John Collison
Related people
- Alexandr Wang
- Chris Sacca
- Evan Spiegel
- Mark Zuckerberg
- Mati Staniszewski
- Molly O'Shea
- Sam Altman
- Scott Kirby
Episodes
- John Collison, Stripe: AI Agents Will Rewire the Internet
- Muse Wonder: Zuckerberg and Wang's $100B AI
- How Toto Wolff built the most valuable team in F1
- The economics and trends of the restaurant industry, with Tony Xu of DoorDash
- The history and future of AI at Google, with Sundar Pichai
- The airline CEO putting Starlink on every plane
- The company building America’s uranium supply from scratch
- The world of voice AI, with Mati Staniszewski of ElevenLabs
- What comes after smartphones, with Snap CEO Evan Spiegel
- Why Sam Altman is against a single superintelligence
- The company building underwater drones for $50,000
Notes
Cheeky Pint · The company building America’s uranium supply from scratch
-
Why Was 20% of Vogtle's Nuclear Cost Overrun Just Interest?
Scott Nolan says interest during a long construction job, not the famous accidents, is a large part of why American nuclear plants got so expensive. Vogtle ran far over budget and behind schedule. About 20 percent of that cost overrun was interest, and that was in a period of historically low rates. When a plant takes about 10 years to build, the overnight construction cost matters less than the interest paid while it sits unfinished. His fix for the big reactors is to build them faster and to build more of them, so the work finally gets economies of scale.
John Collison points to the 1970s slowdown, when mortgage rates reached about 18 percent in 1981 and the cost of capital was well north
Cheeky Pint · Why Sam Altman is against a single superintelligence
-
Why Are Web Forms Fake Work?
John Collison says any web form is fake work. The person filling it in is almost never inventing the information for the first time. A family medical history is the case he uses: the same facts, typed out again, feel like a waste. He would be happy never to do that kind of work at a computer again. Sam Altman agrees.
Companies still treat their HTML and their mobile apps as what defends the business. The defense is either wrong or temporary, because it did not come from assembling a page. The iPhone App Store was the last clear version of that fight. Startups in a YC batch rushed to build an iPhone app for the new distribution. Big companies hesitated, unwilling to hand the customer relatio
-
Why Don't Personal Agents Feel Essential Yet?
Sam Altman says personal agents have not had a Macintosh moment. People like the current ones, including Dots, but they do not say they cannot imagine life without them, which is how people talk about coding agents. The category is a small number of ideas away from being mainstream.
The winning agent will look a lot like the ones shipping now, including Dots and Muse, and it will just work. It will be reliable. Hand it more complex tasks and it will do them. John Collison calls personal agents one of the most exciting things happening in AI. He puts the winning version down to intelligence, context, and tools, and he expects someone to find that shape soon.
-
What Is a Self-replicating Data Center?
Sam Altman says a robotics program is being pointed at making lots of data centers, and self-replicating data centers are within grasp. AI would control the robots, run fabs, build turbines or solar panels, and build the data centers, so compute gets cheap. Most talk about RSI means an AI that does AI research. There is much less talk about data centers that build more of themselves. A chip built for this is described as incredible performance, in units of intelligence per watt, and the same push is learning how to build its own data centers.
John Collison asks whether data centers are already past a labor bottleneck and are stuck on permitting and components. More components for the robots
-
Why Are AI Models Finding PHP Bugs Now?
Sam Altman says the models have been getting good at finding vulnerabilities in the wild, and they keep getting better quickly. People had not pointed them at all the code on the internet. Once that search starts, he expects many more serious vulnerabilities than even the people who already expected a lot.
John Collison says patching the software is not enough. The architecture has to change, which is harder and much more durable. The change would look very hard, except the models are the tool for doing it. A new generation of startups is already talking about a new kind of internet and a new application architecture that assumes the models keep getting smarter.
Cheeky Pint · How Toto Wolff built the most valuable team in F1
-
How Much Profit Does the Mercedes Formula 1 Team Make?
Toto Wolff says Mercedes Formula 1 generates around a billion dollars in revenue and a 30 percent profit margin, about 300 million of EBIT.
John Collison put the stake sale at a 4.6 billion dollar valuation, about 15 times that EBIT. Wolff says the multiples justified the price when he sold part of the team last year, and that the business could probably show about 30 percent more today, though he already has the partner he wanted. The team distributes dividends. He describes many American teams as living in a debt structure, while Mercedes does not. The team also skips heavy construction spending because it does not own a stadium, and it gives up some of the revenue a stadium would bring.
-
Will the Madrid Circuit Add More Straights Next Year?
Toto Wolff thinks the Madrid circuit will change its layout next year to add more straights. The track is not permanent, and the design did not account for how hard the new generation of cars makes overtaking, even though a long banked corner looks spectacular for the drivers and for camera positions.
He loves old circuits as a traditionalist because they represent history. Layouts with vast areas, available because places were less developed, still provide great racing: overtaking, and cars able to run side by side. Criticism of the new Spanish track was that there was not a lot of overtaking. Madrid as a city is spectacular, the track built there is spectacular, and the job done over the
-
How Early Can You Spot a Formula 1 Driver?
Toto Wolff says a kid's karting shows pretty early whether that kid can become a professional racing driver. Even the tiny ones stand out against their peers. An eight-year-old who beats the ten-year-olds in the group is an early sign. A kid going 100 kilometers an hour in the rain, sideways across a corner on a first lap, already has the ingredients for being good.
Racing has no single number that separates one driver from another. John Collison points to cycling, where Tadej Pogacar is the best in the world on threshold power, the power a rider produces for body weight. Threshold watts per kilogram and VO2 max are measurable in cycling. Pogacar's metrics at 12 are unknown, and whether the
The Peel with Turner Novak · Why Stripe’s 1st Employee Bought a Bank | Darragh Buckley, Increase
-
Why Did Patrick and John Collison Sprint While Unloading?
Darragh Buckley says Patrick Collison and John Collison sprinted while unloading a new office because they always asked why the work would not happen faster. A first office arrives empty, so the errand is an Ikea trip for six of the cheapest desks and the cheapest chairs. The usual pattern is to set a box down and walk back to the car. They picked the box up, walked it in, put it down, and sprinted.
He wondered whether the car was parked illegally. When other people are already running, being the one person who does not run stops being an option. The group gets there faster together, without rubbing it in when someone is slower. When the day involves a lot of traveling and movement, moving
-
What Lesson Stuck From the Early Stripe Team?
Darragh Buckley says one of the most impactful and meaningful lessons from the early Stripe team was how much a small empowered team can get done. He walked into an early group with John Collison, Patrick Collison, and Greg Brockman.
Stripe was finally going to talk to the credit card network more directly. On a partner call, the person in charge of certification said the next date was six weeks away and six weeks should be enough. In startup time, six weeks felt as far off as 10 years. Someone on the call asked whether any time was sooner. Tomorrow at 10 a.m. was open, along with a warning that the team was not going to be ready. They signed up anyway. The person on the call still assumed
Cheeky Pint · Muse Wonder: Zuckerberg and Wang's $100B AI
-
Why Did Mark Zuckerberg Reboot Meta's AI Lab?
John Collison names four moves he thinks even an opinionated founder would not make. Mark Zuckerberg treated the recent prior AI effort as not good enough. Standing up Meta Superintelligence Labs is the reboot, and people can read that between the lines. He was willing to pay a lot of money to bring Alexandr Wang over, including 49 percent of what Collison calls a very good grindy business. Wang had worked in AI and had never trained a frontier model, so the hire was a bet on potential rather than on having done that exact job. Zuckerberg also cleared his calendar and did only this for two months.
Wang says that once Zuckerberg decides something is right, he shifts the work to match that be
-
Why Did Researchers Join Meta Superintelligence Labs?
Alexandr Wang says a reason people were excited to join Meta and Meta Superintelligence Labs is that they wanted to work somewhere focused on giving this technology to billions of people. He says that is not the stated goal of every other frontier lab. People are choosing the world they want, in where they work and who they do business with, and those choices push away from a single winner.
The lab's standing start is recent. Llama 4 launched on April 4, and they date that to 2025. Around April 20 or 21, the calls for advice on what to do began. John Collison first says the lab is a little over two years old. The room corrects him: it is a little over one year old.
-
Why Hasn't One AI Lab Run Away With the Lead?
John Collison recalls the 2023 story that one AI lab would get ahead, start recursive self-improvement, and pull away from everyone else. He says that has not happened. Several labs are still at the frontier, Muse Wonder is arriving as Meta's frontier-class model, and open-source models have kept up.
Alexandr Wang gives three reasons. Execution moves around. Every frontier lab has strong stretches and weak ones, and those average out. If a lab in the lead had executed perfectly the whole time, he thinks it would probably have stayed ahead. Distillation is the second reason. He calls it an anti-hegemonic force in the technology, not a plan anyone chose. It packs strong capability into smalle
-
How Does Meta Plan to Pay for Its AI Buildout?
John Collison asks how the AI buildout gets paid for. He says Google went free cash flow negative, and that Meta is not free cash flow negative yet, with more capital spending still ahead.
Mark Zuckerberg says Muse has to be widely used, because the network gets more valuable as people interact through it and as partners integrate. The way to get there is a free base that is still excellent. Revenue then scales in a few places. Using Muse Wonder inside Muse takes a subscription, because that larger model is not economical to give away. Commerce can take a small cut of money the assistant saves or makes. A business-in-a-box version could run pieces of a small company and be profitable for th
-
How Much Does Training a Frontier AI Model Cost?
Mark Zuckerberg puts the figure you would see in public, to operate a gigawatt cluster for a year, on the order of $15 billion. John Collison calls that roughly a $15 billion effort to train a frontier model. Zuckerberg separates that operating cost from construction. Building the capital expenditure is probably closer to $50 or $60 billion, paid up front, and that is one cluster for one part of training.
The final run is not the whole research bill. Each change is tested up a ladder to full scale, and that testing is probably an equal amount again, before anyone counts serving the model. Zuckerberg agrees that tens of billions of dollars is a fair estimate for training a frontier model now
-
Why Is Meta Spending $100 Billion Before the Model Exists?
Alexandr Wang says frontier AI is the opposite of a lean startup. A lean startup gets something working and only then doubles down. You cannot do that here. The bets have to be made before there is proof the thing works, or you end up radically behind. He calls it entrepreneurship on hard mode, and he gives Mark Zuckerberg credit for taking those decisions.
Zuckerberg retells the end of last year. They were excited about the progress and had not seen the models yet. The check they were going to write was, as he put it, like $100 billion or $200 billion, and he asked Alex if he was sure it was going to work. John Collison describes the same constraint as pushing the chips all in before the c
-
Why Do Winning AI Products Look So Simple?
John Collison says winning AI products look simple because the field collapses quickly onto a few designs you could describe on the back of a postcard. Making a good version is still massively complicated and takes huge amounts of frontier work.
At Stripe, spelling out a winning product would take him pages: developer experience, payment methods that differ from country to country, and stablecoin. In AI, transformers worked, then the chat agent interface worked, and companies are now collapsing on a persistent agent that has a computer and can act out in the world. Once people see Muse, everyone treats that shape as clearly how an AI agent product should work, and future AI agent products w
Sourcery · John Collison, Stripe: AI Agents Will Rewire the Internet
-
What Is John Collison's Biggest Lesson From Seventeen Years Building Stripe?
John Collison's biggest lesson from seventeen years building Stripe is that there are no shortcuts. Stripe never had a South by Southwest breakout moment. Every year was a bigger year than the prior year, and after enough years Stripe reached one point nine trillion dollars.
-
When Does John Collison Buy a Skill Instead of Building?
John Collison says Stripe is more likely to build something that sits in the core of Stripe. A brand new muscle, like token routing, would take much longer to learn inside the company. A fast moving market can make an internal build a little bit out of date, because the market is totally different today than it was twelve months ago.
-
Why Won't John Collison Share Stripe's Artificial Intelligence Volume?
Stripe's goal is to increase the GDP of the internet, and a lot of the growth now comes from artificial intelligence, but John Collison will not say the exact fraction. It is hard to measure. Meta runs on Stripe and grew up before artificial intelligence, yet it is increasingly becoming an artificial intelligence company.