Big companies only have one A-team

She starts by conceding one case where the fear is real: go up against search and advertising, and you're facing Google's actual A-team — and those people are genuinely good.

But most startups aren't in that situation. She says a big company's product portfolio is so wide that most of it isn't staffed with the A-team — because there's only one A-team, and it doesn't stretch that far.

If your product is Google's priority number 27, that's really not so scary. Sarah Guo

She has a name for this: organizational physics. The idea is that there's a hard physical limit to how many things a company can do well at once. Unless it was built from day one to fight on multiple fronts, it'll only excel at its core business and be mediocre everywhere else. That's not about good or bad management — there just aren't enough people to go around.

Her point, then, is that founders should feel more confident going after everything outside a big company's core business. Those areas get B-team and C-team staffing — which, for a startup, is open space.

So where exactly is Anthropic's core?

Which raises the question: how do you know where a company's core actually is, and whether you're about to run into its A-team? The host pressed further: what was Anthropic built to be good at?

Her read is that the model itself is the business — the core use case is code and models that can improve themselves, pushing capability forward toward AGI — and everything else is fairly secondary. What backs that up is what it has deliberately chosen not to do.

So why don't the big labs just scoop up the application layer while they're at it? Her answer: they've actually tried both horizontal and vertical applications, and it hasn't worked out — not because they can't see the money in enterprise. Every step of running a software business that actually satisfies customers is hard:

None of these four hurdles has anything to do with model capability. That's also why she thinks the big labs leave room for everyone else.

So if big companies aren't as scary as they seem, where should you look for opportunity? That's actually the question Matt opened the interview with.

She finds opportunity in reverse

Matt had thrown out a hard question at the start: all the old conventional wisdom about software companies is void now, everything from the bottom layer to the top can be rethought — how do you invest in a moment like that? He said it makes him deeply uneasy.

Her approach is to flip the order. Don't pick an industry first — watch what new capability the technology just unlocked, then find a need that already existed and connect the two.

That's how the Harvey investment happened.

In the second half of 2022, what had just become possible was this: machines could generate long stretches of text, and they'd only get better, with smarter retrieval along the way. She says she still doesn't really know what lawyers do day to day, but law looks like a job that consists of reading huge volumes of text and then writing huge volumes of text — and that's probably the core of the profession.

You just work from that premise: we now have a magic box that conjures text, and paired with the right founder, where could you point it? That's not some complicated judgment call. Sarah Guo

Matt cut in, noting that she was measuring time in months. She said, right — plenty of things she didn't believe a few years ago, she believes now. You have to keep updating yourself.

After text, the next direction she zeroed in on was real-time voice, on the theory that AI could now handle a phone call convincingly. The key is not to fall in love with an industry — follow the technology instead: see what it just unlocked, then ask which old problem it can solve.

The next question follows naturally: if that's how opportunities surface, how much time do you actually have to seize one?

AI doesn't waive a single rule of business

The host laid the question out plainly: distribution, margins, the health of the business model — do those old items on the startup checklist still matter, or has AI handed everyone a free pass? Her answer: nothing gets a free pass, but the state of every one of them can shift.

You have far less time to waste than before

She thinks market leadership hasn't been made obsolete by AI — it still matters just as much. What's changed is the window you have to grab it.

Five years ago, someone could spot an undervalued software workflow and say: once my equity vests, I'll come out and build it in three months — back then, that delay wasn't fatal. Not anymore. She uses code generation as the example: three months later, the way people use that capability might have changed entirely. You don't just lose three months — the product shape you were originally aiming at no longer exists.

Her explanation: the moment a good opportunity surfaces, some eighteen-year-old somewhere pounces on it, and the money follows just as fast. So she says the point here isn't really time — it's urgency.

Anthropic's revenue came from developers showing up on their own

On distribution, Silicon Valley's long-standing consensus has been: bottom-up adoption by end users can carry you part of the way, but getting to billion-dollar revenue eventually requires a sales team and an enterprise playbook. She uses Anthropic to push back on the second half of that consensus.

Anthropic getting to $40-billion-plus in revenue was definitely driven by end-user adoption. Developers decided: I want to use the best model on the market, the best framework, the best product — so I just used it, and I ended up spending a lot of money on tokens. Sarah Guo

She added: sure, there's a sales team, but they're mostly just taking orders.

The host then brought up a question he asks every founder himself. Whenever someone answers that their go-to-market strategy is "we'll hire a sales team," he fires back:

Name me a less efficient go-to-market motion. Just name one — something less efficient than hiring a pile of sales development reps and account managers to sell your product. Matt MacInnis

He says founders pause for a moment, then admit they can't think of one.

So what makes a founder good at distribution? Her measure isn't which playbook they name — it's whether they've taken ownership of the problem. She points to Zach at Open Evidence: a deeply engineering-minded founder running a free product who couldn't spend unlimited money on acquisition, so he and his team kept asking how to get every doctor in the country to know about the product for as little money as possible. One thing they did was print and mail out a medical research journal — because for a free product, that's cheaper than running a sales team.

In the end, a company is just a machine

This one comes from Aaron Levie. After Box went public, the two of them were chatting over coffee when he said something that's stuck with her ever since.

Sarah, you need to tell these founders their job is to build a machine that produces cash flow. I didn't realize it before, but my job was building a cash-flow machine with a highly efficient sales motion bolted onto it. Aaron Levie's words to Sarah Guo, as she related them in the interview

The host filled in the first-principles layer underneath: a company's value is simply all its future cash flows discounted back to the present — that's it. The revenue multiples and other shorthand valuation methods we use day to day are just rough rulers pointing back to that same principle.

The host said that once this clicked for him, every other decision snapped into place: why you should worry about burn rate, why you should watch margins, why you grab customers now instead of in six months — because all of it eventually converts into cash flow at some future point.

Why a company has to get big

On ambition, she starts with an answer she herself calls "not exactly inspiring": a venture-backed company only has two good endings.

Both paths demand scale, or you have to become a genuine threat to someone. She says founders start out thinking about some customer problem, but the ending is an entirely different optimization problem: how do you get to $500 million in annualized revenue and keep growing, or how do you become an asset someone absolutely has to buy?

Since the ending demands that kind of scale, the next question is how you choose your path from day one.

Giving up on "nail one thing first"

She says this is Conviction's biggest shift in thinking over the past few years — partner Mike Vernal has written about it specifically. Silicon Valley's traditional mindset is: pick one feature, and make it the best.

The logic sounds clean: a feature is your wedge into a product, a product is your wedge into a suite, a suite is your wedge into a platform. She disagrees, for two reasons.

The first is just looking at outcomes — you can find examples from both the SaaS era and the AI era showing that this theory has little to do with whether a company actually succeeds. The second, which she thinks is the harder point, is that customers don't think this way at all:

If you ask customers, they don't want a million vendors. Being a single-point solution is just a temporary state. Sarah Guo

Following that logic, AI lets a single company do far more, so the founders best suited to this moment are the ones taking on more scope. She's clear-eyed about the cost, too: more ambition usually means more risk — you have to assume your second bet isn't any less likely to pay off than your first.

Old markets are consolidating, new ones are opening

The host raised an inference here that's bad news for investors: if AI compresses what used to be many separate billion-dollar categories into a smaller number of much bigger companies, that's bad news for the return structure of venture capital.

She half-agreed, then reframed it. The existing software market really should consolidate — her own words were that the space shouldn't have this many SaaS companies. But there are still plenty of markets software hasn't entered yet. Her example is Sierra.

She says she invested in Brett Taylor and Clay Bavor as people, not in the "call center software" category. That category has historically sold something quite thin — essentially just the result of "the call goes through" — and barely even gets to claim it saves customers money. Sierra, though, is operating two abstraction layers higher.

That answers the host's worry. Consolidation really is happening, but that's the math for the old market. Move two layers up and you're selling something different — the whole ledger changes.

With direction covered, one thing remains: all of this takes money, and the money sits with investors.

How to read the investor sitting across from you

The back half of the interview turns to the investor side of the table. She thinks founders invest far too little effort here: before fundraising, hardly anyone has seriously studied how the process of getting a VC to write a check actually works.

My framework is simple: treat them like an enterprise buyer. There are other stakeholders involved, you need to understand the process, and there are several steps in that pipeline. Sarah Guo

She gives a two-layer classification and suggests sorting the investor into it before you even start talking.

Social capital is a limited resource

This is the most practical concept in the episode. She doesn't want founders burning energy overanalyzing favors in the investing world, but she does want them to see who's paying the social cost in every interaction.

If I need a favor from Sundar at Google, the number of times I can ask isn't unlimited. People generally keep a running tally in their head — am I adding a burden to this person, or helping them? Sarah Guo

In practice, that looks like this. An investor introduces you to a few of his growth-stage investor friends — the motive might be that those people are LPs in his fund, and he's promised them first look at the companies he backs. That's not aimed at you specifically, but it's not really in service of you either.

She gives two questions you can ask yourself on the spot: did you ask for this introduction, or did they offer it unprompted; and is the investor they're handing you actually the best fit for you, or are they just banking points for themselves?

The other half of the homework is reading what the person actually cares about. Wanting to make partner, wanting to make a lot of money, a genuine love of technology, wanting founders to adore them, just liking being in the room — these are very different drivers. She says that if you understand the people you're doing business with, you can predict their moves much better, and that effort is worth spending.

Her own rule: assume the money could stop at any time

Last comes the current capital wave. The host describes this moment as Renaissance Florence, just with a mountain of capital pushing it along. Her view is that cheap capital gives the same amount of progress a higher valuation, which is why bubbles are actually good for building infrastructure.

That's exactly how a bubble is good for infrastructure. A batch of founders and investors will lose a lot of money, because in hindsight, it wasn't the rational thing to do. Sarah Guo

Her own discipline is to assume the music stops. Land-grab decisions — spending $200 million to train a model, or loosening a sales comp structure from 6-to-1 to 4-to-1 — only make sense once you rerun the math without assuming capital keeps flowing forever. That's the only way to know what you're actually doing.

She points to friends' examples: several software companies with revenue in the hundreds of millions of dollars were once spending on ads with a 14-year customer acquisition payback period — meaning a customer had to stick around for 14 years just to break even. In an era when the cost of capital was near zero, that was rational in a vacuum, but that assumption won't hold forever.

How Conviction built itself in four years

Finally Matt asked a question unlike anything before it: Conviction has only existed since 2022, and building a name the industry actually recognizes in four years is genuinely hard — how did she pull it off?

She starts by saying luck mattered more than skill, then lays out three concrete things she actually did.

First: pick something important, actually go understand it, and be inside the circle before the crowd shows up. She says most people decided to do AI around 2023 — if you already understood it and were already embedded in the ecosystem before that, it counts for a lot.

Second: run the firm like an actual company. She says even she thinks this sounds dumb to say out loud, but most VC firms' self-positioning boils down to "we love founders, we have money, and we pick winners."

That's not persuasive. Sarah Guo, on how most VC firms position themselves

Her breakdown: to a founder, every VC firm looks basically the same — each has good investors and bad ones, each has hit good companies before, and back then, nobody knew who Conviction even was. What actually needs solving is an information asymmetry problem: why should you believe my network is better than the next firm's? Her answer is don't just say it — go do partner marketing, put photos of yourself with the people who matter online, and belief comes a lot more naturally.

She also brings up something partner Mike Vernal says: at any given moment, there are roughly 250 people in tech doing the most interesting work, and that list is always turning over. Their approach, regardless of whether those people are founders, executives, researchers, or operators, is to find them first, understand what they're doing, and then be useful to them. She says that's closer to sales than to marketing.

Third: don't be bland. The upside of starting four years ago was having nothing to lose, so she could take bigger risks. Put opinions online and you'll definitely get some wrong, but she says she's hardly the first person to say something wrong on the internet. And her readers are engineers — get something wrong and they'll email you to point out exactly where, which beats a reply of "who even are you" by a mile.

Near the end of the interview, Matt asked how she wants founders to understand Conviction. She gave two lines that sound like principles:

We're happy to bet on a great founder even before we're fully sold on the idea. But we will never pair a market we think "should exist" with a team we don't think is right for it. Sarah Guo

Beyond that is her view of the whole picture. She says the big labs are all genuinely impressive, but if it all ends with just three companies winning and everyone else reduced to infrastructure, that won't be this economy at its best. What she wants is for innovation to spread faster — with the right money, backing the right people.

Sarah Guo doesn't answer yes or no. She has you figure out first: who inside that company are you actually competing with. Where does what you're building rank internally, and what caliber of person would they assign to it?

Her method for reading Anthropic is exactly this: the model itself is the business, and the core use case is code. The evidence is that it holds back from video and image models — partly out of safety concerns, partly just focus. What a company gives up says more than what it claims to prioritize.

The idea: there's a hard physical limit to how many things a company can do well at once. Unless it was built from the start to fight on multiple fronts, it'll only excel at its core business and be mediocre everywhere else. It's not about good or bad management — there's only one A-team, and it doesn't stretch far enough.

Big companies being less scary doesn't mean you can take your time. She thinks market leadership hasn't been made obsolete by AI — what's changed is the window you have to grab it. Spot an opportunity, sit on it, and when you come back you might not even recognize it.

All three figures come from Sarah Guo's own remarks in this interview; this site has not independently verified them. Following this line of questioning, she also offers two more points:

• Don't cling to "nail one thing first." Customers don't want a million vendors — being a single-point solution is just a temporary state. AI lets one company do far more, so taking on more scope actually fits this era better. This is Conviction's biggest shift in thinking in recent years.
• Read investors like enterprise buyers. First sort out whether you're facing a large firm or a fundamentally early-stage one, someone who can decide alone or someone who needs a room full of sign-offs; every favor you ask spends a limited stock of goodwill, so know who you're spending it on and who it's actually serving.