Why software moats just disappeared

I pulled the transcript of Thrive Capital’s Miles Grimshaw on Dialectic with ytmd, a local captions-to-SQLite tool I built so an agent can save a YouTube video, search it, and read timestamped windows instead of guessing from the title.

The conversation is about 1 hour 53 minutes. Jackson Dahl talks with Grimshaw about investing as an infinite game, why classic SaaS moats collapsed, and how Thrive actually partners with companies like Cursor and Turbopuffer. These are the notes I wanted to keep. They are Grimshaw’s claims from the interview, not independent analysis. Captions can be slightly off on names and numbers.

The job is an infinite game

Grimshaw’s worldview is that you create the board. You do not just try to beat someone on a fixed course. He grew up on adventure racing: 60-hour orienteering events with a map, compass, and checkpoints, sleeping an hour a night on the forest floor. There is a winner, but the course is emergent. You move only as fast as the slowest teammate. The point is whether you can go the distance, not whether you beat a world-record mile. (4:30)

That is how he sees entrepreneurship and early-stage investing: bushwhacking, not running the marked path.

A founder who wins, Stripe or Cursor, compounds in one container for 20 years. A successful VC is sent back to the base of the mountain. After Cursor sold to SpaceX, a friend called him and all he said was: I just need to find another. What do you got? (8:00)

A coach once sent him Camus: imagine Sisyphus happy. Each climb has its own dimensionality even if it looks like the same mountain. Thrive’s job is to join the founder’s expedition, then go back down, blank slate, and hunt a new summit. (8:55)

Most people do not opt into being Sisyphus. The edge is actually enjoying the reset.

Stay hungry on purpose

Success insulates. Your circle gets bigger and smaller at the same time: more people, more yes-men, more inbound from people who already know you. Founders and investors both drift into only talking to winners. Breaking that is active work. (11:24)

He treats curiosity like lifting. Every week, month, and quarter: where are we hunting versus what is coming at us? His bet for 2026 is that the majority of Thrive’s serious investment conversations were proactive, not inbound. Early-stage ones often started years earlier as friendship and support.

Impatiently patient is the Thrive ethos. (17:31) Impatient about learning, meeting people, exploring uncomfortable domains. Patient about the right moment, team, and price.

They spent a year-plus learning AI supply chain and hardware with zero silicon experience on the team, walking into rooms saying “I don’t know,” writing things down, using ChatGPT, and only then made their first investment. Same pattern on Stripe: they partnered at $3B and $5B, then stepped back in when ZIRP unwound and others were fearful. (19:31)

Investing as courtship

He does not think of early-stage investing as selling the founder on money. Capital is the ticket. The real ask is joining the team for a decade. Gut-check on every deal: can I imagine recruiting for this company tomorrow and in year 10? (21:37)

Turbopuffer. Heard about it because Cursor used it for semantic search. Thesis: when knowledge is oxygen for LLMs, search gets reinvented. One call with Simon; he was wildly impressive. Grimshaw flew to Ottawa for dinner before anyone was hounding him, term-sheeted the next day, and Simon said he was not ready. Bootstrapped, maybe profitable. They stayed friends for a year-plus. Then a Monday 9:30pm call: this is that call. Thrive huddle at 10:15pm, committed by 10:30pm. Thrive and Locky are the only outside investors. (14:04)

Cursor. Knew Michael about a year before the Series A, in the infancy of the product. Never got a spreadsheet. Started with the person and the future they see. Michael most prized recruiting help, so they ran case-study conversations on how to sell candidates. Michael later said he learned recruiting from Miles trying to sell him. After the A, Thrive was effectively full-time recruiting. Michael then picked them to lead the B. (23:23)

The best early deals look like courtship. You sow seeds with no guarantee of a round. Speed only works because the relationship already exists.

Love of change is trained

He is the eldest of seven. His mom raised them to be the best version of themselves, not one ideal. Moved from West London to the US at 12. When she sat 11-year-old Miles down to warn him about the hard parts, he said we gotta go. Wrote the tooth fairy asking not for money but to help his sister get into the American school. Later spent six months at a Himalayan boarding school in India because he saw photos and decided he had to go. Learns painful sports on purpose: kitesurfing, alpine snowboarding, flying. (25:46)

The hard part is not talent. It is living in the place of sucking. Uncomfortable experiences that worked became a reward function. Jackson’s line, which Miles stole: confidence is the memory of success. (30:34)

Company genetics

The environment is changing faster than ever, so you should question every premise. The old premise: building software is hard, and maintaining it is harder. A lot of SaaS “truths” were downstream of that and may no longer hold. (32:33)

What still does not change:

1. Create more value than you capture.
2. Someone has to give a shit.

His two Thrive meme questions: What is it? and Who cares? What is the customer actually buying, and who is it for? (34:24)

Do not copy the wrong lesson from winners.

• Google’s 20% time was possible because they had the best business model in history, not because 20% time is a law.
• “Every SaaS company is 10x ARR” is not true. Revenue is not created equal.
• Calling labs “labs” hides that Sam Altman is extremely product- and customer-minded. The better lesson is Cursor: relentless product, tiny sales team, tens of millions of ARR with zero salespeople. They were the customer. (36:11)

Phenotype versus genotype

House-cat kittens and tiger kittens look similar. Early exponential growth is phenotype. Genetics are what it becomes. Spreadsheets are backwards-looking. You are trying to see Michael Jordan as a kid. (38:18)

Two things he actually looks at.

Where is the founder looking? You go where you look, in driving, biking, and companies. Stripe and Braintree looked similar early. Patrick was already pointing at GDP of the internet. Michael Truell was not looking at the IDE. He was looking at abstractions of coding, making coding easier. The IDE was today’s manifestation. (41:16)

The atomic customer relationship. Not ARR growth or time to $10M / $100M. He kind of does not care. Three $50M customers versus 500k customers at $100 are different organisms. Neither is automatically better. Ask: what is that one repeated thing, and what can it become?

Cursor as a first-principles dream: the IDE market is terrible, mostly free, JetBrains maybe around $200M. If intelligence sits at the center of that surface, value per user can be thousands to $10k, not a $200M TAM. Write that on a piece of paper and underwrite the dream, not the historical category. (45:02)

The job of the investor is not to enumerate ways it fails. What if it went right? What if it went really right? Is that implausible? Is this the unique team to do it? Then help make it as sure as possible. (47:45)

Cursor at investment time: models were not good enough to do it all. You needed human oversight, so the product had to couple human craft with more intelligence, from autocomplete to pseudocode to code review. People forget Cursor Tab was their own model. Full-stack early. The bet was product times model for the whole software-development supply chain, not an IDE.

Customers as evolutionary pressure

Early ARR is not created equal. Demanding, avant-garde customers are a fitness function. Stripe would be a different company if Delta were customer one instead of Shopify. Shopify’s scale, Black Fridays, and Toby as a developer’s developer pressurized the company into the future. (50:48)

Advice to a Series A company: pick the customer who helps you live in the future of your market, not the one who writes the biggest check. You need some oxygen, but venture exists so you do not have to sell your genetics for cash.

Why the moats disappeared

People used to talk about magic number, Rule of 40, spreadsheetification. His visual: great software companies built cities. Salesforce did not just build CRM. It built Sales City. Agglomeration, all the bakers on one street, plus wall-clock time plus mindshare. An alien doing sales looks at the map and goes to Sales City. Switching costs, rare network effects, years of accumulated software. (55:11)

AI deleted wall-clock time. You can now build the city in a day. It may have no residents yet, but the construction moat is gone.

Offline, RadioShack, Bed Bath & Beyond, Staples, and Blockbuster worked: pick the top 30 metros, occupy the corner, run it well, nobody else starts. The internet killed that. You had to become Amazon or Walmart, do everything, thin margin, huge surface area, or LVMH, luxury. Mid-market specialty retail died.

He thinks software is heading the same way.

• Do a lot. More surface area, less margin, intelligence at the center. Security already trending this way. Salesforce too if it reimagines around intelligence.
• Or be luxury / white-glove. Palantir as the original forward-deployed engineers. “It’s only FDE if it’s from the Champagne region of Palantir.”

The old rule, never build bespoke, maintenance will kill you, assumed software was hard to maintain. That assumption flipped. More companies should say yes to one customer’s n-of-1 feature. If you cannot ship it fast, you are not using the tools. If maintaining a slight variant is a minefield, your architecture is not adapted to software’s new ephemerality. (1:01:16)

Is software over for VCs? No. We will consume more software, not less, like media after YouTube. The ceiling of any market is value created for customers. The shift:

software that helps you do work → software that does the work

helps you make decisions → makes decisions

helps you coordinate → plans and executes

That raises the value ceiling. White-glove players get paid well because they create a ton of value, tailored. Remaining durability: switching costs, trust, imbued know-how. Craft still matters because, as Patrick Collison put it, attention to detail is how customers decide they can trust you on the stuff they cannot inspect. Interior design as training: once you notice the couch legs, you never unsee them. In an abundant-AI world, intentional human decisions become the signal. (1:05:06)

Change is the oxygen for new companies. Incumbents get caught off guard. That is why entrepreneurial energy is so high. He still loves software because there is no gatekeeper: four MIT 20-year-olds beating Microsoft at IDEs; Tom Blomfield saying he would build a better UK bank, Monzo, about a tenth of the UK as a daily checking account. He hopes he will not be remembered as a software investor, but as a partner to visionaries who started small. (1:08:31)

Intuition is trained judgment you can explain

Early on you do not know the factors, let alone how to weigh them. Over time you gradient-descent into feel. Do not trust raw intuition until you have had reps. Then you must make it legible: if your partners cannot see what you see, that is your failure, not theirs. If you cannot communicate it to the people who trust you most, the world will not get it either. (1:11:12)

Humans are relative, not absolute. You need exposure to true greatness to calibrate. Once you have seen the very best, how great great is becomes obvious.

Mesh Optical. Three people who just left SpaceX, Starlink laser interconnect. Introduced by a founder he never backed but stayed close to. Friday meeting: they authentically love building lasers; vision is photonics manufacturing; they absorbed SpaceX culture so they will be able to lead, not just invent. Saturday meeting, Sunday team pitch, Monday, a holiday, committed, in optical transceivers, a domain they had absolutely zero business knowing. He would not have had the confidence five years earlier. (1:14:48)

How he learns a new domain

Three most important words for an investor: I don’t know. Four words after that: I’ll figure it out. An investment team that cannot say “I don’t know” to each other is dangerous. Story from a public-markets investor: the legendary interviewer would ask an impossible tax-jurisdiction question. The only right answer was I don’t fucking know, let’s talk to the tax person. (1:18:47)

He does not put a clock on learning unless a decision forces one. Walk in naive. Sorry, I did not understand that. Let me play it back. Can I analogize it this way? High ego, I’ll figure it out, plus low ego, willing to look dumb. Get to ground truth, not marketing answers.

What VCs actually do

They do not make bets. They make commitments. If you do not believe it, candidates can tell. Selling starts before the call. (1:22:17)

Recruiting well means helping the candidate make the right decision. A bad-fit hire is a management problem for the CEO. Ask where the candidate is looking. They are the writer of their book. This is the next chapter.

The board job is to help founders make better decisions. The decisions are always theirs. Give perspective, including views you do not fully believe, so they can still be confident after seeing the other world. Not there to push an opinion.

Michael Moritz: demanding in good times, supportive in tough times. A shock absorber. Good times are when you can level up: next recruit, next customer you should not have yet, next capital raise. That is when boards usually check out, and when they should lean in. Bad times: the founder has already thought about every failure mode. They need a shoulder, not more pressure. (1:27:12)

Thrive, Benchmark, and being a punk

Turned down Bridgewater. Curiosity there was about organizations, not macro. Read TechCrunch, Hacker News, and David Skok’s For Entrepreneurs SaaS blog around 2011–12, interned at Matrix, met Will Gaybrick, who joined Thrive and pulled him in. Thrive was then mostly consumer: Warby, Harry’s, Instagram. He pushed software: Segment, GitHub. Early reinforcement learning on greatness: Instagram, GitHub, Josh building Oscar. Josh bets on will, not proven ability. Their latest investment-team hire was 20 and dropped out of college. (1:29:10)

Jared Weinstein: Miles was a punk, as a compliment. Fanatical about the games he picks, inattentive to the rest. Not an A-student of other people’s tests. Early on, could not communicate intuition well.

Investment committee as music studio, not jury trial. Worst version: I want to do this, your job is to find me guilty. Best: everyone came with sheet music, then you jam toward the right answer. (1:34:47)

Leaving Thrive for Benchmark, then coming back, required humility. Ignore others’ perception, admit the fit was wrong. He needed to leave to know the water he was in. He had never been anywhere else. Learned a lot: Eric Vishria on theory to ops, Peter Fenton’s range, Jack Altman. Thrive fit: insanely collaborative, always-on. Turbopuffer at 10pm Monday. A call after the World Cup final. Josh flying NY–SF twice in three days around the OpenAI $150B deal, then showing up Thursday 7am with the next three ideas. He likes getting the Saturday 7pm family-dinner call. (1:36:16)

From Josh Kushner: dream bigger. Bet on the person who wants it, not the most credentialed. Will plus can-do. And not just winning, the way we win. Kind and competitive, often versus yourself. (1:43:11)

Kids, agency, endurance

He will not pretend to know what school will look like. Two things: be a kind person, a daily muscle, and agency. Intelligence is abundant. The scarce thing is what question you ask and where you want to go. Favorite interview question: a relationship that came from a cold email. Cultivates curiosity at home. Do you have two ChatGPTs? Pull up voice mode whenever his son asks a good question. (1:45:24)

Endurance: I was never faster than anyone else, but I’ll go longer and suffer more. No shortcut. Train the mind. First marathon at 17, two days’ notice, Boston. 150-mile New Zealand stage race with his wife, 8,000 feet up and down per day, half the field DNF. After that, rim-to-rim-to-rim felt easy. Once you have been there, confidence compounds. It is a mental bit-flip more than physiology. (1:48:10)

What I am keeping

1. Change is oxygen. Incumbents freeze. New companies form. Hunt change. Do not wait for it to inbound.
2. Be impatiently patient. Learn aggressively. Transact only on fit.
3. Look at genes, not growth charts. Where is the founder looking? What is the atomic customer loop? Who is pressurizing the product?
4. Wall-clock time was the SaaS moat. AI deleted it. Mid-market software is RadioShack. Survive as Amazon-scale surface area or Palantir-style luxury.
5. Value created is still the ceiling. Software that does work expands the market even as unit construction cost goes toward zero.
6. “I don’t know, I’ll figure it out” beats fake certainty. High-ego drive, low-ego learning.
7. Commitments, not bets. Recruit as if you are on the team. Demand in good times, support in bad ones.
8. Collaboration over courtroom. Make the instinct legible. Win the way you win.
9. Enjoy being Sisyphus. The reset is the job.

Transcript pulled with ytmd. Full episode: Dialectic with Miles Grimshaw.