The revolution will be ticketed

Real revolutionaries wear suits.

If you are a bold entrepreneur who wants to change the world with ambitious artificial intelligence technologies, how should you do that? Two answers seem obvious:

1. Build ambitious artificial intelligence technologies.

2. Talk about building ambitious artificial intelligence technologies.

Arguably, the second thing is just as important as the first thing.¹ You are selling sweeping visions of prosperity and productivity as much as you are selling technology. Leading artificial intelligence researchers—the people you need to build ambitious artificial intelligence technologies—want to work at big companies with lofty missions that can make them rich. Customers want to buy stuff that will help them survive the looming mass obsolescence event. Venture capitalists want to fund companies “with more ambition than ever.” One AI startup periodically publishes essays about how their ambitions are somewhere between destroying the world and saving it; they are about to raise $50 billion.² Another AI startup told investors that they should consider their investment a donation because “it may be difficult to know what role money will play” in the world if they are successful; it was the greatest fundraising pitch in history. To change the world with ambitious artificial intelligence technologies, do not talk quaintly about “making something people want;” talk about an age of 100x productivity; talk about recalibrating your gradient; talk about ushering in new eras of maximal abundance.

Before you can do any of that, though, you will have to first talk to the government about building ambitious artificial intelligence technologies. To incorporate a company, you have to pay various fees and fill out various forms, and some of those forms will ask you about your company. As a matter of accuracy and as a matter of ego, you might want to memorialize your bold ambition into your company’s founding charter. You will click the dropdown—“What does your business or organization do?”—and look for the appropriately audacious selection: Revolutionize labor; or, Launch intergalactic data centers; or, the Jetsons. You would probably be happy with Artificial Intelligence. You might even settle for Software.

But, no. Those will not be options in the dropdown. The IRS bureaucrats are unfazed about your ambitions. The BLS econometricians do not recalibrate their gradients. They do care if you have already raised $4 billion, if you have so many Porches that you get stuck in your traffic, or if you live in a moon house floating over San Francisco like the mothership. They care about the consistency of their statistics. And so, when you fill out their forms, you will have to declare, officially, to the United States registrar of permanent records, that your business does Information Technology.

It will feel like a terrific indignity. IT is the opposite of ambition; IT is anathema to ambition. IT is administration. IT is paranoia. IT is a hall monitor, blocking websites. IT is your mandatory anti-phishing training; IT is the disabled skip button on the video player playing your mandatory anti-phishing training. IT is your flaky VPN. It is a PC. It is Microsoft Teams. It is a stiff suit and a lanyard; it is khakis, a polo with “IBM Medallion Partner” on the sleeve, an OGIO Renegade RSS backpack, with both straps on. IT is the brakes.

Bold entrepreneurs and ambitious artificial intelligence startups are all gas. They do not follow the rules. They experiment; they dangerously skip permissions; they build with unhinged and explosive abandon. They run swarms of phishing software directly inside their VPNs. IT is supervising Jira. AI is replacing Jira; AI is agentifying Jira; it may be difficult to know what role Jira will play in the world AI is building.

Information technology, pfft. You are building a cathedral that will change the world, and all the government can see is the bricks.

---

And, I mean, just look at how much AI is changing the world. It is changing Block:³

today we’re making one of the hardest decisions in the history of our company: we’re reducing our organization by nearly half, from over 10,000 people to just under 6,000. ... we’re not making this decision because we’re in trouble. our business is strong. gross profit continues to grow, we continue to serve more and more customers, and profitability is improving.

but something has changed. we’re already seeing that the intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company. and that’s accelerating rapidly.

And earlier this week, Coinbase:

Today I’ve made the difficult decision to reduce the size of Coinbase by ~14%. … AI is changing how we work. Over the past year, I’ve watched engineers use AI to ship in days what used to take a team weeks. Non-technical teams are now shipping production code and many of our workflows are being automated. The pace of what’s possible with a small, focused team has changed dramatically, and it’s accelerating every day.

And, just yesterday, Cloudflare:

We’ve made the decision to reduce Cloudflare’s workforce by more than 1,100 employees globally.

The way we work at Cloudflare has fundamentally changed. We don’t just build and sell AI tools and platforms. We are our own most demanding customer. Cloudflare’s usage of AI has increased by more than 600% in the last three months alone. Employees across the company from engineering to HR to finance to marketing run thousands of AI agent sessions each day to get their work done.

Of course, there are questions. Is AI changing these companies, or are they using it as an excuse to correct past mistakes? Have they oversold themselves on AI’s hype? Many people are saying this.

Still, regardless of the motivation,⁴ the companies are changing:

Cloudflare – [We are] defining how a world-class, high-growth company operates and creates value in the agentic AI era.

Coinbase – We’re reshaping Coinbase to lead in this new era.

Block – We’re going to build this company with intelligence at the core of everything we do.

Coinbase – We are adjusting early and deliberately to rebuild Coinbase to be lean, fast, and AI-native.

Cloudflare – We are reimagining every internal process, team, and role across the company.

Block – Block is showing what it looks like to fundamentally rethink organization design.

Cloudflare – [Cloudflare] cannot rest on the workflows and organizational structures that worked yesterday. We’re confident that our reshaped organization will be even faster and more innovative as we continue building the future.

Block – The intelligence lives in the system. The people are on the edge.

Coinbase – [We are] rebuilding Coinbase as an intelligence, with humans around the edge aligning it.

Block – Player-coaches [will] replace the traditional manager whose primary job was information routing. A player-coach still writes code or builds models or designs interfaces.

Coinbase – Every leader at Coinbase must also be a strong and active individual contributor. Managers should be like player-coaches, getting their hands dirty alongside their teams.

Block – Companies move fast or slow based on information flow. Hierarchy and middle management impede information flow. … Block is building what comes next.

One way to read all of this is that AI is reinventing Block, Coinbase, and Cloudflare. But the other, more literal way to see it is that Block, Coinbase, and Cloudflare are reinventing Block, Coinbase, and Cloudflare. They are defining how their companies will work; they are rebuilding their companies with intelligence at the core;⁵ they are rethinking organizational design; they are building what comes next.

Decades ago, companies like Microsoft changed the world with Windows and the personal computer. But they did not author what happened after. That was figured out on the ground, across a million IT teams, testing new ideas against their specific problems, rebuilding their companies around the computer and the internet. It was in partnership with Microsoft, perhaps; as Medallion partners, perhaps; but the final cathedral of the modern corporation was designed by Jeff in IT as much as it was by Silicon Valley’s storied founders.

The same, it seems, will be true again. Box CEO Aaron Levie:

Whether it’s existing consulting firms, new ones that emerge, FDEs from agent vendors, or new internal agent engineering roles, the amount of work that is going to be created to implement agents in enterprises will exceed anything we imagine today.

AI companies cannot do that work. They can help with it—they can spend billions building consultancies and deployment companies—but it too will ultimately be carried out on the ground, across a million “internal agent engineering” teams, testing new ideas against their specific problems, rebuilding their companies around “intelligence.”

It is a frequent question now: What will work be like? If we are revolutionizing labor, what will labor become? AI companies are building the new technology, and it is easy to assume that they will be the ones who tell us the answer.

But the long tail of teams experimenting with AI will be much bigger than the AI companies themselves. OpenAI and Anthropic may build the model and some of its generic products; everyone else will figure out how to use them. And it will be their discoveries that become tomorrow’s standards.

Or, framed differently, if you are a bold entrepreneur who wants to change the world with ambitious artificial intelligence technologies, are you building the cathedral, or are they? Who is the real architect of the future—who is the most trailblazing pioneer?—the AI company, or the internal teams that are thinking about, uh, business technology and the internal flow of information?

---

XaiXspaceXanthropic

Last week, we talked about AI companies, their computers, and the cyclical race they’re all running:

1. To build frontier AI models, you need a lot of computers.

2. Frontier AI models are so popular that even big companies like OpenAI and Anthropic do not have enough computers.

3. So the companies have to choose: Do they use their computers to train new models, or do they use them to sell those new models, through applications like ChatGPT and Claude Code?

4. In some sense, you have to do both—you can’t sell a bad model, and a great model is no good if you don’t sell it.

5. So, you devise a two-step plan: First, focus on building the best model. Then, once it’s the best, you sell the model.

6. But! Once you build the best model, everyone will want to buy it—more people than you have capacity to serve.

7. What do you do??

8. You probably use the computers to sell your AI. Money has a way like that.

9. When you do that, though, you open the door for your competitors, who will simultaneously be selling less of their AI and in need of better one. So, they will probably focus on building the best model…

10. Etc., etc., etc.

However, in the original post, there were eleven steps. Between steps 5 and 6, there was a joke:

Sometimes, [building the best model] doesn’t work. You spend a zillion dollars, and do not build the best AI. Ah, awkward, an A for effort, I guess.

So what do you do then? One option is to spend a zillion more dollars and try again. But the other, easier option is to give up and sell all of your computers:

Anthropic on Wednesday announced a deal with Elon Musk’s SpaceX to use all of the compute capacity at his company’s Colossus 1 data center in Memphis, Tennessee.

As part of the agreement, Anthropic will get access to more than 300 megawatts of compute capacity, and it also “expressed interest” in working with SpaceX to develop multiple gigawatts of compute capacity in space. Anthropic said the deal will directly improve capacity for its paid Claude Pro and Claude Max subscribers.

Right, if you are in a bike race, and you are close behind the leader, it’s not that hard to stay in the race. From inside the peloton’s slipstream, you are one sprint from being back in front. But if you’re in 29th place, you can’t draft behind the leaders, nor are you likely to catch up to them. In which case, maybe biking just isn’t your thing.⁶

1. If you talk about it really well—that is, with a plaintext manifesto—you can raise billions of dollars before building anything at all.

2. Anthropic “has received multiple preemptive offers to raise fresh capital of around $50 billion at a valuation in the $850 billion to $900 billion range.” Yeah, uh, I have a feeling the number you’re going to need to hit there is $853 billion.

3. Do you think the internal email used capital letters? Do you think Jack Dorsey rewrote the internal email without capital letters before he posted it on Twitter? What do you think he was thinking when he did this?

4. And putting aside the bigger social and moral questions.

5. Both Block and Coinbase used almost identical language here: Intelligence at the core, people on the edge.

6. But also, maybe SpaceX is just ahead of the curve? In response to reader’s comment last week, I wondered if this might be the eventual story:

1. OpenAI, Anthropic, and others build huge proprietary models.

2. They buy tons of computers to run them.

3. The other models all get pretty good; there isn’t much differentiation between them; nobody really cares that much one about which one they use.

4. OpenAI, Anthropic, and others become generic AI compute providers because they have tons of computers.

Today’s cloud providers sort of backed into the cloud hosting business, in part because they were buying tons of computers to run their very computer-intensive businesses. If models converge but computers endure, eventually, what will be more valuable—selling models or selling computers?