Sarah Guo | Invest like the best Summary

Invest Like the Best, EP.489 | Patrick O'Shaughnessy & Sarah Guo

1. Key Themes

The 250-Person Theory of AI Progress

Sarah and Patrick frame the entire AI frontier as being driven by a surprisingly small group of researchers and entrepreneurs — roughly 250 people. Conviction's core strategy is to know all of them, be close to them, and support them before they need capital. This is not a volume-based sourcing strategy but a density-of-relationship one.

"There's sort of 250-ish people that he thinks about, or you guys think about, that are some combination of entrepreneurs and researchers and doing the most interesting things on the frontier. The people that are actually showing up in the morning and pushing this whole thing forward. And one of your goals as a firm is to be as close to those people, know them all and be as close to them and support them in as many ways as possible." 00:07:25

Technology-Forward Thesis Construction: Identify Model-Native Workflows First

Rather than starting from customer pain and working backward, Conviction deliberately identified application areas that were structurally well-suited to what models could actually do in late 2022 — and then found the founders who believed the same. Harvey and law is the canonical example.

"One of the things that we were really looking for in the first year was application areas, workflows and professions, tasks that we thought were a good fit for purpose for the models, which is a very technology forward approach... if you think that we can do next token prediction with language, and you knew that in late 2022, then law is structured language... you need to read a lot of documents, and we had retrieval, and you need to generate text, and there's a lot of precedent text." 00:09:05

Researcher Morale and the Disempowerment Problem at Big Labs

A non-obvious theme: top AI researchers at large labs are feeling psychologically displaced — both because they believe models will do their job, and because the scale of compute required makes individual contribution feel irrelevant. This has meaningful implications for talent migration to startups.

"I definitely think there's a large contingent of researchers who would feel that one of two things is now true. What I do doesn't matter anyway, because the model is going to do it. Or two, the only thing that matters is compute scale. And both of those are somewhat disempowering." 00:12:50

Compute Independence as the Next National Security Narrative

Sarah argues that the bottleneck to AI leadership is not software or talent but energy and compute supply chain sovereignty. She frames this as analogous to energy independence, with equally long lead times and political obstacles.

"There is not a version of the world where we rebuild our industrial base without automation in the United States... I think we're going to start talking much more about compute independence. That's a big problem." 00:39:29

"The version of us having enough compute over the next 5 to 10 years... I was talking to the leader for infrastructure at one of the hyperscalers earlier in the week, and he's like, there was nothing that was going to move the needle for us at sufficient scale before 2030. That's depressing." 00:14:04

Open Source AI: The Cat Is Out of the Bag, Focus on Safety Testing Instead

Sarah takes a pragmatic, non-restrictive stance on open source AI: restricting it only harms law-abiding American businesses while doing nothing to stop adversarial actors. The productive path is rigorous safety testing, not access restrictions.

"If you did restrict use of open source models in the United States, you'd basically just restrict law abiding American businesses and slow them down... the actual attackers or people who have adversarial uses of these things are not affected by your restrictions. You're restricting your own people." 00:37:04

Biology + AI: The Conventional Wisdom Is Wrong, Software Can Win at Pharma

The long-standing VC orthodoxy said you can't make money selling software to pharma — you have to make drugs. Sarah has flipped to the opposite view after seeing Chai Discovery's traction with top-10 pharma companies, arguing AI fundamentally changes the unit economics of computational biology.

"The conventional wisdom was the only way you make money in biotech or serving pharma is by make a drug and then get biobux deals... Dumb software investors, you can't make money selling software to pharma or build platform businesses at pharma. The debate is, does it change with models? I'm a strong yes." 00:45:44

Jevons Paradox Applied to AI: More Productivity Means More Work, Not Less

Sarah explicitly invokes Jevons paradox to predict that AI-driven productivity will expand total economic activity, not shrink it — pointing to software engineering as the first visible proof point and expecting it to replicate across every function.

"I'm hopeful that a year from now, we see Jevons' paradox in practice as we have agents and products that do more of the mundane more effectively in all the domains of our lives. It should look like the transformation that has happened in software engineering... Do you work less now that you are more productive with AI? Yeah, I work more." 00:53:40

Conviction as a People-Dense, Anti-Volume Firm

Sarah only sees four to six new companies per week — a deliberately low number — and spends two-thirds of her time on portfolio companies. The edge is not deal flow volume but depth of community and technical understanding.

"I probably see four to six new companies a week. It's not a very high volume. My first couple months at my old firm, I saw 500 companies." 00:26:38

2. Contrarian Perspectives

Pedigree-Based Investing Without Business Intuition Is Dangerous

The dominant mode of AI investing right now is proxying judgment to brand names — who else invested, who are the founders' references, what lab did they come from. Sarah argues explicitly this is insufficient and dangerous, even when the person is genuinely excellent.

"If I don't understand what they're doing, I can't have an opinion on their judgment... I asked an extraordinarily good investor friend. We had a debate like you do. And I was like, I don't understand. What is this company going to be that would be big? His explanation was essentially, do you know the quality of this person? The technical theory and the business don't make sense to me. And people are making large-scale research bets without any intuition for them or without any opinion on them. I'm like, it's not all going to work." 00:25:54

Recursive Self-Improvement / AGI in Two Years Is a Belief That Has Existed for 10 Years

Sarah gently challenges the prevailing consensus that we are 1-2 years from exponential intelligence, noting that this belief has been held for a decade with Andrej Karpathy himself as a self-aware example.

"Andrej Karpathy will actually in a very self-aware way say, I thought it was two years away for about 10 years. And he thinks it again, to be fair, who can say?" 00:11:40

Grand Strategic Layer Frameworks Are Not Useful for Investment Decisions

The VC industry spends enormous energy debating which "layer of the stack" wins — model vs. application vs. infrastructure. Sarah thinks this is a distraction from the more useful question of where the next 99% of diffusion goes.

"Coming up with some grand strategic framework for like, what layer is going to win here, I think is not useful to me. And I feel like people spend so much of their investing energy thinking about this versus I want to spend my energy figuring out if we're 1% of the way in, what is the next 99% of diffusion?" 00:52:51

Cynicism About Entrepreneurship Is Nonsense; Substance Still Wins

Sarah pushes back on the Gen Z-era view that entrepreneurship is primarily a marketing, network, and Twitter game. She frames this cynicism as actively corrosive and empirically wrong.

"There's so much cynicism about playing the game, be it marketing or fundraising. And I hate that... if we just do the right thing by the customer, we will win. Feels a lot more complicated than that. I think he's right. That seems to be working." 00:34:08

Venture Investing in Semiconductor Companies Is No Longer a Bad Business

The longstanding received wisdom in VC was that semis are not venture-investable due to capital intensity and buyer fragmentation. Conviction has updated away from this given concentrated demand from hyperscalers and supply chain independence motivations.

"Venture investing in semis companies, Lim Boo Tan aside, was like a god-awful business for the longest time... Bella, a partner on our team, started looking at a bunch of these companies and it's obvious that the demand is there... the market is different today. We are seeing consolidated at scale demand for accelerators or even supply chain independence because the big buyers of it want it too." 00:43:35

3. Companies Identified

Harvey AI legal platform. Mentioned as the flagship example of Conviction's technology-forward investment thesis — identified because law is structurally well-matched to language model capabilities (document retrieval + text generation + large precedent corpus). Founded by Winston and Gabe.

"We also took a very specific view of what is now possible and then what is valuable within what is possible... Winston and Gabe, they believed that AI would transform the practice of the law in a very AI-pilled way... from the kernel of I can look at a landlord tenant agreement in California and answer a question, somewhat trivial to, I can project to doing an Activision Blizzard M&A and doing 85% of the work." 00:09:59

Physical Intelligence (π) Implied reference — Sunday Robotics is the company Sarah describes; she references Tony Zhao and Changqi at "Sunday Robotics." Context strongly indicates this is Physical Intelligence. Mentioned for extraordinary speed of iteration from Stanford basement to manufactured semi-humanoid robots doing home tasks, with a targeted beta launch end of 2025.

"This entire team believes that we are going to have general semi-humanoid robots doing things in people's homes first and beta end of this year... We have done hundreds of iterations of hardware on model data collection and translated it into tasks and tested in all these real world environments." 00:21:27

Chai Discovery Computational biology / protein structure AI company. Conviction's first check. Working with multiple top-10 pharma companies in significant ways to accelerate R&D. Mentioned as evidence that software can win at pharma, breaking the conventional biotech wisdom.

"We're the first check in a company called Chai Discovery. And Chai is working with a number of top-10 pharma companies in really significant ways to accelerate some part of the R&D process." 00:45:19

Suno AI music generation company. Mentioned as a missed investment — Sarah passed and regrets it, using it as an example of systematically underestimating consumer demand for AI creative expression tools.

"Mikey Shulman at Suno is building an amazing business doing music generation. Shame on me. I knew Mikey, a mutual friend of ours who's an investor, like asked me to like invest and I stupidly said no." 00:50:41

Sigma Data analytics / BI company. Mentioned as a portfolio company example alongside Notion and Rippling where conviction through a non-obvious early period paid off.

"Mike and I have both had the benefit of being part of the journey for some companies where it took a minute to begin to work... Sigma, Notion, Rippling, you wouldn't bet on a company hoping that it's going to take four or five years to like find the thing." 00:48:07

Rippling HR/payroll/IT platform. Mentioned as a conviction-through-ambiguity portfolio success story. CEO Cod (Parker Conrad implied by context of "do right by the customer") cited as an inspiring example of substance-over-marketing thinking.

"Cod is like this. I find him very inspiring. And Tuhan is like this. I find him very inspiring where he's just like, if we just do the right thing by the customer, we will win." 00:34:08

Notion Productivity/collaboration platform. Mentioned as another example of conviction through a non-obvious early period.

"Sigma, Notion, Rippling, you wouldn't bet on a company hoping that it's going to take four or five years to like find the thing." 00:48:07

Paxilica Jacob Helberg's initiative to map and build redundancy into every part of the AI compute supply chain. Mentioned as the most concrete example of the compute independence thesis in action.

"Jacob Helberg is working on something called Paxilica. And it was like, okay, for every part of the supply chain, can we invest in more capacity and figure out what the independent paths are?" 00:41:55

Base 10 Sarah's prior firm before Conviction. Mentioned in passing as an example of conviction through a non-obvious early period, and as context for her low volume / high conviction style having formed there.

"The first couple years at Base 10 were like very non-obvious as well." 00:48:36

Ramp Finance/spend management platform. Mentioned as a Vanta customer and general reference for fast-growing companies.

Cursor AI coding tool. Mentioned as a WorkOS and Vanta customer, reference point for fast-growing AI companies.

Anthropic Frontier AI lab. Mentioned repeatedly as one of the key labs whose strategic priorities (AGI/ASI, Claude, coding, ChatGPT-equivalent) shape the competitive landscape for application companies.

OpenAI Frontier AI lab. Mentioned repeatedly in the context of lab strategy, compute scale, and the competitive ecosystem.

WorkOS Developer infrastructure (SSO, SCIM, RBAC, audit logs). Mentioned as the enterprise-readiness layer used by OpenAI, Cursor, Anthropic, Perplexity, Vercel.

Vanta Security and compliance automation. Mentioned as serving 16,000+ companies including Ramp, Cursor, Harvey.

Rogo / Felix AI platform for Wall Street / finance agent. Mentioned as sponsor — Felix turns prompts into finished client-ready PowerPoint decks, Excel models, sourced research.

Ridgeline End-to-end investment management platform with embedded AI. Mentioned as sponsor and referenced as the platform for firms serious about AI strategy.

4. People Identified

Tony Zhao PhD student turned co-founder at the robotics company Sarah describes. Identified as having contributed most of the interesting ideas in robotics AI over the last four years alongside his co-founder. Described as extraordinary for his practical, data-centric approach to the robotics generalization problem.

"I think that they have contributed dual-handedly most of the interesting ideas in robotics AI over the last four years. That's a pretty weird thing for two very young people to do." 00:20:04

Changqi (co-founder, same robotics company) Co-founder alongside Tony Zhao. Same attribution — equally credited for the stream of influential robotics AI research and the company's remarkable execution speed.

"My partner Pranav and I were introduced to Tony Zhao and Changqi at Sunday Robotics when they were PhD students at Stanford. They worked at Toyota Research and DeepMind and Tesla, and so they weren't just academics by any means." 00:19:35

Andrej Karpathy AI researcher and former OpenAI/Tesla lead. Cited as a self-aware exemplar of the "two years to AGI" belief pattern — noting he has held this belief for ten years and holds it again now.

"Andrej Karpathy will actually in a very self-aware way say, I thought it was two years away for about 10 years. And he thinks it again, to be fair, who can say?" 00:11:40

Mikey Shulman CEO of Suno. Called out specifically for building an amazing business in AI music generation, and used as a learning example by Sarah — her passing on the investment was a mistake driven by underestimating consumer creative demand.

"Mikey Shulman at Suno is building an amazing business doing music generation... My intuition is just wrong." 00:50:41

Brett Taylor Enterprise AI founder (Sierra). Cited as an example of the rare founder whose judgment is so demonstrably correct across repeated decisions that Sarah would back him even without understanding the specific domain.

"If Brett Taylor wanted to like dig in volcanoes or do dog streaming or something, I'd be like, yeah, of course, man... Brett's doing enterprise AI. I understand what he's doing." 00:24:54

Winston (Harvey co-founder) Co-founder of Harvey. Cited for his belief that AI would fundamentally transform legal practice, which aligned with Conviction's technology-forward thesis.

"Winston and Gabe, they believed that AI would transform the practice of the law in a very AI-pilled way." 00:09:59

Gabe (Harvey co-founder) Co-founder of Harvey, same citation as Winston.

Jacob Helberg Policy/national security figure working on Paxilica — an initiative to build compute supply chain independence. Named specifically as the person working on the most concrete response to the compute independence problem.

"Jacob Helberg is working on something called Paxilica. And it was like, okay, for every part of the supply chain, can we invest in more capacity and figure out what the independent paths are?" 00:41:55

Mike (Conviction partner) Sarah's co-founder and partner at Conviction. Described as a very technical person, exceptional engineer, and source of deal-shaping insight. Originated the "250 people" framework.

"My partner Mike and I started with a set of preexisting relationships and understanding." 00:07:04

Pranav (Conviction partner) Partner at Conviction who co-led the introduction to the robotics founders. Mentioned as part of the investment team that identified Tony and Changqi.

"My partner Pranav and I were introduced to Tony Zhao and Changqi at Sunday Robotics when they were PhD students at Stanford." 00:19:35

Bella (Conviction partner) Partner at Conviction who led the semiconductor company analysis, driving the firm's updated view that semis are now venture-investable.

"Bella, a partner on our team, started looking at a bunch of these companies and it's obvious that the demand is there." 00:44:02

Reid Hoffman Co-founder of LinkedIn, partner at Greylock. Cited by Sarah as someone who took a risk on her when she was 23, representative of Silicon Valley's meritocratic willingness to invest in people based on idea quality rather than pedigree.

"Reid Hoffman, who hired me at Greylock, I started when I was 23." 00:56:14

Dylan Field Co-founder/CEO of Figma. Mentioned as a trusted outside reviewer Sarah would send investment memos to for perspective, and as someone who encouraged her to start Conviction.

"I used to like take the memo and like send it to John Lilly or Dylan Field or something." 00:24:01

John Lilly Longtime venture partner and friend. Mentioned as a trusted sounding board for investment memos and as an early encourager of Sarah starting her own firm.

"John Lilly, who's been a longtime partner and friend... there are a few folks who were just like, you can definitely do it." 00:57:07

Ravi Gupta Now co-CEO of Ithaca. Named as one of a small number of advisors who gave Sarah perspective when starting Conviction — described as offering advice they'd consider trivial but that she found formative.

"Ravi Gupta, who is now co-CEO of a new thing called Ithaca, Dylan Field and Elena Natalinsky, John Lilly, who's been a longtime partner and friend." 00:57:07

Elena Natalinsky Named alongside Ravi Gupta and Dylan Field as a key advisor during Conviction's founding period.

Philippe (Coatue) and Thomas Sarah references Philippe Laffont and Thomas (likely Thomas Laffont) at Coatue as inspiring examples of entrepreneurial investment managers who approach their business creatively.

"The creativity with which Philippe and Thomas go to or that Josh at Thrive approach their business and the encouragement they have for others to approach their business." 00:30:21

Josh (Thrive Capital) Josh Kushner at Thrive Capital. Cited alongside Coatue as an example of an entrepreneurial investment manager whose creative firm-building Sarah finds inspiring.

"That Josh at Thrive approach their business and the encouragement they have for others to approach their business. Well, you can do new things and you should go express your opinions in the form of your investment management firm." 00:30:21

Ashim Channa, Anil Bussery, Joseph Ansonelli Named collectively as Silicon Valley figures who demonstrated meritocratic generosity — evaluating Sarah on her ideas and potential rather than pedigree at a formative stage of her career.

"Ashim Channa and Anil Bussery and Joseph Ansonelli and Reid Hoffman, who hired me at Greylock, I started when I was 23." 00:56:14

5. Operating Insights

Start at Conviction, Then Work Backwards to Find the Holes

Sarah's investment decision process begins with a fast instinctive read (an immediate 8 or 9 on a 10-point scale), and then uses the remaining time before closing to specifically identify gaps in her understanding — not to rebuild conviction from scratch but to invalidate it. This is faster and produces cleaner decisions than the slow climb to consensus.

"I'm very instinctive on people. I often know I want to do something immediately... What I'm then doing between that and a real decision is often figuring out what are the holes in my understanding where my judgment of their premise or them is incomplete or wrong." 00:22:42

Write the Full Memo and Send It to a Trusted Outsider Before You Decide

Even when solo, Sarah wrote full investment memos in the style she learned at Greylock and sent them to trusted outside investors for pushback. This ritual of externalizing your thinking before committing forces clarity and surfaces blind spots that internal enthusiasm hides.

"I'm a memo person. Even at the very beginning of the firm, when it was just me, I would like write the full memo, perhaps Greylock style, ship it off to a friend that was an investor who I trusted for their perspective outside of the funds." 00:23:35

The Best Source of Non-Consensus Information Is Founders Whose Customer Behavior You Don't Understand

Rather than consuming pundit narratives about AI strategy, Sarah deliberately seeks out founders whose businesses are producing behaviors she finds surprising or counterintuitive. Those are the signals that update your model of where diffusion is actually going.

"If it is founders who have companies that are creating a behavior you don't understand, somebody working on research in an interesting direction, businesses that are like here's my plan for AI, all that is super educational. I think the circular logic sometimes of what do people believe about the big lab strategy today — it's actually not that instructive for your decision making." 00:51:37

Leave Deliberate Calendar Slack for Curiosity, Not Just Deal Flow

Sarah explicitly protects time for learning that has no immediate deal attached — pharma executives, public markets investors, researchers. This is a structural discipline, not an accident, and is how she maintains the information advantage that her investment decisions depend on.

"I am leaving room in my calendar for learning, feeding curiosity... Now with a pharma company who is thinking about how AI is going to transform their business. This is very interesting to me because I've learned a lot about what they believe about the future." 00:28:04

6. Overlooked Insights

Autonomous Marketing Departments Are Already Happening Inside Portfolio Companies — Right Now

In passing, Sarah describes a portfolio company where the head of marketing has built a fully autonomous AI marketing department for the company himself. This was a throwaway example illustrating Jevons paradox, but it is actually a signal that the "AI agents replacing functional headcount" thesis is not theoretical — it is live, and it started with marketing, a function often assumed to require the most human judgment.

"I'm thinking about in one of our portfolio companies, the marketing department is like a person in the house. This is a company that serves lots of customers. They need to do very traditional things, sales enablement content. What happened was the guy in charge of marketing is interested in like creating leverage for himself. And he's like, I made an autonomous marketing department for us, the company." 00:54:08

This implies that the first wave of genuine agentic enterprise deployment is not coming from top-down enterprise software vendors, but from individual operators inside companies who are quietly self-deploying — which means the diffusion is happening much faster and more bottom-up than enterprise AI investors are modeling.

Gaming and Entertainment Is Massively Under-Indexed — And Sarah Stopped Herself From Saying Why

In a single sentence, Sarah references gaming and entertainment as a domain where she believes people are dramatically under-indexed, and states she "would love to talk about why" — and then drops it entirely. This appears to be a live, high-conviction thesis that was never elaborated.

"I would love to talk about why gaming and entertainment is going to be totally different and people are super under-indexed on it." 00:32:40

Given the rest of Sarah's framework — identify workflows structurally suited to current model capabilities, find founders who believe in the transformation — this suggests Conviction may be actively building a position in AI gaming/entertainment, a sector that has received dramatically less attention than legal, coding, or biology in the current AI investment cycle.