity in the gaming market

• Deal-making between gaming and entertainment companies often fails due to process misalignment and not having all stakeholders (especially legal) involved early enough

Full episode guide:

(00:00) - Introduction and background
(05:32) - What makes successful game adaptations work
(14:16) - Why the Borderlands movie went so badly wrong
(18:04) - Hollywood's age problem and disconnect from gaming
(23:18) - The entertainment industry's "survive till 2025" mentality
(27:19) - Netflix's gaming strategy
(30:10) - The challenge with streaming-only movies
(34:16) - Community-driven IP and Five Nights at Freddy's success
(42:15) - Future trends in gaming and entertainment
(47:48) - Practical advice for deal-making between gaming and entertainment companies

Where to find Simon:

• LinkedIn: https://www.linkedin.com/in/simonpulman/

• Website: https://www.pryorcashman.com/simon-pulman

In this episode:

• How Kevin used Reddit to seed his initial audience

• The positives and negatives of building for consumers in 2024

• How he tackles the search for product-market fit

• Why he loves big hats

Full episode guide:

• (00:39) Is 2024 a good time to build in consumer?

• (02:48) The two viable paths for consumer startups

• (06:00) What Kevin learned from his last two startups

• (07:44) Finding product-market fit and why he focuses on DAUs above all

• (09:49) How AfterHour works

• (15:01) What Kevin learned at Stripe and YouTube

• (19:37) How Kevin traded his way from $35,000 to $8M

• (24:54) Community as a moat

• (26:46) The death of friend-based social networks?

• (28:55) Balancing startup life with fatherhood

• (30:00) What’s the deal with the big hat?

Where to find Kevin:

• LinkedIn: https://www.linkedin.com/in/imkevinxu/

• Twitter: @imkevinxu

• Website: https://afterhour.com

In this episode:

• Why standard pitch templates are good for investors but bad for you 

• The three-act structure Siqi uses to tell compelling startup stories

• The four crucial questions you need to answer for investors

• How to pick a good name

• The risks of investing in "prestige track" founders

Full episode guide:

• (00:24) Introduction of Siqi Chen and Runway

• (01:15) Why traditional pitch templates don't work

• (03:05) The five emotional stages of an effective pitch

• (05:15) Three-act structure for startup storytelling

• (10:01) Four key questions investors need answered

• (12:33) What’s in a name?

• (17:51) How to balance brand with performance marketing

• (24:02) Lessons from a decade of angel investing

• (27:25) How to evaluate founders

• (35:35) Closing remarks and contact information

Where to find Siqi:

• LinkedIn: https://www.linkedin.com/in/siqic/

• Twitter: @blader

• Website: https://runway.com