Jul 3, 2025 · 1h 12m · sourcery
Why a16z Bet $15 Million on THIS Controversial Startup · Sourcery with Molly O'Shea
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of Sourcery, host Molly O'Shea interviews Andreessen Horowitz partner Brian Kim to unpack a16z's $15M investment in controversial startup Cluely, shifting VC metrics in the AI landscape, and category-defining investments across audio, health, and creative tools.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Molly holds 24.5% of the talking time here. How this is scored →
Molly as informed peer 4.1 Guest teaching 2.5 Guest disagreement 1.5 Molly pushing back 1.9
speaking balance: gold is Molly, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 45:20 Rejection of pre-2024 legacy debt premise
Brian explicitly disagrees with the assertion from Molly's CTO dinner that older software companies cannot compete with post-2024 AI startups, pointing to successful reinventions by Replit and Bolt.
Hardest push from Molly ▶ 58:14 Challenging the AI hype cycle and profitability timelineMolly challenges the unchecked growth narrative by asking when AI startups will be forced to transition from burning capital for growth to delivering real profitability.
Biggest teaching moment ▶ 33:39 Schooling on AI discernment and regulation trapsBrian educates on how average internet users struggle to detect AI-generated content compared to tech-native cohorts, while demonstrating why reactive regulations inevitably create unhelpful friction.
Molly holds their own ▶ 1:02:50 Host corrects regulated trading terminologyMolly quickly steps in to correct Brian's characterization of Kalshi prediction markets from betting to federally regulated trading.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Molly as informed peer | Guest teaching | Guest disagreement | Molly pushing back | Why |
|---|---|---|---|---|---|---|
| How Brian Kim Met Cluely Founder Roy Lee | 3 | 2 | 1 | 1 | Molly invites Brian Kim to share the origin story of meeting Cluely founder Roy Lee. Brian tells an engaging narrative about tracking down Roy and winning him over with Costco steak, while Molly mostly reacts and listens. | |
| Thesis Behind a16z's $15M Investment in Cluely | 5 | 3 | 2 | 4 | Molly brings in public skepticism from social media comments regarding why a16z invested $15M into Cluely and asks about the professional risks of viral marketing stunts. Brian breaks down fund dynamics, ownership requirements, and why running into fires fits the firm's ethos. | |
| Momentum as a Moat in the Fast-Paced AI Era | 5 | 2 | 1 | 2 | Molly references Reid Hoffman and Kirsten Green at the Humans in the Loop Summit to discuss rapid obsolescence in AI. Brian expands on his thesis of 'momentum as a moat', explaining why historical product craft rules are modified in the current wave. | |
| Building for Global Scale & 'Cheat on Everything' | 4 | 2 | 1 | 1 | Molly cites Alfred Lin and Yuri Kim on building globally ambitious products and asks if Cluely fits that mold. Brian defends Cluely's provocative 'cheat on everything' tagline as a universal productivity assistant. | |
| Sponsor Segment: Brex | 4 | 2 | 1 | 1 | Following a Brex mid-roll sponsor read, Molly asks Brian about his early investment in ElevenLabs. Brian describes booking an emergency flight to London to lock down the deal before competitors. | |
| AI Context & Investment Thesis Behind Function Health | 4 | 2 | 1 | 1 | Brian discusses the power of context in consumer AI and explains his investment thesis behind Function Health. Molly enthusiastically validates the product with her own biometric testing experience and referral results. | |
| Creative AI Tools, Deepfakes, and Regulation | 5 | 4 | 3 | 4 | Molly presses Brian on the dangers of deepfakes and mass automated creation tools, asking if government regulation is necessary. Brian gently pushes back on Roy's assumption that consumers instinctively spot fakes while arguing against heavy-handed regulation that creates technological gunk. | |
| Exit Dynamics & VC Fund Size Strategy | 4 | 2 | 1 | 2 | Molly asks about frozen exit environments and liquidity options for consumer tech companies. Brian articulates that rapid revenue growth naturally solves liquidity issues through private secondary markets and tender offers. | |
| Sponsor Segment: Fourthwall | 5 | 5 | 4 | 3 | After an ad read for Fourthwall, Molly raises a thesis from a CTO dinner that pre-2024 companies carry fatal infra debt compared to post-2024 AI natives. Brian disagrees directly, citing Replit and Bolt as examples of older companies successfully reinventing themselves. | |
| Disaggregating Platforms and Unleashing Creative AI Energy | 4 | 3 | 1 | 1 | Molly mentions Coatue conference data showing ChatGPT usage rivaling social media platforms. Brian outlines his framework on disaggregating ChatGPT into specialized standalone companies and shifts in how he analyzes seed traction metrics. | |
| Product Shipping Velocity and Momentum as a Moat | 4 | 2 | 1 | 1 | Brian describes compounding shipping velocity as the ultimate non-metric metric for early AI startups. Molly reinforces the point with examples of hyper-growth startups like Whop that ship updates continuously. | |
| AI Application Margins, Unit Economics, and High Growth | 5 | 4 | 3 | 4 | Molly pushes back on AI hype cycles by asking when startups must pivot from pure top-line growth to profitability. Brian demurs and argues that AI application software already enjoys public-company level profit margins and declining compute costs. | |
| Team Efficiency, Revenue Per Employee, and Unbundled Infrastructure | 5 | 2 | 2 | 3 | Molly references Coatue's revenue per employee comparisons between AppLovin and Microsoft before bringing up Kalshi prediction markets on tech layoffs. Molly playfully corrects Brian that Kalshi is trading rather than betting. | |
| Anticipating Breakthroughs in Consumer AI and Productivity | 4 | 2 | 1 | 1 | Molly and Brian run through Kalshi-style rapid fire predictions about Roy Lee, Jony Ive's OpenAI hardware, and ambient wearables. Brian explains why audio form factors will win over smart glasses and pins. | |
| Autonomous Vehicle Scaling, US Markets, and Time to Build | 4 | 3 | 1 | 1 | Molly and Brian discuss the rollout of Waymo in New York and LA compared to San Francisco. Brian provides geographic context by pointing out that two US metro areas represent a larger market than continental Europe. | |
| Sourcery Newsletter Overview and Subscription Call to Action | 0 | 0 | 0 | 0 | Host solo outro promoting the Sourcery newsletter and asking listeners to subscribe across podcast platforms. |
Statements from this episode (25)
Brian Kim: VCs have two to three weeks to close inflecting startups
“When things are inflecting, you probably have two to three weeks to catch it. And by catching it, I mean, convincing the founder, doing the diligence, getting to a handshake, And that's really hard to do.”
Kim: a16z check sizes are dictated by strict fund ownership requirements
“We do have some ownership requirement given the size of the fund and how big we need to believe the company becomes and There's some triangulation of where that needs to work. So maybe the founder says, Hey, like I have a need for thirty million dollars, but t…”
Kim: Generational tech giants often start with controversial or questionable use cases
“Ultimately, we want to back companies that become generational and become very, very large. And a lot of the origin story of companies can be questioned when looked back in from the history or from sort of current looking back, like Facebook started was hard o…”
Kim: a16z does not shy away from polarizing startups like Cluely
“I'm a lot less worried about it, worried about that because as you know, we, as one of my partners put it, we run into fires. Yeah. Right. We're not, we don't shy away from that. So that's okay. What I am more focused on is making sure the image and the brand …”
Kim: a16z backed BeReal because its early user retention was exceptionally high
“In fact, I back be real with or without the growth curve because the retention was so high.”
Kim: Cluely has good margins and founder claims profitability
“We know the product has good margin. He's said a couple of times he's profitable.”
Kim: a16z closed its ElevenLabs deal in four hours in London
“I ended up flying to London sort of told Maddie, I'm like, I happen to be in London tomorrow. Let's chat. And we ended up getting to terms and an agreement in the next, you know, in the four hours I was on the ground.”
Kim: Less than 1% of people fluently use AI today
“Think about the number of people who actually fluently use AI today. I would say that's less than one percent today.”
Kim: AI on 7 billion devices is one or two steps away
“My guess is that there's a strong belief that AI for general population, the seven billion devices out there will become very popular and will become democratized. I can't wait for that to happen. And I think we're probably just one or two steps away from that…”
Kim: Regulating bleeding-edge AI will produce 'more gunk', not innovation
“In terms of, like, you know, enabling companies and people to continue to build in this bleeding edge, I think any effort to curb that will probably not result in more innovation. It'll probably result in more gunk, if you will.”
Kim: Markets will eventually reward startups scaling to $500M run rate
“So when I see companies that are going from zero to 50, zero to a hundred, zero to 500 in run rate in a very short time, to me that says, There's an incredible value that is being created and market may or may not be efficient in the short term, but sure as he…”
Kim: IPO status matters less due to private secondary liquidity options
“Now we're at a place where private companies are very large. They've been staying private for a long time. They're secondary. They're tender offers. There are different ways of getting liquidity back as well. So I sort of look at this trend as larger companies…”
Kim: Pre-2024 Startups Aren't Crippled by Legacy Tech Debt
“The idea that Pre 2024 companies have this infra debt, and it's so hard to unjumble, and you can't innovate, you can't grow fast. I don't think that's true. What I do think is true is you need founders who are understanding of the technical advancement that's …”
Kim: a16z prioritizes velocity in revenue and user growth over retention
“So what I now look for is velocity. And what I look for in those are two to three things, velocity in user growth or revenue. Nowadays people are very willing to pay for tools, especially in this early age. So tools that are actually effectively monetizing and…”
Kim: AI Startups Often See Revenue Retention Outpace User Retention
“What is different now is because of AI, because you can introduce different tiers, you can introduce credit for overage usages or what have you actually see revenue retention being higher than unique user retention.”
Kim: a16z Memos Compare Shipping Velocity Across Competitors
“There are a couple of memos that I've written with a team where we try to compare the product shipping velocity across all players in, in, in the same sort of space.”
Kim: AI app cost structures allow high growth to equal high profitability
“The current dynamic of the cost structure, especially for apps companies, Really favor this equation of gross almost equals profitability. The margin structure of some of these companies that we're investors in is incredible.”
Kim: AI has broken the revenue-per-employee metric with mind-boggling scale
“I think six months ago, Molly, if we sat down and you were like, Oh, what are the interesting metrics? I might've said revenue per you revenue per employee. I don't even know if that makes sense anymore. There are so many companies that are just, that's mind b…”
O'Shea: AppLovin Vastly Outpaces Microsoft in Revenue Per Employee
“Applovin's revenue per employee is, like, huge delta between Microsoft. Microsoft is kind of, like, flatlined a little bit, but, like, that, that is, like, a true trend we're seeing.”
Kim: Tech industry layoffs in 2025 will be lower than in 2024
“I would actually bet that there would be less.”
Kim: AI will not increase unemployment because productivity creates more tasks
“See, this is this is why I don't think unemployment goes up or layoff goes, goes off, goes up because it's just, there's more to do.”
Kim: Jony Ive and OpenAI device will resemble AirPods rather than glasses
“My guess it's an ambient device. I think it'll be closer to I AirPods of the world than glasses.”