What this covers
The conversation is with Hunter Walk, who moves from his earlier work on virtual worlds and YouTube to his current life as a seed investor at Homebrew. The first half draws on how those platforms worked. Second Life ran an internal economy partly to ration the computing load of physics simulation, and it permitted real-money trade where early MMORPGs banned it, accepting that the world would inherit the biases of its economy. Avatars there still kept real-world conversational distance. On YouTube, audio and video fingerprinting turned a takedown system into a leave-up system with ad splits, Creator Grants advanced cash against a temporarily higher platform share, and steady creators produced revenue curves that could be forecast.
The second half turns those lessons into an investing method. Walk prefers Problem Size Over Market Size: BuildingConnected's TAM looked like $100 million, yet it now serves over 160,000 businesses. He tests Founder-Market Fit by asking why someone would spend ten years on a problem. He also describes a Shadow Economy of attention that goes unmeasured until a platform exposes it, and cites theSkimm, Ipsy, and Chime as examples of under-served audiences.
The more contested material includes his view that millennial habits will prove durable, the opening he saw for Shield in government AI, and how autonomy investing changed once conviction flipped. He also discusses what SoftBank's large late-stage checks mean for seed investors.
Walk argues that durable opportunity lies in tools and platforms that lower barriers to creation and share monetization with creators, and that early-stage investors should back founders with deep empathy for hard problems, often following attention before dollars.
- Platforms that natively monetize for creators align incentives and create a durable creator class
- Large, urgent, valuable problems matter more than market size, and founder-market fit predicts success
- Shadow economies of attention and creation grow unmeasured until platforms reveal them, so early investors can find them first
Native creator monetization aligns incentives and keeps platforms self-reinforcing and competitive.
- Once platforms help creators earn money, a native class of prosumers emerges who can quit other jobs and spend time on the platform, which strengthens the platform's ecosystem.
“I think you end up with this native creative prosumer class.”
- At Google, AdSense gave bloggers and community site publishers a contextual advertising option that let them earn money from content, replacing pop-ups, pop-unders, and flashing banner ads.
“We gave them a contextual advertising that they could just put on their page and make money by creating content.”
- The strongest creative platforms have monetization native to the creation experience and pay creators directly, so the company and creators have aligned incentives instead of a platform built only around extracting money.
“I believe the strongest platforms also have a monetization native to it.”
- YouTube's Creator Grants gave creators $2,500 to $10,000 in cash advances, repaid through a temporarily higher platform share such as 80/20 until the advance was recovered, to let creators hire editors and make more videos.
“we just garnished that off of the rev share as they uploaded new videos.”
- The strongest creative platforms form a virtuous cycle in which technology reduces barriers to creation, communities provide audience and collaboration, and monetization rewards creators, and Walk's three prior products all followed this pattern.
“all within a virtuous cycle of technology is a bar that pushes down the barriers to creation. It allows people to be creative. It allows it to be done within communities”
- Platforms that refuse to pay creators, such as Snapchat and Instagram, leave a competitive flank open, and Snapchat is now moving toward creator economies while Instagram remains hesitant.
“if you're not helping them make money... you are A, doing them a disservice and B, you're leaving a competitive flank open for you.”
- Ipsy's YouTuber co-founder turned her trusted audience into a multi-hundred-million-dollar business because subscribers paid for curated goods she personally used and recommended, not for the content itself.
“She wasn't just lending her name to a third party that was going to throw a bunch of crap in a box”
- Ad-driven models that plan to scale and then plug in standard IAB ad units lead to a race to the bottom, so Walk prefers premium sponsorship and brand deals over commodity ads.
“I have not invested in any businesses that talk about we're just going to get to scale and then we're going to plug in an IAB standard ad unit. That's the race to the bottom these days.”
- People will pay for community and identity rather than purely transactional value, so media businesses should think of their model as fan club 2.0 instead of subscription 2.0.
“In media, I call it fan club 2.0 more than subscription 2.0.”
- Many creators who posted only weekly were not limited by talent but by lack of about $1,000 to invest in editing help, and advancing that cash lets them produce more content.
“All these little barriers, which basically just came down to I don't have $1,000 free, available, to forward invest in my business”
- Platforms that fail to help creators monetize and pay their rent are doing creators a disservice and leaving a competitive flank open for rivals that will.
“you are A, doing them a disservice and B, you're leaving a competitive flank open for you.”
- Anchor's phone-based audio tools lowered the barrier to podcast creation so much that it powers about one in three podcasts on the iTunes store, and many new users are people with existing audiences.
“Fast forward, Anchor now powers one in three podcasts on the iTunes store.”
- When a platform builds better creation tools, experimenters who started with no audience rise to build followings, and the platform's next step is helping them monetize.
“What happens is the people who were just experimenting, some of them start to rise up.”
Investors should back founders committed to decade-long, large problems, screening for structural risk and fit.
- In venture, investors should identify structural reasons a company might fail, namely technology difficulty and market readiness, rather than listing every possible risk, and only then ask what happens if it works and whether the upside is worth the effort.
“you can't get too caught up in listing all the things that could go wrong. You have to understand the structural reasons this company might fail.”
- With only two partners, Homebrew cannot search the haystack for startups, so it builds a reputation that pulls founders toward it through inbound introductions and public content.
“I say a venture is finding needles in haystacks. Because there's just two of us, we don't have time to jump into the haystack but we are very good at building magnets to pull the needles towards us.”
- Walk asks founders why they want to spend 10 years on a problem, and founders who recoil from that question, focused only on 12 to 18 months of capital, are poor fits because lasting companies take a decade to build.
“The ones that recoil too much of that question were probably not a good fit for.”
- Investors should ask why now, which asks whether the future the founder describes is accurate and timely, as a complement to why you, which asks whether this team is the right one to solve it.
“Why now? Which is the compliment to why you.”
- Homebrew prioritizes founder-market fit: founders who recoil from the question of why they would spend ten years on a problem are poor fits, while those with a thoughtful multi-phase plan and personal empathy for the problem are strong.
“I call it founder-market fit. We care a lot about how do you describe the problem you're solving? Why do you want to spend 10 years of your life solving it?”
- Hunter argues investors should evaluate problems by whether they are large, urgent, and valuable, since such problems suggest a viable business in enterprise or consumer markets, rather than relying on top-down market size estimates.
“We care about problem size more than market size. If a problem is large, if a problem is urgent and a problem is valuable, that's an enterprise in notion more than a consumer social.”
- Hunter argues that a memorable name and public visibility from a communications role helped him find people with character, and he uses that attention to spotlight young or uncomfortable people with quality of character greater than reputation.
“every time I can use that to put a spotlight on somebody who quality of character is greater than reputation”
- All venture-backed companies rest on a tripod of product, distribution, and team, and investors should deploy resources and expertise against all three.
“I think all these companies are fundamentally built on a tripod of product, distribution and team.”
- A patient investor should back the best companies in the early part of a technology innovation cycle, then stop investing in the later part and harvest, so as to move on to the next wave.
“You try to invest into these best of class companies in the first half of the innovation cycle.”
- Because Homebrew's funds are small and its return depends on a few power-law outcomes, the price paid for ownership matters less than the number of strong picks, so macro valuation swings do not change the strategy.
“My numerator is always going to stay the same. No matter what's going on in the world, my denominator is always going to be large enough that if we're good pickers ... ownership trumps price.”
- Early Esports and Twitch attention signals predicted entrepreneurial markets, and Homebrew rejected any traditional venture-style deal flow in favor of unexpected inbound deals.
“People are filling arenas. Watching Twitch. The hours that they're spending on Reddit.”
- Professional decisions should be made by choosing only projects you would put on your tombstone, meaning projects that represent you and where you can have a unique impact, not merely a job.
“I said, I only want to work on projects that I put on my tombstone. They represent me. They're not just a job.”
Good investments are timed, contrarian bets in overlooked markets before belief flips.
- Homebrew prefers one- or two-horse races, where only a few teams are attempting a new market, over twenty-horse races, because if the bet is right it gains a unique head start on pioneering the market.
“Trying to look for what I sometimes call one or two-horse, horse races, versus 20 horse, horse races.”
- Homebrew invested in Cruise after the founder sent a password-protected video of his Audi with aftermarket self-driving hardware, showing long stretches with hands off the wheel and feet off the pedals, and the thesis became write the check when videos like that arrive.
“He sent us a password-protected video of him driving down to Demo Day in an Audi”
- Conventional venture screens such as requiring a warm intro, rejecting cold emails, and demanding a large TAM slide would have caused Homebrew to miss BuildingConnected, so they are poor filters.
“If I had just gone through all of the venture tropes, "You need a warm intro. I don't want to read your email." I never would have opened their email.”
- Investors should ask why now, meaning what timing change makes a future already visible in a market accessible, such as robotics for produce harvesting, because why-you and why-now together determine whether an opportunity works.
“Why now? Which is the compliment to why you.”
- BuildingConnected, which Homebrew invested in during early 2014 via a cold email from its CEO, now has over 160,000 businesses on its freemium platform, processes over $100 billion of new projects monthly, and expects revenue to more than triple this year.
“They now have over 160,000 businesses on the platform. Over $100 billion of new projects going through at each month.”
- Investors can find value in areas where people disbelieve a technology, and once belief flips, value appears quickly, so Homebrew moved to areas like public sector autonomy and freight that the market did not yet believe in.
“That people are disbelievers but when it flips to if you're right and it flips to believing, there's a lot of value there very quickly.”
- Investor conviction in autonomy lagged, so early investors had an edge, but once belief flipped, many startups raised large rounds and small focused funds then had to move to the frontier.
“When it flips to if you're right and it flips to believing, there's a lot of value there very quickly.”
- A dedicated mission-driven AI team in the public sector has a floor of being acquired by a large contractor and a ceiling of becoming a major contractor itself, so the bet is attractive even under pessimistic assumptions.
“the worst-case scenario is they are a very valuable acquisition by one of the multi-multi-billion dollar contractors. The best case scenario is they are the next great contractor.”
- GM's acquisition of Cruise amounted to roughly 2 percent of its market cap spent to protect the other 98 percent from disruption by autonomous vehicles.
“Essentially, GM spending 2% of their market cap to protect the other 98%.”
- Shield, which builds AI and drone-based tools for the public sector, was a contrarian bet because government is a slow, political market, but the founders' military experience and the government's refusal to buy Chinese drones like DJI made the market defensible.
“The short answer why we made the investment besides being an amazing team was US government's not going to buy DJI drones.”
- A small number of experts understand autonomy deeply, so small teams can outperform large companies in that field, and the best autonomy talent is unlikely to work on government contracts at large tech companies.
“We've learned that small teams can punch way above their weight in this area. There's still a fixed number of engineers who really understand this stuff.”
- Pre-PC industries like construction and agriculture were overlooked by software startups that assumed they were stupid or broken, when those industries had worked before computers because they depended on customer relationships.
“Agriculture, construction, logistics, all these things somehow worked before you had computers.”
Platforms that lower creation barriers and monetize natively form a durable venture lane.
- Shadow economies are hard to measure, sometimes by design, and grow large before platforms report figures, as the YouTube creator economy was substantial before anyone realized creators could earn millions and Amazon self-publishing may dominate bestseller lists.
“I would define shadow economy as economies that are tough to measure and sometimes by design.”
- People pay for symbolic virtual goods like esports skins because they signal identity and tribe, so commerce and merchandise can scale for creator audiences alongside physical products.
“But it turns out, people are willing to pay for things that bolster their identity.”
- Successful media companies diversify revenue across several legs, such as sponsorship, subscription, and affiliate commerce, and these legs become available at different audience sizes and credibility levels, so the stool grows over time.
“Their three-legged stool will eventually become a six-legged stool, eight-legged stool.”
- Online communities like Second Life are very hard to create but also very hard to kill; Linden Lab's Second Life, in its 18th year, remains profitable with around 500,000 paying members, some of whom spend more time in the world than out of it.
“The cut to the chase thing is online communities like this are very hard to create but also very hard to kill.”
- Creators funded through YouTube programs often moved from side hustles to mini media companies by using one widely watched video to cross-promote other videos and brands, such as moving from fashion into travel or lifestyle.
“We start to see these mini media companies spring out of what was at that time 20-year-olds, 25-year-olds, 30-year-olds”
- Platforms that lower barriers to creation, enable community audiences and collaboration, and include native monetization form a durable venture lane, and Hunter believes this lane will remain attractive prospectively.
“I think it's a hugely durable lane.”
Investors should bet early on founders and stay hands-on through trust-building phases.
- Silicon Valley founders who looked down on traditional industries and tried to fix them from afar without spending time with customers often failed, because customers and industries require respect and direct engagement.
“I think ultimately, a lot of those businesses failed because people don't want to spend time with their customers.”
- The kindest thing one can do is bet on someone early by extending capacity, bringing them to one's table, or giving up one's seat, as David Hornik did by inviting Walk to an early invite-only conference when Walk was a mid-level product manager.
“You deserved to be at that first one. The world didn't know it yet but I did.”
- Founders should test whether a practice is a value or a tactic, because values like transparency must be redesigned as a company grows from six people in one room to hundreds, while tactics can be discarded.
“You have this litmus test for what is a tactic? Was it a value?”
- The two most important phases of an investment are the first 90 days, when trust and context are built, and the period between Series A and Series B, when the founder realizes the business is working and the investor steps off the board.
“the two most emotional phases for me that I really enjoy are the first 90 days after an investment.”
- Founders get the most value from an investor who can both pressure-test the tactical plan and check on the founder personally, because most investors specialize in one or the other and leave founders isolated.
“Find yourself somebody who can do both. ... Are you building something you're proud of?”
- Walk's hiring strategy favored people early in their careers with high ceilings, and he believes emotional returns from putting people into business for the first time are more meaningful than returns from investing in already-known talent.
“I'd be lying if I didn't say that the ones where we've put them in business will always have a special place”
Virtual worlds need designed economies and social norms to manage load and real value.
- Second Life used an economy with land purchase, maintenance costs, and subscriptions to manage the computing load of physics simulations, since without it a few users could overwhelm each server running 16 acres and 5,000 objects.
“The way that we chose to do that was essentially with an economy. You have to buy land. It costs money to maintain your land.”
- Avatars in virtual worlds keep real-world conversational distance and social norms even though physical constraints like bad breath or needing toilets do not apply, because users behind keyboards feel uncomfortable violating those norms.
“Because there's people behind these keyboards, it makes them very uncomfortable to violate the norms that we're used to in our person to person interaction.”
- Second Life chose a Burning Man design, dropping users into a world where they must survive with others' help, instead of a Disney World design, which probably capped how many people wanted to participate.
“For better or for worse, we chose Burning Man. That probably actually turned into a ceiling on the number of people who wanted to survive in the desert.”
- Unlike early MMORPGs that banned selling gold or goods for real currency, Second Life allowed real-money trade so users could build things with real value, accepting that the world would take on the biases of its economy.
“We didn't want to go that far. Because we actually wanted people in the world to build things that had real value and transact them.”
- Linden Lab ran the LindeX floating-point currency exchange, taking a transaction fee of roughly 2% for securely converting real dollars into Linden dollars, which effectively rented virtual land at scale and made the platform a best-in-class business model.
“we ran the LindeX, which was a floating-point currency exchange system where in exchange for, I forget exactly, but let's say 2% transaction fee”
Portfolio discipline favors ownership percentage and partial liquidity over higher entry prices.
- Walk expects that Homebrew will take some money off the table in growth rounds of its most promising companies, to return capital to LPs, because later-stage investors are paying forward valuations and it is prudent regardless of company forecasts.
“I bet I would tell you that we took money off the table in growth rounds and some of our most promising companies.”
- SoftBank's late-stage checks, which sometimes buy a third to half of a company, offer early investors a chance to take money off the table through secondary sales, but seed investors may be left along for the ride as the company is restructured.
“Then it has the investments where they're buying a third to half of a company.”
- With funds under $100 million making six to eight investments a year, Walk believes ownership percentage matters more than paying a higher price, since power-law outcomes from a few great companies dominate returns and the fund's numerator stays fixed.
“ownership trumps price.”
- West Coast seed deals are starting at higher prices, with founders raising rounds like 3 on 8, 3 on 9, or 3 on 12, and Walk says it is a place where Homebrew is never the highest check.
“The West Coast deals are now starting at, a lot of people saying, 'Well, I'm raising 3 on 8, 3 on 9, 3 on 12,' type of stuff”
- Small teams can punch far above their weight in autonomy because few engineers understand the field, and once the market flips from disbelief to belief, value rises quickly and later-stage startups become less attractive for a small focused fund.
“We've learned that small teams can punch way above their weight in this area. There's still a fixed number of engineers who really understand this stuff.”
Problem size and demand signals matter more than market size for finding opportunity.
- BuildingConnected's initial $100 million TAM looked small because existing software was expensive and bad, so it represented 100% of spend but only 1% of the problem's value, and the company now has over 160,000 businesses on its platform.
“That $100 million represented 100% of spend but 1% of the problem value.”
- Whether customers are looking for a product (pull demand) or must be sought out (push demand) determines the cost of acquisition, and whether customers must be educated first is a related test.
“Is the customer looking for you? Do you have to look for them? That goes a lot into what's the cost of acquisition going to be.”
- Investors can find future dollar flows by tracking where attention is going, since leading indicators such as esports arenas, Twitch viewership, and Reddit hours show where entrepreneurs will later find monetization.
“If you can durably figure out where attention is going next, you'll have entrepreneurs that will surely figure out where the dollars come from after that.”
- Walk cares more about problem size than market size, and a problem that is large, urgent, and valuable is an enterprise opportunity more than a consumer social one, because such problems usually admit a path to revenue.
“We care about problem size more than market size.”
Commerce and events, not ads, offer creators the strongest monetization upside.
- Under YouTube's model, a wedding video with a U2 song might stay up with ads that pay the artist and YouTube fractions of a cent, rather than requiring the uploader to license the song in advance, which Hunter called one of the most incredible business innovations of the digital age.
“maybe the worst thing that would happen would be at the end of your wedding video, an ad would pop up that Bono would get his 10ths of a cent”
- Hunter does not invest in companies that plan to scale and then plug into a standard ad unit such as IAB format, because that approach leads to a race to the bottom in ad pricing.
“I have not invested in any businesses that talk about we're just going to get to scale and then we're going to plug in an IAB standard ad unit. That's the race to the bottom these days.”
- Merchandise, commerce, and events have the highest monetization upside for creator audiences, while ads trend down and pure subscriptions trend slightly up, because people spend money on things that express identity and tribe.
“I would have a down arrow on ads, a slightly up to the right arrow on pure subscription and all the way up arrow on merch, commerce, and events.”
Virtual worlds reproduce real-world expectations and status hierarchies through design choices.
- Virtual worlds include toilets in homes even though avatars never need to use them, because users are conditioned by physical life to expect them.
“It's because your conditioned. Like I need a toilet. Your avatar doesn't need to do that.”
- Second Life tried to stop real-world wealth from carrying over by allowing trade in its own currency, but the designers accepted that the community would develop its own biases and hierarchies, so the social status of creative producers would replace real-world status.
“We thought about it as of course the world is going to take on its own complexion and own biases. How can we make sure that you can be successful in Second Life without being successful in real life?”
- Prohibiting sales of in-game goods for real money, as early MMORPGs did, both keeps value inside the company and blocks players who buy power, which is why Second Life chose to permit real-value transactions so people would build things of real value.
“We didn't want to go that far. Because we actually wanted people in the world to build things that had real value and transact them.”
- In virtual economies, the people with the best coding and design skills become the most valuable participants, which inverts high-school status and reflects how digital creation confers economic power.
“this is high school except the nerds are the most popular kids because they're the ones that drive the economy.”
- Heavy early usage, such as users online 12 to 18 hours a day, is ambiguous and could signal either success or a need for concern, a question social media debates still face.
“Wondering if that was a sign of success or something we should be worried about.”
Measuring user behavior in virtual worlds yields forecastable, valuable data.
- Because user-created objects had no semantic labels, Second Life combined quantitative measurement with ethnography, hiring an in-world journalist to explain what users were doing since a world of a few hundred people was already too big to understand alone.
“It ended up being a lot of art and a lot of science. The art part was we literally hired a blogger, a game blogger to be our in-world journalist.”
- Avatar customization built from 0-to-100 slider bars made user appearance measurable, so the team could compare early adopters with later users and see how self-presentation changed, and at scale this data would be extremely valuable.
“it's really just a bunch of slider bars with points between 0 and 100.”
- YouTube could forecast creator revenue from early view counts because a steady creator produced predictable views and ad revenue per thousand playbacks, so each new video gave a reliable revenue curve.
“You really start to see almost a predictable revenue curve coming off of any piece of content they uploaded.”
- Technology lets financial institutions accumulate and analyze data, share it in new ways, and build direct, scalable relationships with consumers, and this data-driven shift is where Homebrew sees opportunity in fintech.
“Any time that technology allows an industry to start doing a few different things, A, accumulate data. Understand and analyze that data in different ways.”
Under-served mass audiences, not affluent elites, are the valuable market to target.
- Under-served audiences gather quickly around new content sources and stay durably, and millennial women are the most under-served valuable audience, which Walk attributes to his experience at YouTube and Twitter.
“under-served audiences gather quickly and stay durably with new sources of content”
- theSkimm, a daily email newsletter started by two former NBC producers, reached 75,000 mostly female readers within six to eight months by word-of-mouth forwarding, even though every VC they pitched rejected them, claiming email was dead.
“Six, seven, eight months later, they had 75,000 mostly women reading it everyday.”
- Chime built a debit card with no overdraft fees and no minimum direct-deposit penalty, and it chose test markets such as Sacramento and Kansas City, targeting tens of millions of under-serviced Americans instead of affluent San Francisco engineers.
“One of the brilliant things they did early on was their choice of test markets. Sacramento, California and Kansas City.”
- Millennials use banking, news, and parenting differently because they grew up with different technologies and value systems, and Walk believes these generational differences will be durable rather than a passing phase.
“It's saying that these generations of consumers who grew up with different understandings of technologies, different value systems are going to be durably doing these things differently.”
YouTube's scale and valuation show how hard platform value is to forecast.
- YouTube grew from about 100 million playbacks a day in its first 18 months to over 4 billion playbacks a day during Walk's tenure leading product, which was when Google had acquired it for roughly $1.5 billion.
“We went from 100 million playbacks a day to over 4 billion playbacks a day.”
- When asked how Google knew the $1.5 billion YouTube price was right, CEO Eric Schmidt said it was either way too high or way too low and that the answer would be known in 10 years, and Walk believes it was too low.
“It absolutely wasn't the right amount. It was either way too high or way too low and I'll tell you in 10 years.”
- YouTube used audio and video fingerprinting to let copyright owners identify content automatically and choose to take it down, track it, or monetize it with an ad split, turning a takedown engine into a leave-up engine.
“We essentially turned what could have been just a take down engine into a leave up engine.”
AI and SaaS waves augment traditional industries and sometimes serve overlooked markets.
- Shield builds AI for the public sector, and Homebrew saw an opening because the government would not buy Chinese DJI drones, and because elite AI talent at Uber, Tesla, Apple, Facebook, and Google was unlikely to pursue government contracts.
“US government's not going to buy DJI drones.”
- TrueAccord uses machine learning to choose the right channel and offer for each debtor, aiming for a win-win collection process that could eventually let it buy debt and collect it itself, rather than selling it cheaply to abusive collectors.
“He believed he could use machine learning across all these data to understand what value proposition and what channel to use.”
- The first wave of Homebrew's SaaS investments digitized workflows in industries that predated PCs, such as construction, legal, and logistics, and the current wave adds AI and machine learning to traditional SaaS.
“It's essentially the digitization of workflows that hadn't yet been digitized.”
- Outlier applies AI to business data to tell managers which questions to ask and which metrics are anomalous, so knowledge workers are augmented rather than replaced, a shift beyond first-generation dashboards.
“They call it AI for your BI. As opposed to just having systems that show you graphs”