CaddieOpen in CaddieDownload for Mac

Why Founders Like Jensen Huang Are Irreplaceable | Tobi Lütke, Shopify

Shared by Caddie Library · Updated
David Senra1:09:43Open on YouTube
Chapters show their main points

The bottom line

Founders and makers create the value that drives human thriving, and AI now lets them build far faster and better. So society should celebrate builders rather than vilify them, because fewer people starting companies slows progress for everyone.

Key takeaways

  • Wealth is created when builders turn potential into products. 14:17
    • Lütke owns only 5 to 6% of Shopify. The other 95% is held by the public, employees and early risk-takers.
    • Bezos suggested a list of wealth created for others. Amazon at $2 trillion with 200 billion owned means about 1.8 trillion created for others.
  • A founder's vision cannot be passed on, so the company ends when the founder leaves. 7:15
    • Tim Cook extended the runway of products made under Steve Jobs. Lütke says the market cap should still be credited to those products.
    • Nvidia kept shipping tensor cores that nobody asked for, and they became the basis of AI. Lütke calls that product of vision, not of competition.
  • AI has ended hand-typed programming, and the job is now directing agents with big ambitions. 33:53
    • Lütke says weak results usually come from insufficiently ambitious prompting. Giving the AI the problem, not a preset solution, works better.
    • He rewrote Shopify's internal source repository over a weekend, for about $2,000 of tokens. Merged pull requests are now 10 to 100 times more frequent.
  • Software is about to become nearly bug-free, fast and malleable. 45:44
    • Formal verification lets AI prove a system correct. Asking AI to build a TLA model of a codebase, Lütke says, You will find a thousand bugs.
    • He expects most human-written production code to be replaced within six to twelve months, for security and for speed.
  • Hostility toward builders could make many people opt out of starting companies. 15:15
    • Lütke cites CEO assassinations and polls saying up to 60% on one side of politics find occasional violence against CEOs justified.
    • For now, new business formation is at an all-time high, by his reading of Labor Bureau numbers.

There is no such thing as a succession for a founder.

Tobi Lütke

Chapter summaries

0:00 Why founders like Jensen are irreplaceable

Lütke treats Jensen Huang's exit as the end of Nvidia as it was, whatever the scandal behind it.

  • Lütke's first reaction to the scandal is blunt, and his second is that the loss matters more. 0:02
    • Jensen resigned about four days after affair allegations, which grew into a claim of a Chinese spy.
    • A TikTok then alleged more affairs, up to three. Lütke's first thought: don't have affairs.
    • He had posted that Nvidia ends today and becomes a thing cosplaying NVIDIA.
    • He wrote that it would be run by custodians trying and failing to preserve the flame he ignited.

1:22 Donating blood 3 times to buy a Riva TNT

Nvidia built for a decade before AI needed it
Tensor cores and CUDA came from vision, not demand, and they became the base that AI runs on.

Nvidia's value comes from decades of stubborn building toward a vision, which a board cannot replace.

  • Jensen's products shaped Lütke's own life and the whole industry. 1:22
    • As a teenager he ran a paper route and sold things to classmates, then gave blood at three German hospitals to afford a Riva TNT.
    • Accelerating linear algebra made modern video games, then CUDA made the chips programmable.
    • The hospitals paid donors, and they weren't linked, so he could make the rounds.
    • For ten years nobody understood CUDA, and investors asked why Nvidia was doing it.
    • AlexNet, the first model trained on GPUs, changed that. Today you can't plug a monitor into them, yet they are still called GPUs.
  • Nvidia is a product of pursuing excellence for its own sake, so custodians cannot keep it going. 3:05
    • Jensen made something important unbelievably good before anyone demanded it. Now he acts as a statesman, deciding where the scarce chips go.
    • Lütke blames a board that made a snap decision under scandal. He sees no record of custodians prolonging a company running this hot.
    • The company is worth over $4 trillion that would not exist without him. Lütke says the West is richer for it.
    • Ten or twenty years on, a chart of progress may show a slowdown after this decision. The burden of proof for removing him should be much much much higher.
    • Michael Moritz wrote that no greater distance known to a man exists than the single footfall that separates a CEO from a founder. 6:29

7:15 Tim Cook was still extending Steve's runway

Even the best successor runs a different company
Founder-run (Jobs)Custodial (Cook)
RoleCreates the productsExtends their runway
New productsThe core lineupOne or two
DirectionSets itPreserves it
Market cap creditProducts built under himGrowth came on his watch
Lütke credits Cook's market-cap gains to products developed under Steve Jobs.

Even the best successor runs a different company from the founder's, and a refounding is possible but a long shot.

  • A managerial company can succeed after a founder, but it is a new company. 7:15
    • A succession plan can turn a founder-run company into a custodial one, and some of those work.
    • Lütke calls Cook presumably the best ever at taking over and keeping direction. Only one or two new products came on his watch.
    • A company can sometimes get a refounding event, and there are examples.
    • The founder's company ends the day the founder leaves. Being replaced by something more valuable takes extreme luck, and the odds are low.

10:47 Only founders can reinvent their companies

Reinventing the company is what founders do, and the market keeps misreading it.

  • Meta shows a founder remaking his own company, which an investor wrongly said would change least without him. 10:47
    • Lütke calls that view completely wrong. Meta was mocked for paying researchers, and is now at the AI frontier.
    • Many firms have Meta's resources, and none made that bet.
    • He says people always have it both ways about such moves.
    • He recommends James Burnham's The Managerial Revolution, written before World War II. He thinks it calls founders something like bourgeois entrepreneur.
    • He calls his own account of why founder firms work junk science. He thinks the cause is understudied.
    • Lumping all billionaires together is a category mistake. Builder billionaires add value that otherwise would not exist.

13:14 The 95% of Shopify he doesn't own

Founders keep a sliver; most created wealth goes to others
Shopify5% founder95% everyone else
Amazon10% founder90% everyone else
Lütke owns 5–6% of Shopify, and Bezos's roughly $200B stake leaves about $1.8T of Amazon's value with other people.

Founders start the thing, but the value is shared widely, and that value pays for a fair society.

  • Money enters the world when a product is made, and it pays for society's safety nets. 13:14
    • Shopify exists because of choices he made in 2004 and 22 years of persistence. It is still a team production of thousands.
    • Healthcare, options and safety nets all depend on money created by entrepreneurs turning potential into products.
    • Jensen's teams turn sand and rock into something that thinks. That is why you gladly pay for a ChatGPT subscription.
    • Lütke wants a more balanced view of makers in society.

15:16 Rising violence will make founders opt out

If criticism of founders becomes a safety risk, many people will stop building.

  • Founders can take derision, but danger locks people out. 15:16
    • Entrepreneurs are used to naysayers. Once it becomes a safety issue, many will not start.
    • Lütke links the violence to telling everyone that all problems come from someone else being rich.
    • He answers the focus on his stake: it is a big number because of all the value that was created. 16:05
    • David Senra's reaction to the polling is insane.

16:07 Bezos wants a different rich list than the Forbes 400

A fair society rewards creating wealth before redistributing it, and public giving can inspire more of it.

  • Bezos thinks the Forbes 400 turns wealth into a contest, and wants a list of wealth created instead. 16:07
    • Senra met him for a planned 30 minutes, and they wanted three hours. He told Bezos he built a magic button that brings anything to your door.
    • Lütke agrees, and Senra suggests building the list with AI.
    • Delivery once took two days and is now closer to six hours.
    • Senra joked that people then put a guillotine in front of Bezos's house.
  • Creating wealth comes before redistributing it, or society slides toward zero-sum. 17:30
    • Lütke says few people ask where money comes from, yet it is the most important question.
    • If growth stops, moving a dollar to someone means taking it from someone else.
    • The Middle Ages had no steam power. Energy came only from people and beasts, so growth was near zero.
    • Wealth went to the landed gentry, or people went on conquest.
    • The modern bargain lets creators keep a tiny fraction and share nearly all. GDP growth is that new money.
  • Giving publicly can amplify good, though anonymous giving is purer. 19:27
    • Senra cites an anonymous $50 million gift to a Los Angeles group that helps former gang members. He contrasts Carnegie, who put his name on libraries.
    • Lütke is conflicted. His father-in-law, after years of persuasion, put his name on a large gift for a mental health hospital in Canada.
    • Lütke calls money directed energy. Visible giving makes it high status and draws more money to a cause.
    • Senra praises Michael Dell, who gives now, not at death, through kids' accounts. Elon Musk said he would do similar with SpaceX shares.
    • The robber-baron era traded guilt for libraries. Lütke rejects the guilt: creating products for others is the virtuous act.

24:33 Markets are the original superintelligence

Markets are a collective intelligence that needs competition to work, except where a founder's mission does the driving.

  • Markets coordinate people far beyond what any one mind knows. 24:38
    • Prices change in real time with supply and demand, so supply appears wherever there is demand.
    • Lütke can call a plumber in Toronto and get his house fixed without knowing how. He calls this the "super intelligence" of the city.
    • The oldest clay tablets show traders hauling goods to the next town and tallying who owns what. He calls them entrepreneurs of their time, about 12,000 years ago.
    • Free people make goods for each other and get paid, which lets them make more. He sees this as what happens in any free society.
  • Competition should stay in every field, except where a founder follows a true mission. 25:44
    • He says he is not a full laissez-faire capitalist. He wants monopolies avoided and competition in every space of value.
    • When competition stops it "very very rarely" goes well. The usual exceptions are founder-run companies.
    • Nvidia had no true competition for a long time, yet it shipped a new card revision every year "like clockwork".
    • When custodians take over, motivation is induced from outside. A founder has an "intrinsic burning flame that needs to be fanned".

26:58 Jensen reserved chips for the little guys

Jensen Huang used his control of AI chips to keep supply open to small players, and Lütke suspects his removal was orchestrated.

  • Huang uses his position as the only address for GPUs to keep access broad. 27:14
    • Lütke says Huang reviews every large order and makes sure startups can still get chips, though he could earn much more otherwise.
    • He allows that this may be self-interest, since startups become future customers.
    • He says he has no insider information and is relaying stories he has heard.
    • Nvidia also championed open source AI and invested heavily in it, though many people "really really really do not want" open source AI to take off.
  • Lütke suspects Huang's ouster was a coordinated takedown, and says it shows how fragile a founder's seat is. 28:05
    • He calls the sequence of scandals "too organized", like "how you write it in a book as a takedown story". The question he asks is who benefits.
    • He thinks the Chinese spy claim looks very unlikely on current information.
    • He admits the view "might look dumb in retrospect" and that he could be wrong.
    • What frightens him is how fast someone can be unseated after 33 years of building the most valuable company on Earth.

30:16 The computer programmer job is over

The Web 2.0 era of hand-written code has closed, and its builders now work as makers who direct machines.

  • The era of typing code to build web products was a distinct period, and it has ended. 31:42
    • Lütke says David Heinemeier Hansson made this case in a Rails World talk, framing 2026 as the end.
    • Web 2.0 meant turning an "academic document viewer" into a platform for real software.
    • Basecamp's 2004 release marks the start. Many companies, Shopify included, were built on Hansson's Ruby on Rails.
    • He describes the exchange as honest. Makers gave open source gifts, then built products customers could cancel at any moment.
    • Some practitioners will remain, but he calls hand-tweaking "very fleeting" now.
  • Former programmers now hold a better job, making things by describing what they want. 34:10
    • He says people now talk to computers "in the form of wishes" and think harder about the whole system.
    • He has reached his lifelong goal of being a full-time technologist at a company he considers kickass.
    • He became CEO partly by "cleverly hiring everyone under myself", and mostly to create jobs he wanted for others.
    • He now builds core systems and tools that everyone at Shopify uses.

35:59 Insufficiently ambitious prompting

Hardware speed gains were spent on abstraction layers
Abstractions bought engineer productivity when engineers were scarce, but agents need rules, not layers, so the tower can come down.

Agents change how code should be written, so the main human work becomes specifying and verifying systems.

  • Lütke diagnoses weak prompting by having the AI analyze a person's own prompt history. 36:12
    • He asks the AI to cluster all of someone's prompts by type and give good examples from the start and middle of a project.
    • The AI can do the meta-analysis itself and show where the person presumed a solution.
    • Most of the remaining work becomes what he awkwardly calls "verification engineering", meaning acceptance tests that show the result is good.
  • Agents need rules, not abstractions, so code must be written differently. 39:32
    • Abstractions existed for productivity, because demand for engineers far exceeded supply.
    • That productivity was bought by spending nearly all hardware speed gains on taller stacks of abstraction. He says we don't need it anymore.
    • An 1984 Mac word processor opens instantly, while Word today shows a 30-second splash screen.
    • Human engineers resent strict linter rules as "tools of oppression". Agents love rules and will happily write tens of thousands of lines rather than read 50 files of abstraction.
    • He calls code review a "shitty" human-run form of formal verification, and says there is now too much code to review.
  • Formal verification becomes the most important engineering field because it proves systems correct. 43:26
    • His exercise: ask a tool like Claude Code to find all implicit state machines in a codebase, then model them in TLA.
    • He says this shows human failings, because the code was not as good as everyone thought.
    • He knew formal verification only from academia, and from mathematicians using Lean to formalize proofs. He did not expect it to matter this much.
    • Next, run a security tool such as Deep Sec. Then rebuild on the verified spec in a safer language, such as Rust.

45:45 Software is about to become nearly bug-free

Software is about to get much faster, nearly bug-free and reshapable by its user.

  • Computers will run far faster and bend to what users want. 45:45
    • Speed gains of 100x in many cases, with programs starting instantly again.
    • Even a 10-year-old laptop will do far more than people expect.
    • A user who dislikes something tells the computer, and it rewrites the parts involved.
    • Lütke says all computers are entirely malleable, fitted to your job exactly where you want it.
    • The "computer" may disappear, replaced by full automation in the cloud or an AI you talk to.
    • He adds that all of this is going to happen extremely fast now 46:44.

48:58 AI as an instrument

AI rewards those who learn it deeply, and those people learn together in small, fast groups.

  • AI should be learned like an instrument, not used as a task accelerator. 49:01
    • Anyone can pick up a guitar and pluck a melody, but Hendrix plays the same guitar at another level.
    • A small group can already play AI at that depth.
    • Bring the problem, not the solution, and stay open to answers you never considered.
    • The host compares this to AlphaGo finding moves humans never imagined. Lütke names Move 37 and says that's exactly what I'm doing.
    • Chasing single tasks keeps you in a smaller frame of reference.
  • Shopify engineering now runs on factories of cooperating agents. 50:15
    • Each agent has its own job, loosely mirroring an organization.
    • They coordinate over chat or message boards on long-horizon tasks.
    • Progress comes from better harnesses around the AI, not only from better models.
    • Cybersecurity is one reason to rewrite software. He says human software is just not secure enough.
    • A coming compute crunch is another. Laptop supply chains are tight, so software must use fewer resources.
  • Peer groups at the same level keep him at the frontier. 52:23
    • Groups inside Shopify admit only people who can keep up, and keeping up means contributing.
    • He has worked every weekend this year and says he is made for this moment.
    • His rule for success is to be too good to ignore.
    • The shared playbook covers implicit state machines, linting and formal verification that circumscribe the problem space.
    • People who know these methods outcompete everyone who doesn't, and many share them openly on X and podcasts.

54:51 A weekend rewrite that used to take years

Agent factories turn year-long rewrites into weekend projects
Blacksmith (human coding)Factory (agent coding)
New database layerSpecialist team, a yearA day
Source repo rewriteNot worth attemptingFriday decision, Monday production
Pace of changeWeeks per pull requestContinuous
What limits outputNumber of engineersCompute cost
Once time stops being the constraint, only cost varies, and work that was never worth doing becomes worth doing.

Agent factories turn huge rewrites into weekend projects and make once-impossible work worth doing.

  • A working factory turns a Friday decision into Monday production code. 55:05
    • Lütke rewrote the internal source repository that thousands of Shopify engineers and staff commit into.
    • It was loosely based on his earlier tool Wargit.
    • GitHub fit the pace of humans who worked for weeks before each pull request.
    • He calls the gap with non-users a modern factory versus a medieval blacksmith. Time stops being a concern and only cost varies.
    • He says "entire" is figurative. The point is the speed.
    • A new database layer that simplifies millions of lines, before a Rust rewrite, once needed a specialist team for a year. Now it takes a day.
  • Shopify's internal-tools culture made it ready for AI. 57:53
    • Working on internal tools is high status there, while most firms give them to interns.
    • AI can now autocomplete nearly any internal tool.
    • Internal tools tolerate downtime, unlike production software that needs four nines of uptime.
    • DHH told the host Shopify's internal tools are better than anything he could buy.
    • Shopify builds for itself first so it can play the tools like an instrument, and only then productizes them.
    • Lütke expects a billion people to make things on computers, using any language. He calls the moment absolute magic 59:20.

1:02:23 Why computers shouldn't make you feel bad

Better software turns users into makers, then entrepreneurs
Free markets, not 'capitalism', are the engine, and each sale funds the next iteration.

Better software lets people stop blaming themselves, and that confidence pushes them toward building for free markets.

  • Clumsy software makes people feel incompetent, and fixing it unlocks ambition. 1:02:23
    • Users fighting Word's page breaks blame themselves, not the program.
    • That bad feeling makes them avoid computers for solving problems.
    • Lütke says the fault is always the computer's, never the person's.
    • He sees a chain: better tools help people finish work tasks, then spark off-hours projects, then entrepreneurship.
    • He expects existing software to improve first, with no splash screens and instant responses 1:02:18.
  • Free markets, not "capitalism", are what produce human thriving. 1:04:26
    • Entrepreneurship means offering something good enough to the market. Payment is energy flowing back to the creator, funding the next version.
    • He says "capitalism" mostly names the places where markets stop being free, through monopoly or regulation.
    • Patents are monopolistic and may be worth it only for drugs.
    • The host cites Branson: a business is an idea that makes someone else's life better.
    • GDP growth is really energy growth. Anything that slows the flywheel means less human thriving.
    • Entrepreneurs are just people who picked up skills and make goods and services for each other 1:08:54.

In the growth library

Tobi Lütke, Shopify, founder and CEO of Shopify. Tobi Lütke argues that founder-led companies cannot be replaced by custodians, and that AI is turning software builders into makers who work through factories of agents, formal verification and far more ambitious prompting. The result worth hearing is his own: a weekend rewrite of Shopify's internal code system, work that used to take a multi-year team.

Lessons that draw on this:

More from David Senra · The whole library, and ask it anything

Transcript2105 lines
Why founders like Jensen are irreplaceable 0:00
0:02

Give me your thoughts on the firing of Jensen. >> I mean, um, with the timeline, it's I mean, he he resigned. He resigned, I think, four days after some allegations of him having an affair. I think in short order it evolved into an affair with a Chinese spy. And then there was like a TikTok of um like another person, another person, I think the up to three or so of of of of of alleged affairs. Uh, then a resignation. And so, yeah, what are my thoughts? I mean, yeah, don't have affairs.

0:34

Definitely sort of the first thing that comes to mind. >> This is why I just asked you the question. And I'll just read this to you real quick. These are own words. There's no replacing a founder like Jensen. There's never has been and there can never be a true succession for a founder like him. NVIDIA ends today and it's replaced with a thing cosplaying NVIDIA. It will be run by custodians trying and failing to preserve the flame he ignited. So really the reason I asked the question is it's like your thoughts on why founders like Jensen, and I would say most founders are irreplaceable.

1:00

>> Yeah, and especially Jensen because he's one of one. We see the NVIDIA of now right like a like what is it? Four plus trillion dollar company. NVIDIA's product is that it takes essentially rocks and makes them think. It's the closest thing to magic human humanity has ever come. It is something that is so clearly a product of his un indomitable

Donating blood 3 times to buy a Riva TNT 1:22
1:21

vision. When I was a teenager, I had a made my money I I I sold things to my classmates. I had a paper route and because I was trying to make money for the things that I wanted like usually computer upgrades. Then I fell short. I went to donate blood because the hospitals were paying you in Germany for this. And they didn't they were not integrated so you could make rounds. And I remember donating blood in three different hospitals in a row to make enough money to buy almost magical device what's called a Viva TNT, one of Jensen's products, a 3D accelerator.

1:57

Like this was very very very long time ago. Jensen has been part of this industry and has kind of ignited the magic and the spark of this industry in a very, very way. Like accelerating linear algebra made the video games that we know today. But this dogged pursuit of his, just making this better and better and better, the vision of making them programmable, CUDA. For 10 years no one knew what CUDA really did. It's sort of kind of people didn't get it and yet I I don't know how you preserve something where where you have only naysayers. It's a huge investment. The investors were like why are you doing this?

2:31

And then the dawn of a um you know AlexNet which was I think the first um machine learning model that was trained on these graph graphical accelerated GPUs are not like this a graphical processing units. You can't even plug a monitor into them anymore but they're still called GPUs. So that's how Jensen Jensen has given us the gift of a graphical accelerator, given us the gift of a machine that does something so fast that we can use it to make synthetic intelligence now and God knows what else and all these industrial things. It is a singular company he's built. It's a company around a vision that, if we just take a thing that's really, really important and make it unbelievably good in it even in the presence of no computation, just dogged pursuit of absolute excellence for its own sake but we would reap the benefits from this. And, um, on top of it, he's like, "Now that everyone needs this product,

3:27

this is literally the most valuable product on planet Earth." He's been a like a statesman of presiding over this incredible um responsibility and of of of allocating these these things and where they go and bootstrapping an industry and supporting everyone. It's like it's Jensen is like irreplaceable. It's Jensen's exoskeleton, right? And so um look I I extramarital affairs obviously like marital indiscretions I like these are not good things but like it's like only Jensen can like run NVIDIA. It's it's it's it's an it's a nutty thing. I mean you can put him on administrative for or like figure out something But he built this company.

4:04

Like the company is worth what? Again, I think I remember it's over $4 trillion. Um, and it's like, you know, that's all value on planet Earth that doesn't didn't exist without him seeing this, without him building this. It's like you like everyone is living in a is a good deal richer in the West because Nvidia exists that creates money out of nothing and creates these products out of the potentiality of the ether, and then more value exists in the world and then you can build products on top of them which are even more value entering the world. And this is like trillions and trillions and trillions of dollars. And if this stops now because they're turning Nvidia into some kind of legacy company because the knuckleheads on his board just kind of like seemingly have made a snap decision in the face of scandal or whatever, like the eye of the storm, it's crazy to lose this. And I just I just think that there's kind of

4:58

no track record of custodians truly managing to prolong something that's running as red hot as like as Nvidia. I think this is of societal impact. Like progress on planet Earth at some point, if you can measure it, like 10 years from now, 20 years from now, you might look at the chart and you could potentially mark the moment where sort of after Jensen's like after the current set of decisions have played that played themselves out in the market, there will be a deceleration because of this, because like again, it's an unbelievably hard thing to run a high-functioning large company that's doing important things. And there's really not many people who can do it. I don't think there's really anyone who can can do it, definitely for Nvidia as well as so like well, we need to preserve people in the in in this world as long as we want them. And we also should celebrate people for what they truly accomplished. And who knows what

5:49

the real story is and what all the components are to it. But like the burden of proof needs to be much much much higher for saying that he she he's not the right person for running this company. >> Yeah, I think what we were talking about before we started recording was both of us I like the last time we had this conversation, and oh, we actually had the same job. Both you and I are trying to create more entrepreneurs. And maybe I'm like so like, you know, I'm so deep in the ocean, I don't understand that like like the water is wet.

6:15

But like the idea was like unfathomable to me that regardless that they would When people were like, oh, you know, Jensen's going to lose his job, people were speculating, I was like, this is impossible. Like the founder is so important because I just we have a deep understanding of how important they are. I just want to read something to you because I think Michael Moritz he wrote this book on Steve Jobs called The Little Kingdom. He publishes it.

6:36

Steve Jobs is still there. I think Steve Jobs is 30 years old. In 2009, Michael Moritz writes an updated version called The Return to the Little Kingdom and in the forward he has one of the best lines cuz when I was reading your tweets about Jensen, this is what popped to my mind. And he talks about the difference between the founder and a professional CEO. He says there's no greater distance known to a man than a single footfall that separates a CEO from a founder.

6:58

CEOs are for the most part products of educational and institutional breeding. Founders, at least the very best of them, are unstoppable, irrepressible forces of nature. And I think your perspective and what you were treating on Jensen is like he is a force of nature. There is no repeating. There's

Tim Cook was still extending Steve's runway 7:15
7:14

no replacing him. >> There is no such thing as a succession for a founder. You can have a succession plan as a in a structured sense of moving a founder run company and transitioning into a custodial run company and a managerial company if if you will. And there's obviously success stories in the managerial world too, right? Like it's I mean Tim Cook did a very very good job extending the runway of a products that were created, like keeping like building them on the clock and so on.

7:45

I think there was even one or two new products under his leadership. So something kept working. He's presumably the best in the world that's ever existed as in terms of taking over from a founder and actually keeping the direction. Now a lot of the market cap increase was during his time and he should he should get the credit for it, but it's like you have to allocate the market cap to the products. And the products are all developed under under under Steve.

8:14

So it can't it can be done in this way. There's even a case we made that it is possible to potentially give a company a refounding event. There's examples of this too, but like really in all situations you you you're having a transition from one thing to another. And the thing the company that was under the founder ends the day the founder is no longer no longer in charge. And maybe if you're lucky, it's replaced with something else valuable.

8:39

If you're extremely lucky, maybe it's going to be replaced with something that might even be more valuable. But like I think your the betting odds are just really really low there. >> One of the things I love about Ramp is their mission is easy to understand. They want to help your business save time and money. And that is why they have built Router. Do me a favor while you're listening to this, go to router.com. Router was built to reduce your company's inference costs by matching every request to the most efficient AI model. Router automatically balances cost, speed, and quality of output. Router is built for CTOs and loved by CFOs. Your engineering team gets the best model for every workload, and your finance team gets lower inference spend. One customer of Router has the perfect description of why your company should be using Router. Choosing the right model makes a meaningful difference to our AI spend. We run

9:32

billions of tokens through Router and have reduced our model costs by 92%. Tokens are money. Save both. And you can do that by going to router.com. One of my all-time favorite quotes is from the book Zero to One. And it says the single most powerful pattern I've noticed is that successful people find value in unexpected places, and they do this by thinking about business from first principles instead of formulas. That is exactly what AppLovin has done with their advertising platform.

10:00

AppLovin connects you with over a billion potential new customers in mobile games. AppLovin allows you to capture undivided attention. AppLovin ads are full-screen videos that are watched for an average of 35 seconds. That is retention that blows other ad platforms out of the water. And you can launch on billion potential customers. Other businesses have seen immediate results, scaled to hundreds of thousands of dollars of spend per day, and increased their revenue by millions. So you want to get started quickly before all of your competitors are on AppLovin. And you can do that by going to applovin.com. That's applovin.com.

Only founders can reinvent their companies 10:47
10:47

>> Why do you think this is so widely misunderstood, the importance of the founder? Like let me give you an example. A friend of mine was talking to somebody, and this person said out of all the technology companies that are still founder-led, the one that would be the least different if the founder left was Meta. And I go, what? And I go, who's this person? And they're like, that this person's an investor. I go, no, that person's an idiot.

11:07

>> Yeah, no, that is completely wrong. People always have it both ways, like you know, just like a year ago or whatever, maybe a year and a half ago too, maybe Max, like everyone's complaining about suddenly Meta is hiring all these researchers for a lot of money, and that doesn't make sense. And now they're like have frontier-level like Meta is an AI lab at full Pareto frontier. Like like try that. Like there's a lot of companies in the world now who have the resources of Meta, and none of them have made that decision. So like it's like giving like reinventing your own company as an as an aspect of founder-run companies, and like you you can point at basically anyone for founder-run company. Now, I can't go in a bit of a my junk science take on why this is, but like, it depends on where you want to go in this conversation, really; because I think it is under ,like understudied, what what,

12:01

what is the, the operating bit that's causing founder-run companies to, to, to be so effective. But it's not a new topic. Like there's the I think the best books on this topic are written I mean probably the best book straight on this topic I've read is written by um John Burnham called the Managerial Revolution. I think he this before World War II. They didn't wrote founders. I think he called them the bourgeois entrepreneur or something like this, you know. So um but but but but but it it nails the the the thesis and it's under discussed and I think it's causing people to uh make a lot of principles of category mistakes and like billionaires that universally hated and people are allowed to hate whoever they want but like it is really really crazy to not make a difference here between builder billionaires and others. You know, again the people who um build companies, people who build products or there's at

13:03

least one billionaire who wrote books, they are adding value to the world in a way that like it it would simply not exist otherwise. Like someone has to start the book and someone has to write it. Someone has to

The 95% of Shopify he doesn't own 13:14
13:14

start a company. I started Shopify. It exists because of a couple of decisions I made in 2004 and then the decision to keep at it for 22 years. I'm obviously not responsible for all the value of Shopify. This is a team production. We have thousands of people work here and I own very small percentage of the company is like 5 6% at this point. So like 95% is owned by you know the public, by employees, by early people who took bets and you take took huge risks.

13:44

This is how money is created. We want great societies. We want fair societies. We want um people to not suffer, people to have healthcare, people to have options, people to have safety nets. And so you know all all these kind of things are created from an energy in the form of money that is created by the entrepreneurial spark mostly of taking potentiality like potential that's out there and converting this into products and services that we then trade for each other and the money comes into the world at the moment of product creation.

14:19

Jensen and his teams are taking sand and rock and making it think and be able to answer your entrepreneurial questions that you secretly harbor in the quest of one day starting your own company. That is a thing that only is possible and is extremely valuable to you and therefore you are happy to pay for the ChatGPT subscription because of all the work by the builders. I think it's hugely important to have a more balanced view of the importance of makers in society because again, it's a pretty tough job by itself and I think most founders can deal with the sort of constant division. Founders tend to be and entrepreneurs tend to be used to lots of naysayers along the journey. You have to overcome a lot of people telling you you shouldn't be doing this. So there's some sort of school of hard knocks there that everyone's gone through and that's okay.

15:15

The moment it becomes a safety issue,

Rising violence will make founders opt out 15:16
15:17

right, which it increasingly becomes, that will just mean a lot of people locked out. So this is >> mean safety issue? >> Well, I mean we've seen assassinations of CEOs now. We've seen there's a lot of violence. There's polls which say that occasional violence against CEOs can be justified, which I think up to 60% of people on on one side of the political spectrum agree with. >> Come on. >> And you know >> [ __ ] insane.

15:45

>> I mean, this is also a consequence of telling everyone that all problems come from someone else being being rich. >> I think your framing is really smart about okay, everybody's focused on the fact that you own 5% of of Shopify and they can do the numbers and it seems like a big number, but you're like, what about the other 95%? So >> It's a big number because of all the value that was created.

Bezos wants a different rich list than the Forbes 400 16:07
16:07

>> I got to spend time with Bezos. He's one of my favorite entrepreneurs. We're obviously one of the greatest [ __ ] crazy. The complexity and the performance of the the company and everything else. >> have it. >> And so we were supposed to meet for 30 minutes because he listens to my other podcast founders. And I think he's going to do this show. But um we were supposed to meet for 30 minutes. We wanted to talk for 3 hours. And one of the things I told him, which was the most insane thing ever, because his point was just like uh I was like, dude, you made a magic He's done a lot of stuff, but I go, you made a magic button I can press and anything come Anything I want in the world comes to my house in 2 days. And now it's like 6 hours or it's even crazier. And I go, and they put a [ __ ] guillotine in front of your house.

16:46

>> Yeah. Yeah, right, right, right. Exactly. >> Did you make a magic button that I could press and anything Like that's an insane thing. >> Yeah. >> And one of the ideas he had was he thinks that the Forbes 400 list is actually like really detrimental to society because it's like we're like, this person has 200 or 100 or 50 and it's like makes us some kind of competition. He goes, there should be how much there should be a list of how much wealth the entrepreneur has created. He goes He goes I I think at the time I I forgot to mark cap. Like we just make this up.

17:10

Amazon's worth $2 trillion. I own 200 billion of that 2 trillion, which means I created one or I helped create 1.8 trillion for others. >> Yes. >> The idea that we haven't built I love your like we can just put things into existence with AI now. We should just build this together and just have a [ __ ] list and then we'll just put it everywhere about the wealth created by other people. >> I completely agree and it's just like it's it's it's again, like where does the money come from? It seems a question that a lot of people just are not curious about, but it is the most important question because only when you understand it do you understand who are on like who are people on the side of human thriving truly, right? Like it's like don't get me wrong, but redistribution of uh uh money is also an incredibly important topic, but clearly it's a secondary topic to creating it because like first you need to make

17:56

something and then you can figure out who needs help, right? And so I a lot of parts of the world in the West, we are no longer concerning ourselves so much with the creation of money and a lot more with redistribution. And the problem is if this continues, at some point, you know, the growth stops, at which point you only have redistribution. The only way to then get $1 from one side where you think it shouldn't exist to a place you think it's needed even if you're right about this, you have to take something away from others and this is where things get really really really bad because that's basically the situation that the Middle Ages were in that's where there was no growth that we didn't have steam power the only way to bring more kinetic energy into the world was by someone being born or by like a human or a beast of burden and that was very slow so the growth was basically nothing

18:46

therefore only thing that existed was rival with competition for resources and therefore it all went to the land of gentry or you you went on conquest for for resources and that's a bad world like this is a one world we rejected and replaced with a with a vastly better world where we allow people with the vision to create things to to to to make goods make services for each other which then brings more money into onto planet earth into the economy that is what GDP growth is then allow them to like like or incentivize this kind of behavior by allowing people to own a very small fraction while sharing almost all of it with civic society and so this is a beautiful system.

19:27

>> There's something that popped to my mind, and I actually think it's a problem that I didn't think of until we're we're talking about this now. Plenty of examples of well-known entrepreneurs that are giving away money quietly and anonymously; they do not... they, they're like, "I want to do this, but I don't want any credit." Back there's an organization I think I, I forgot this like three or four days ago, I was in downtown LA. There's an organization that is helping rehabilitate gang members; a lot of them came out of prison, a lot of them are covered in tattoos. They're doing tattoo removal, they're getting them jobs, they're getting them education, and it's completely one off donations. And a world-famous wealth creator, entrepreneur that built products donated, I forgot the, it was like 50 million but they insisted it was anonymous.

20:11

The problem with that, now listening to you speak, is just like: okay, so it's public how much your stock is worth, and at the same time, all these entrepreneurs that the wealth is public, they're doing all this great stuff for, for society and giving back, but they insist it's anonymous, We can go back to history. You and I both share uh a love of history. It's like, well, what did Carnegie do? Carnegie did like he spent a ton of his wealth building I think a couple hundred of the libraries at the time, obviously no internet, like in communities that made a difference. And he insisted on starting in communities where he got his resources from, like a lot of the the the natural resources you need to make steel. And as a result, like he made society better, but then he would made it known that it was his. Now you could say he did that for [ __ ] ego and everything else, but I think that is

20:52

actually important. Doing anonymous donations to me is like the highest form of giving, but it's also like there especially when we're all connected, it's like you almost have to kind of let people know what you're doing. Think about one of the most beloved, you know, multi multi-billionaires in the country right now, I would say would be Michael Dell. You're not going to find a lot of people that hate Michael Dell. He's one of my personal heroes. I I love him. He's been so supportive of me. His books are amazing. The way he thinks, he's just an incredible person.

21:22

I could spend a ton of time with him. And I think what he's doing is one of the the the best things I've ever seen. It's like, I'm not going to wait till I die to give the money away. I'm going to do it now and I'm going to do it for the kids of America. >> Yeah. Yeah, these um kids' accounts were I think one of the most important sort of breakthrough piece of infrastructure that has been built in recent times, right? Like because there's I think a way to like like do this now, give money to all the kids. I think there's some talk about uh depositing shares uh of companies into kids' accounts.

21:55

>> is doing this for shares of SpaceX I think Elon said he's doing the same thing. >> This is a wonderful form of like a like of giving, which is I think it's very novel. It gets around a lot of the problems that that otherwise exist with with giving. I agree. So like I'm I'm conflicted on the on an anonymous donation thing. I'm partial to it, but like it is somewhat selfish counter counter intuitively. The largest charitable donations in Canadian history, I I'm proud to say, was done by uh a a member of mine. Well, right like Bruce McKen, who's like financed and and company. My my my father-in-law did um a very, very large, enormous donation. And we convinced him over over years to eventually own it.

22:38

Like it built I think like mental health research hospital in Canada. The reason why it was important to own it is because if you do it anonymously, it is incredibly like it it's sort of unambiguously morally pure for sure. However, money is a form of directed energy. It comes with a lot more than just the monetary value. I think you want to make it a high status thing to do this type of donation like this well researched. This is an cause that's important to me and funnel it in your direction because you give signal to everyone else. First of all, [snorts] others people can see themselves in this act of putting money into a cause that they like really that that's really important to them.

23:21

And then also you're just like creating an amplification effect. Like there's more money flowing into this cause now because this is >> So Michael Downe, Brad Gerschner started this and now everybody's coming in. They're like it's a great idea. I didn't think of that before. I have resources. Why don't I do that too? >> I don't even think it's necessarily true in this way, but like mainstream understanding of a robber baron times was that there's a lot of labor exploitation. So you would basically accumulate guilt along the way of creating wealth.

23:50

And then you would sort of make it okay by building libraries or so. That that's that is a template. You know, like that was sort of a compact that existed until I think recent times. I think the sort of last person to go for the compact was, I would say, Bill Gates is sort of in for this, right? Like at least the narrative arc. Again, I don't really believe in the guilt part of this because again, I think the real virtuous thing that you can do on Cladiv. My own philosophy is like [snorts] what is more virtuous to do than create products and services for your fellow humans? Right? Like what that already is the thing that you're supposed to be doing, right? Like

Markets are the original superintelligence 24:33
24:33

the money component is like humans make markets. Markets are the original form of super intelligence. Like we are all part of markets that are vastly more intelligent than us. They are in real time change the prices of everything based on supply and demand. If there's demand for something it'll like supply will exist. I can in my city I can just call a plumber and fix everything in my house if that needs fixing. I have no idea how to do this and I don't need to face up the reality of being completely incompetent in these matters because the super intelligence of the city of Toronto allows me to just like solve this problem on demand for myself. Like markets exist everywhere.

25:11

The oldest writings we find like tablets, clay tablets things there a bunch of people go to the other town over with with a cart full of goods and they tally who owns what and so that they can track who will make what profit. Those were entrepreneurs of their times. This is like 12,000 years ago or something and we are creating markets. Free humans create markets amongst each other. They make goods for each other. They make compensation for this which then allows them to do more of it and this is just what happens in a free society, right? I'm not a full laissez-faire capitalist. I think you need to avoid monopolies. You need to ensure that every space of value has competition in it. If competition stops it very very very very rarely goes goes well.

25:59

The examples of where it went well are usually the founder run companies, right? Like I think like Nvidia had no true competition for a very very long time yet you would never notice because like clockwork every every year there was a new revision of the cards with new features. Suddenly we had tensor cores that we had no idea why anyone would want them because we used them to make I don't know Call of Duty run better, right? Like so why why do we need tensor cores?

26:24

But those are tensor cores that were the entire reason we needed them like in the end because that's what you need for the for for AI acceleration. That was not a product of fierce competition, that was product of vision. So, in a market or in a space where you have people following true mission, you don't need to rely on markets and competition to get the best outcomes. In all other cases you do. The moment the custodians take over, the motivation will be induced externally rather than being an intrinsic burning flame that needs to be fanned to glow brighter and brighter.

Jensen reserved chips for the little guys 26:58
26:58

>> To tie that idea to what you referenced earlier that you viewed Jensen as like the statesman of the entire AI industry. >> Again, getting these GPUs, without them you can't do AI training. There's no other address you can call. You need a lot of them and you place a large order. Jensen will review every one of those orders. I mean, he makes plenty money, so like he's fine but like at many times he could make a lot more money. Again, and I I have no insider information here because you know, the stories I hear or what I'm being told, he always makes sure that there's availability for this the smaller guys, the the the the startups. Presumably he wants more customers in the future, so there's some self-interest in this, but he doesn't have to do it. He made sure that these things were available to everyone. He championed open source AI, which has a very high likelihood of being involved somehow in these events

27:49

because there's a lot of people who really really really do not want open source AI to to take off and Nvidia was heavily investing in them. I should say I I should say. So like again, we'll get more information. This might look dumb in retrospect and I might be completely wrong, but this looked like knob. Let's be honest. This was like surgical affair, Chinese spy, another affair, another affair. Now, of course, they did potentially have those affairs. Maybe there's even a Chinese spy component that was like currently just looks extremely unlikely that that's a real thing from the information available right now, but this is knob, right? This is like cryptography. It's too organized. This is exactly how you write it in a book as a takedown story.

28:31

The question is who's benefiting from Jensen not having this job and there's a pretty large list and it depends on there's multiple motives that could exist. So, I think we're we need to understand the real story. But what's frightening is how quickly you can unseat someone who started a company for 33 years and turn it into the most valuable company on planet Earth, doing an incredibly good job running running it and, again, having to make enormously important decisions, you know, about locations and roadmap and, like, of, of of, you know, the currently most important product on planet Earth, really. So [snorts] yeah, there's clearly more to the story.

29:09

>> The best founders are relentless about saving time and only focusing on what matters most. The founder of FedEx, Fred Smith, had this great line that he would repeat. He would say, you have to be absolutely brutal in the management of your time. That is why many of the best founders in the world are using dot in ChatGPT. Dot is OpenAI's AI agent that works for you continuously even when you're not actively using ChatGPT. Dot works across your apps, can execute multi-step tasks and help you stay organized. You give it a responsibility and it starts working, giving you more time to focus on the work that matters most. One way that I use my dot is to help me research upcoming guests. It reads my email and keeps an eye on my calendar so it knows who my upcoming guests are. It then finds relevant information, interviews, articles, and their top posts on X. 24 hours before we record, my dot sends me a document with

29:59

detailed notes and links to everything. What used to take you hours, dot can do in minutes. Save more time and keep your business moving with dot. Dot is included with ChatGPT Pro, business, and enterprise plans. Again, that is chatgpt.com/get-dot.

The computer programmer job is over 30:16
30:16

>> All right, so let's transition to what you were saying earlier. I think I don't even think I think we were talking about this before we started recording. I had your close friend DJH on the show. He's just like a bottle of energy. I think you just like tee him up and he just goes. [laughter] I think that's probably the the the episode I spoke the least in cuz I was just like I was also been a fan of his writing for like 15 years. So I was like [ __ ] it's kind of cool now that like you know, we had dinner the night before, we're hanging out at his house.

30:41

Like it's it's really cool to get to like meet the people you admire. And you know, his a lot of his writing shaped my thinking as well about like how to build a business. You said something interesting where you're like well uh and I watched his his speech about him no no longer being a programmer and he puts like the the day like 2026 is when it all ended. So you said something fascinating you were like well >> is a like a talk he did at um Rails World um conference like of a again technology started a lot of us used to build our companies shop included.

31:11

I recommend even for non-programmers to watch at least the these clips because I think David did an enormously good job here like framing something that's really really important and the entire audience there was there to learn about how um what's next in this technology and David basically said this is no longer a job. >> We'll link it down in the show notes, but explain how you said web to web 2.0 because I love how you tie everything back like you always looking back to figure out what's going happening forward too. So web 2.0 as of 2026 is dead. That idea, can you just explain that in totality?

31:44

>> You'll hear me say a lot of uh similar things to what David I think articulated better than I will be able to. These were things we've been talking we we we we have a hobby together doing endurance racing and these are things we've been philosophizing. I mean we've been part of one of the greatest creative periods in human history with the jobs that allowed us to be not just participants but also make a dent into them like really sort of early 2000s. It was called web 2.0 um uh then it's it's not a term that's important but like it's important term that's important to me.

32:17

It was a lot of us busy raising our hands and saying hey we have figured out how to take this like academic document viewer called web browser and use it to make software. It's sort of very different from the way we making desktop software in all these important ways because there's some restrictions in the medium and of course like a client server architecture and so on. But like we figured out how to make good web applications. David's best known product is Basecamp and I think really could be associated with the true starting point at its release in 200 2004. So [snorts] the browser is now good enough.

32:56

We figured out some sense of mastery over the technologies and we know how to put it together to for for for making phenomenal products and then we did like enormous amounts of the technology industry are companies that many used David's technology Ruby on Rails again including Shopify. So we we built these companies again we made gifts for each other in the open source world and then we took vows to make gifts for our customers that we charged some money for and then our customers they if we did a great job then through the wonders of the market they subscribed. In most cases they could cancel at any moment. This is true for Shopify. So it's a very very honest relationship of value exchange. You know this is this industry of web tool and we are in the AI age and at least the jobs that we had were computer programmer which I associate with you know typing characters into a terminal or editor to to to program

33:51

computers. That doesn't exist anymore. I mean there might be practitioners still and there might be like some situations where you really really really want to tweak exactly everything but like that's like very fleeting now. So really there is a period where this existed. It's over now. And everyone who used to be a computer programmer has started a new job now. And and I think in in almost everywhere this is going to be a better job. It it's it's like it's it's we are now just creators of things. We are makers. We are talking to our computers in the form of wishes. We are thinking much harder about the system we want instead of just the next step to getting somewhere that's like good enough. We are thinking about what could be. We like this vastly more ambition.

34:41

There's vastly more back and forth. I can go in all sorts of depth here on what this job now is. >> Explain what you mean by there's now more depth. >> I truly live in this world now. I have since we last talked, I have realized my life's ambition of becoming a full-time technologist, creator of products at a kickass company. That's what I always wanted. This is like started company because uh I couldn't find a company that I would call a kickass company. I felt I had to create it. I became the CEO partly by cleverly hiring everyone under myself and to preserve the flame of a great company because that's easy to screw up. And but like mostly I created the jobs I really really wanted for other people.

35:26

And now I can at the same time while doing my job also be someone who makes things. So I'm making I think pretty cool. I hope everyone else agrees that the company contributions to like Shopify core systems, I built products, I built tools and technologies that thousands of people use in the company. In fact, everyone uses in the company. That is remarkable. My main advice is when people are not quite able to get all the value out of this AI driven development, it's like usually

Insufficiently ambitious prompting 35:59
35:59

insufficiently ambitious prompting. It's really phenomenally interesting to sit down with people and saying and sometimes funnily enough I just sit down with them and asking them how they prompt. I'm like I ask them, hey, go through all the history of prompting. Like I I literally go to their AI and we say go through all the prompts and just sort of cluster them in types of prompts and like give us good examples of prompting and especially start, beginning, mid of product and like just like analyze it. And and like you know, the AI can actually do the meta analysis itself. And then we look for this and just saying, look, you had some preconceived notions about how to solve the problem which you told it to like investigate, which is good, but you didn't understand the problem. So give it a chance to actually say there might be much better ways to solve it, or maybe there's a you know, maybe this isn't the problem at all if

36:53

we change something else over here. So they often sort of working with machine is actually in a way trying to inspire it to understand like the the problem you're trying to solve. And then most of the work ends up becoming it's not a good term, but for lack of a better one we call it call it verification engineering, which is like creating acceptance tests. Like here's how you know for sure that the thing that exists, the thing that was created is good. Again, I'm not going to bore you with the technologies going into this, and also I think all of us are going to change because we're sort of working with what we've got right now.

37:29

But this will be a huge part of our industry now where we make sure that we use automated systems to know that like a system is known to be good. We we had unit tests before, we had static typing before, which are sort of things at least static typing is sort of formal verification, but the field of formal verification is now by far the most important field in engineering. Look, I've lived in this industry. I pride myself on knowing almost everything that's going on to the degree that it that that von Neumann can. Formal verification was a thing I had I knew it existed in academia and knew mathematics was using things like Lean to to formalize proofs, but I probably knew this from AI research. And like I had no idea that this was going to be the most important part of engineering in the future and now it is. And so this is how quickly everything is changing.

38:21

>> why it's going to be the most important part of engineering? >> Because you can prove that some a system is correct. This is and this is extremely important for complex systems. Like I I just like let me say something about what I've learned here in about agentically engineering. And this is specifically when you're rebuilding systems with agents. And tying this back to David and Ruby and why David proclaimed pencils are down, we are no longer hand chiseling code, we are now working with agents.

38:48

We didn't have enough engineers in in a world or even close. Probably 0.1% of the amount of the there was demand for. Like it's just too few people went into this. Maybe it's an aptitude thing, but like it's just I think for whatever reason there was just like way more demand than supply of engineers because there was so much demand for writing code. Therefore, the most important thing that most of us could work on and this is why you hear so much about frameworks. Like Shopify is built on top of Ruby on Rails. I said this like three times already. That's a really really important thing. What is Ruby on Rails?

39:21

Ruby on Rails is a again a framework that just sort of like constraints the possibility space and just sort of says here's a known good way of building a web application. It's full of abstractions and abstractions are sort of pretension. Like you pretend that something is simpler than it is. You can just like make a you you you can talk to the database by just making a new object and telling it to find find by name one and like it it does all this. It does all the hard work of plumbing these things together and dealing with wire protocols and the complexities of retry logic and busy databases and like all these things are kind of pretended to not exist in an abstraction. So what are abstractions?

40:02

Our entire industry was like built abstractions. Why? Because abstractions cause productivity. They actually unfortunately have a large negative consequences, but none of this mattered because we needed productivity. Productivity over everything. We needed to get one engineer to be able to write 10x, 100x the code of an engineer wouldn't use these abstractions. So productivity was absolutely everything and we invested one could say wasted almost all advances of computers getting faster simply for productivity gains. Like we we made higher and higher higher stacks of abstractions. This is why when you find an really old computer and start it on like like started like an old Mac or so it just boots in a no time and you start an application, it's there and it's like this thing is from like trailers of a Macintosh. Like it's this thing is from 84 and like the word processor opens instantaneously. Why does

41:01

today when I start Word have a splash screen that takes 30 seconds, right? >> The agent is doing this. I love that the he's like I bought a new Apple computer. It took 40 minutes and then his new his new Linux operating system >> Yes. Yeah, it like it you can install the entire thing faster than you can probably load Word, right? Like and so why is this? Because we were extremely wasteful because we had to be because we need productivity. Okay, all this to say we don't need productivity anymore. So this is crazy, right? Like so working with agent is actually in so many ways the opposite.

41:32

No engineer wants to work in a codebase with insanely strict rules. These are called linter rules, right? Like rules which say yes, you wrote valid code. It might even work, but it's not allowed around here for reasons of that are important to the codebase. So valid code becomes errors because every company has their own way of doing things and bans certain ways and allows certain ways. These things are seen as tools of oppression. You have to like not go too far, otherwise no engineer wants to work on your codebase anymore because it's like oppressive. The agents love them because they give them rules.

42:08

They love rules, but not abstractions, because they are perfectly happy to write tens of thousands of lines of code that are correct for the thing they're doing, and they don't want to go for 50 files and read all the layers and layers and layers of that that the abstract tower of Babel has built up to understand what the actual consequences are. Okay, that's important. So we need to write code entirely different now with agents.

42:37

How do we know that it's good? Because now we are producing too much code to review because another aspect of building abstractions is the end result of code is not that much. Therefore, we can have humans reread this in the form of a code review to make sure that it's okay. Code review is basically our human run formal verification shitty version of it. All this is changing now. So formal verification now it's actually specified to you basically say this system has these many states like that can exist. Here's like a model of [snorts] the state machine that runs the Shopify checkout or something. And now please verify that every single possible state the system can get into is actually thought through and is tested and is specified.

43:21

What you will invariably find is within if you ask your AI to like and everyone can try this. Everyone who is like any codebase can try this. Go into something like Cloud Code and just ask your AI find all implicit state machines in this codebase. You say get me all the state machines to be that are implicit currently and sort of ad hoc in the system and maybe explicit. First of all, you will learn so much about human failings in writing code. You you will immediately realize you were not as good as writing code as we all thought because the machine will make everything better the moment you tell it. And then afterwards you say now make it's called a TLA model. It's like it'll change 15 minutes of this podcast is released about how we do these things.

44:06

But right now you ask it now make a TLA model for it to make sure that all the cases are specified. You will find a thousand bugs in your codebase, any code that we've ever pounded it at. Then afterwards you go and say okay, by the way use some tool the cell made a great tool called Deep Sec. Use Deep Sec to look for security vulnerabilities in my code. And then you will definitely not think that humans should be writing any code anymore after that's done. We are currently in the midst of an unbelievable awakening to of of cyber security, which is like, I think most companies are in some form of code red, just like trying to fix everything, and we probably have to replace all human code within the next probably six or 12 months that are running in production on planet earth.

44:56

So the way to do it is either by improving existing code bases to be sound through prompts and things like I mentioned, or by building verification systems that make it so that you can really specify what the actual system truly should be, that you make sure that it's all considered, which is something you can collaborate with great AI on. And then, against these formal verification methods maybe unit tests, maybe production traffic, whatever you have you will rebuild a completely net new code base in a safer language than the one you might have used before, like Rust or something like this. that's will be the code of the future.

45:39

Many many consequences in this like that are side effects of this kind of activity. I think software will

Software is about to become nearly bug-free 45:45
45:45

basically be near bug free. It will be 100 x faster in many instances they'll definitely get back to the world where things just start instantaneously and we will be amazed by how much you can do on with a run of a mill or like even with a 10 year old laptop out of a sudden. The quality of our software is going to just go through the roof. Computers will do the things that their users want. If computers don't do the things that their users want you will tell your computer that you would like it to be able to do something different and it will rewrite the parts that involved in the thing that you didn't like before. We will live in a world of entirely they they all computers are entirely malleable in a sense that it's going to be the computer that supports you and your job perfectly in exactly where you want potentially not a computer at all potentially just fully automated in the cloud potentially

46:38

behind an AI that you talk to. Those are options for you for everyone and all of this is going to happen extremely fast now. >> Deal is how the best founders turn the world into their talent pool. I've been studying how history's greatest founders operate for a decade and one thing they all have in common is they understand that and hiring the very best talent is your most important priority. A players recognize other A players, which is why top companies like Ramp, Shopify, 11 Labs, Uber, and DoorDash all use Deal.

47:08

Many of the top founders I know have personally invested in Deal after using their product. And what they discovered is that Deal is the best company in the world at building infrastructure for global hiring. Deal will help your business hire, pay, and manage any worker anywhere in the world so you can retain the best talent anywhere and spend the rest of your time focusing on what you do best, delivering value to your customers. The founder of 11 Labs has a great description of the value Deal can give your company. He said, "We built 11 Labs to break down language and communication barriers. With Deal enabling us to hire and support exceptional talent anywhere, we can accelerate our innovation and bring more voices, stories, and ideas to every corner of the world." Deal is trusted by over 40,000 companies and growing fast.

47:52

Learn how they can help your business by going to deal.com/ senra. That is deal.com/ senra. >> Okay, I'm trying to understand everything that you just said and I it's very fast. Thank you for doing that by the way. That was incredible. It's but it started from this point where you're like, well, big issue here is you're insufficiently ambitious prompting, right? And then, the more you spoke from that starting point, I almost feel like you're describing the difference between human intelligence and artificial intelligence. And when we went through these prompts, you're assuming too much that you knew you actually identified the right problem, or that you knew maybe you had an inkling how to solve it. There's a weird thing that happened in my mind when I'm listening to you speak. You that that documentary they did on Demis and Deep Mind, right?

48:37

And they talked about like eventually they just studied every other at the very beginning they studied every move that humans could possibly do and go. And then eventually [clears throat] that only got them so far. But what it got them even further is the machine coming up with things that humans would never even >> Yeah. >> comprehend or even think of. >> Move 37? >> Something like that. Is you're describing a very similar thing in programming. Yes or no?

AI as an instrument 48:58
48:58

>> that's exactly what I'm doing. >> AI certainly is an excellent current task accelerator, and I think this is how it's usually used. I think AI is best thought of as an instrument. If you first pick off pick up a guitar, maybe on one string you can maybe even figure out how to play a melody or something like this, but like you don't really know how to produce music. It takes a while, right? Like and of course, but there's no difference between truly the guitar you can pick up just like this and what like I don't know, Hendrix plays. At some point you understand like you can play an instrument truly because you understand it at a at a level.

49:32

It's a small group of people, but there are people in the world who who can play AI at this level, um, who are really understanding the tools, even though it's a very fast-moving base. And I think I'm reporting back a little bit from what the what the world looks like from that perspective in that, instead of accelerating your tasks, learn the instrument; learn how to, um, work with AI in such a way that it, like you, you provide it with the problems you want to solve. Show up with [snorts] an openness for a wide variety of solutions because it turns out there often are things that you might never have thought. With the tools becoming more sophisticated, it's not just the AI is turning sophisticated, it's also what's called the harnesses that are using the AI is becoming more sophisticated.

50:15

Almost all engineering at Shopify now is happening through these things called factories which uh uh like constellations of AI agents with different jobs kind of loosely mirroring what organizations um were like who collaborate with each other for chat or for message boards to accomplish much more long horizon tasks. It's much better to be more ambitious with AI and have it do reps on solving the problem than it is to, uh, accomplish tasks; because I think accomplishing just the tasks keeps you in a sort of smaller frame of reference, which I think will be very hard to get to the right places, at least in the engineering world, fast enough. Again because, and this is the other part between cyber security number one, and then also, um, a resource crunch that's probably going to happen with compute and in the next year, there's really good reason why a lot of software probably has to be rewritten. We need it

51:12

to run faster to use fewer resources because like if you want a laptop right next year by now, like it's just everything is getting very very tight in the supply chains and also that uh uh it's just human software is just not secure enough. >> How do you make your own self maximally ambitious? Like are there other people that you're going to to learn? Like okay, so I guess we'll back up cuz we talked about um you you're like, dude, I love the Sam Altman episode you did, but I had to skip over the first five minutes. But this is not something I'm prompting. So DHH comes on the show and he's like, Toby, nerd snipe me. I didn't know anything about AI. I was dismissive and now I'm like envious that I can't think about AI the way he does. I hear this from a million different founders by the way on and off the show. And then Sam Altman, I'm just [ __ ] sitting with him before the show recording, he

51:57

started bringing up Toby and the fact that he gets this insane like he learns more about how CEOs and founders are using his product from you than anybody else and you give him this crazy feedback. So like you're obviously people are looking to you, but like who are you learning this stuff from? Are there resources, specific people, companies and like how are you making sure that the same advice you're giving the other people don't like be maximally ambitious? How are you doing that for yourself?

52:23

>> I, I'm extremely fortunate in that I think if you can create self-sustaining small groups, like uh, along both lines like within Shopify I just have groups of people where everyone is kind of at the same level. And it's like maybe there's the implicit: you're part of this particular club because you can keep up, and to keep up you also have to contribute. So like there hasn't been a weekend this year that I basically didn't work throughout the entire weekend because like I just like I have zero ability at this point to stay away from progress, computers and stuff like this. I don't know. I I I'm made for this moment. I I think I I didn't know this, but I like every decision I've made in my life pointed at this moment that to be maximally prepared and useful, um but like I certainly think about it in this way.

53:09

Again, [snorts] if you want to be successful in life, it's very simple. You just have to be too good to ignore and there's opportunities. You got to be like among the first people to that are being called when something interesting happens or you need to be the obvious choice when a promotion when when a position Like it's be too good to ignore. That's that's all. Like figure out your own way there. Um that's up to you. That's the beautiful thing is there's millions of ways to do it, but be too good to ignore. Inside I have like groups of people that I can talk to who are like way ahead of me on many many things and I can learn from them and so on. So it becomes self-sustaining. These groups exist in many companies. A lot of people share what they know on X or in podcasts. Like I mean, I, I, I unschooled the close to the edge of my understanding, uh, of earlier here. Like the the, the talk track

53:56

of, um, insufficiently ambitious prompting, the underscore around implicit state machines, the the, the fact that we have to move from productivity towards much more constrained systems through linting and through formal verification that circumscribe this problem space and which you want solved by the AI. You do all those kind of things and you can outcompete everyone who doesn't understand them right now, right? So like there's a lot of sharing.

54:22

>> Can you say more about the gap between the productivity of the people that do understand what you're saying and the people that do not? >> I think it's the biggest gap of productivity that's ever existed in any industry, even invariant time probably. It's incomparable. It's a modern factory versus a medieval blacksmith kind of territory. There's so many zeros in terms of difference. Time has almost stopped being a concern. The variable is just how much are you willing to invest in terms of cost.

A weekend rewrite that used to take years 54:51
54:51

>> We don't understand why time would stop a gets her? >> Again, it's called building factories. It's a dumb term and it will be renamed. If you have an like an agentic factory that like really works well in the problem domain you're trying to solve problems in, [snorts] you can make a decision on Friday and you show up and uh like on on on Monday with production code. I'm not making this up. I literally just did this like I completely rewrote our entire like I mean not entire thing but like hey and I rewrote like it is a figurative like [laughter] a term at this point not to be taken literally.

55:24

The system of our internal source repository system that hosts all the code in Shopify Core data sort of loosely based on a tool I shared called Wargit a while ago. Yeah, that's in production thousands and thousands of engineers plus the rest of the company everyone commits into this thing. It's like I had to update it because we are now merging pull requests at a rate which is like I don't know 10 to 100 times higher than it has been at any other point in history in the company because of the acceleration.

55:53

So it need to be faster. So it now it's now it's faster and that was a weekend project not a multi-year rewrite. >> Okay, so that was a weekend project 5 years ago would have been taken how long? >> I don't even know if you would do it, because it's like it's kind of why, like, you you would use GitHub. And and the reason why you would use GitHub is because GitHub was perfectly good enough for the pace at which humans wrote code, which is like: you worked for weeks and then you make a pull request. Maybe there's multiple pull requests at the same time, but it's like it, it's bounded by the amount of engineers you have, and so the it, it never gets like that crazy. But now, if you decide, "I need to rewrite our entire database layer like system because I have this beautiful idea that can cause millions of lines of simplification in the systems that we then have to rewrite in Rust," which is

56:45

so like doing pre-work basically to make the next task cheaper and more solvable, then you can build all of that in a day. You would never ever have done this because it's too hard. It's like it's a too hard thing because you you'd have to find you need specialized skills to staff it. You would then have to keep them being a really really driving team for a year and at the end you get something that you then can check to figure out if your assumptions were right. It's like in a company that makes software for entrepreneurs and commerce and and so on. That's too much of a side quest, but now it's like it's worth for $2,000 of tokens, right? So >> You just mentioned your mission. So DHH uh he said when he was on the show so he's on the board of Shopify and he says you should see some of these internal tools that they're building. It's like better than anything that I could buy is what he in his words. And he was like begging

57:40

for you to give them or sell them to him, right? How do you decide what are these tools you're going to open source? What are you going to keep for yourself? What you could you sell as you know additional services to your customers? >> So Shopify has always been a internal tool in company which makes sense because we are tooling company, right? We make tools, we tool makers in Shopify we make tools for entrepreneurs. Therefore we make tools for making tools for entrepreneurs. It's like very very native to us. I think we talked about this before in an episode already.

58:10

This culture of internal tools it's a bit of a two-sided coin sometimes because like we do put a lot of effort. It's it's a very high status thing in Shopify to work on internal tools. In every other company I as I understand it it's not the best people on boards. It's the hardship assignment inside of a company. It's usually where you get some interns you put them on internal tools. Yes yes yes. >> And you're saying in Shopify >> in Shopify is opposite. So like maybe not quite opposite but like it's it's some of best engineers work on internal tools. We found this to be correct for us and easily worth it. So [snorts] guess how prepared we were for the AI age because of this. Now we already have a culture of internal tools. We have all this infrastructure for internal tool making. The AIs can now auto complete basically every internal tool.

58:51

Internal tools are wonderful to work on because you can just deploy them and your colleagues say if it's any good, and if it isn't, you make changes. And if it you need some downtime, who cares? Like, like production software is like, I mean, you need, like, four nines of uptime and whatnot. And you know, I'll show you some stuff after the episode. I can easily blow your mind. It's like it it's again the way Shopify feels internally right now. And I really, really hope this is true for everyone else who's so lucky to work in tech companies. It is magic. It is absolute magic. Like we are all kind of wizards conjuring up beautiful new things from almost from spells we cast into factories and agents and AIs and like it it's just it is like I've never seen anything like this. I mean, I really hope that Florence Florence might have felt like this during the Renaissance, but I kind of don't even think so,

59:47

right? Because it's just crazy. I think people have not fully priced in and understood how amazing things are going to get now. There's such a limiting factor that existed around software. There's so many creators. Like we're going to have a billion people easily who will make things with like on computers. And man, there is a lot of incredible brains, incredible minds that just cannot make things right now because they just don't have they didn't have the skills until recently.

1:00:19

But learning how to work with these systems is getting easier every day and involves using of language as as main tool. And it will support any language. So it's not even like that you have to learn English first. It's such a positive thing for my mind and we definitely have very much arrived in this sort of world. I think we're tracking very early, but my god, is it nice. Like we what we want to do is we want to like get our customers to be like the part of industry that sort of experiencing this first as well. So this is the the products we are creating is like again, first we create products for us because we need to understand the new tools that exist to a degree that we can play them like an instrument. And only after we've like figured out what works, what doesn't, does it make sense to productize it into our core products? Because otherwise it doesn't come from a point of fantastical

1:01:13

insight. And I think software that doesn't come from a point of fantastic depth of insight is transient and there's too much of it. Like it, it maybe both a little bit for a moment, but it will always be replaced by once the sort of a beautiful hidden harmony behind all the tools that people have created has been spotted and then is taken as a core foundation on which entirely new products are being built. And that moment is now rapidly arriving.

1:01:42

Like I said, like software will become fantastical, malleable, essentially bug-free, incredibly fast, just delightful in every way it can. People don't have to care about software. For people it just has to work, but like the makers of the software should love it and be proud of what they built because otherwise there's no chance that people will like them like it and it's not worthy of creating true fans, which is I think what great great software does.

1:02:07

>> Say more about the fact that you don't think how great things are going to be is priced in. >> First of all, because the shape is going to change. Initially I thought what would happen is first the quality of all existing software is just going to go up. Again, bug-free, fast, no splash screens anymore, this kind of stuff. Again, this doesn't sound like much, but

Why computers shouldn't make you feel bad 1:02:23
1:02:23

people love their computers when they are everything happens instantly and they really really dislike computers and usually feel it's their own fault somehow when everything's sluggish and they kind of can't accomplish things. When software is like too verbose and too complex, or you can't find where the feature is that you want, like you ever been in, like, I don't know, Word and you just sort of try to get the line, the page breaks right, and there's like some things that just like go to the next page and you don't quite know why and like, it's like people oddly never end up being mad at Word for that.

1:02:54

People end up being thinking they are bad at computers, right? They they internalize this, but then also that's a negative feeling and then they don't want to use computers to solve problems, which I also understand because it made them feel bad and like I don't think a computer should ever make anyone feel bad about themselves. If you want to throw a computer across the room or in a garbage bin or so, that is not your fault, that is the computer's fault. Now we can fix this and make things just that look the way people want and at you ask it the question and it will help you.

1:03:21

That is by itself will cause more people to accomplish their tasks at work and then potentially due to much much better experiences be more ambitious, maybe in the off hours with their own projects and that will also work. And then that might lead to times where people might reach for you know the the the institution of entrepreneurship to to to make products that you can sell and that you can share with others. [snorts] And I think again, these are the things I want to see. That's my like entire philosophy and bias. So I I I I state those as the most important things to accomplish because they are the most important things for me.

1:04:01

>> I'm going to interrupt you because what I love about you and why like our conversations is because you're an evangelist for entrepreneurship. And the greatest definition of a business I ever heard actually came from Richard Branson and he says all businesses is an idea that makes somebody else's life better. And so what you've just beautifully described for us is the fact that these tools, this new technology if you embrace it, it just allows you to make products that make other people's lives better.

1:04:24

>> Yeah. I completely agree with what Richard said. The entrepreneurship part of it is essentially just saying, hey, I got something that I think is good enough to join the market with it, making it available through the wonders of the market to the system that a shares it with everyone. Now now once it's on the market, people can access it. They can exchange money for it to get to gain gain it. And of course, that money is a form of energy directed backwards at the person who created it, which then causes more of these products to be created, more iteration, better products, maybe a next version of it later.

1:05:02

I just feel we need to appreciate the system more. I always like cringe when I see people talking about capitalism and stuff like this. Capitalism the capitalism is not the point of it. Like it's it's free markets are the point, but like it's capitalism is just sort of like a term that like usually offered in by by talking about the sort of small bits and pieces around the margins that the markets stop being free, usually in form of monopoly or often regulations frankly. Patents are monopolistic, right? Like and and and disable the free markets. They might be worth it societally, especially for drugs, but nowhere else are they beneficial and still do them.

1:05:41

Capitalism can go either way. Free humans are entrepreneurial in the form of putting things into a market. The quality of our lives is determined by how vibrant this market is and how many ideas and how many products and how many services are shared for it. And so it is our project on planet Earth as part of any nation, I believe, to cause the market to become more vibrant, right? Like to make more products, support them there, to create value, to take it out of the potentiality of of a of planet Earth and make it actually realizable, then the market grows. We call this GDP growth, but really it's it's energy growth because of all of it expanding.

1:06:25

That means there's more opportunity, there's more story, there are more successful people that might speak to you as models and you think you can potentially do the same thing as that person that you're observing. And as long as all these pieces fit together, man, are we in a good shape here. And I believe that almost anything that makes this flywheel go slower leads to less human thriving. Obviously I'm very far now with my explanation out of the overtone window here of like how people tend to think about these markets.

1:06:57

But am I wrong? And if I'm not, at least broadly, why is the current narrative in media outside of your wonderful podcast and very few other places so different about this? Why are the people who are building and creating companies so maligned to the point of drumming up violence? Violence. Now, everyone has a different tolerance for these kind of things. And again, the economy is our collaborative multiplayer game. And the market is the thing that we all trying to make go faster because it leads to more thriving, more vibrant society.

1:07:37

If the destination of this, like of, participation again this moment of entrepreneurship, the moment that you declare, "I've got something that I'm going to try in the market "if that is starting to come just with too many downsides, especially in terms of bodily harm, man, like, then we are going to go into the time of maximally most distributed participatory potential and opportunities that probably planet will ever see. More people could make companies and create products than any ever before.

1:08:08

And so many people will opt out because [ __ ] that's not worth it. I think that's dangerous and I think benefits no one. The good news is I I just looked at the Labor Bureau numbers. New business formation is at an all-time high. It is actually increasing. So right now things are looking good. But like I think there's a very very very very troubling trends here that are just like they they need to be reversed. Otherwise, this is going to get go poorly. It is so unbelievably important to just like spend a moment like celebrating the good parts of what's happening when anyone starts a company, small and and and and big. And it's so important to tell the stories in a way that like I think they become relatable. Like like they're all just like all the entrepreneurs are just they're just people. They decided to pick up a lot of skills along the way and put themselves out there. They're just trying to make

1:08:59

goods and services for each other at the end of the day. >> that's my definition of entrepreneur. Has idea, does them. As the first person on the receiving end of what you just said and the main takeaway I hope that I took away and that I hope that people listening take away is that all you need to do is just use technology to build tools or products that make other people's lives better. As always, man, thanks for making the time.

1:09:19

>> Thank you. >> I hope you enjoyed this episode. Please remember to subscribe wherever you're listening and leave a review and make sure you listen to my other podcast, Founders. For almost a decade, I've obsessively read over 400 biographies of history's great entrepreneurs, searching for ideas that you can use in your work. Most of the guests you hear on this show first found me through Founders.