We Hired a Top Programmer and Put Them on Slide Decks: Think Bigger Means Revisiting What Was Once Out of Reach

The world’s very best programmers are spending every day making slide decks and writing reports for us—I mean agents like Claude Code and Codex. They’re on call twenty-four hours a day for just a thousand-odd yuan a month, yet what they do is work boxed in by our own thinking: making slide decks, sending emails. What are we doing? It’s a sheer waste. If the goal is to make more money in a short time, we should be figuring out which ways make more money and have AI do those things, rather than have AI make more slide decks.

Calling them the world’s top programmers is no exaggeration. This February, Anthropic ran an experiment: sixteen Claudes worked together for two weeks, with not a single line of code written by a human, and built a hundred-thousand-line compiler from scratch—one that can even compile the Linux operating system—for under twenty thousand dollars in total. At last September’s World Finals of the International Collegiate Programming Contest, there were twelve problems; OpenAI’s model solved all twelve, while the best-performing human team solved eleven.

So how do people actually use them? A report Anthropic released this June shows that on weekdays people use Claude to write business emails, write marketing copy, and make slide decks, switching only on weekends to talking about relationships, asking about health problems, and asking about investments. Across all conversations, the ones that produced code made up only about one-sixth. It’s the same with OpenAI. Last September, OpenAI researchers and several economists published a report with the National Bureau of Economic Research (NBER), sampling over a million ChatGPT conversations: among work uses, writing ranked first, at forty percent; across all conversations, those related to coding were only four percent.

Let’s ask it a different way: if the best programmer in the world were working at our company today, available anytime, and for free, what would we have this programmer do? For example, the owner of an antique-watch business might at first feel AI has nothing to do with them—they just sell watches. Asked this way, though, the possibilities suddenly multiply: a separate page for every single watch, each watch’s era, design, and the story behind it written up, the website rebuilt from the ground up, the customers studied, and even the small customers who could never get attention before finally served.

A lot of things didn’t get done in the past not because they lacked value, but because they were too expensive: no programmers, no designers, no research team, and no time. Now all of that is here. In the past, many people lost because they couldn’t do it; from now on, more and more people will lose because it never even occurred to them that this can already be done. AI has raised what each of us is able to do. This April, OpenAI co-founder Greg Brockman said on the Big Technology podcast that what AI will raise next is how big a problem people dare to take on. But how big we dare to think is something AI won’t raise for us automatically.

At the in-person sharing session in Chongli this time, a few women from our group—a director, a teacher, and someone in advertising and marketing—did wonders with AI in just half a year. They used AI to coordinate the large-scale events of the Chongli 168 ultra-trail race, a trail race with more than twelve thousand participants this July, and they used AI to build all kinds of interesting platforms and apps. I’ve said before that people who aren’t programmers can build AI products too. Us guys, on the other hand, spend every day just using AI for consulting and making slide decks.

None of them are programmers; they’re professionals in their own fields. Professionals in every field are in the same position: they used to be close to the problems and far from building a product, and now that distance has shrunk. We don’t need to become the best programmer in the world. What we need to know is which problem in our own industry is actually worth having the best programmer in the world solve.

Take healthcare. Gene-level health management used to require an entire team. Blueprint, the program run by American tech tycoon Bryan Johnson, was covered by Fortune in 2023: Johnson was spending over two million dollars a year, with thirty doctors revolving around one person. Now, with AI, the cost of what that whole team did is almost gone. Anyone with AI’s support can get to that level, and there’s so much open-source data to use as well. I’ve done it myself with Claude Code: I connected my own genetic data and checkup reports to public medical literature databases and produced a half-year follow-up plan.

Think bigger isn’t about shouting bigger goals, nor does it mean we have to build a ten-billion-dollar company. It means looking back at the things we couldn’t even afford to touch in the past because there were no people, no money, no time—things that, today, are worth serious thought for the first time. This question isn’t a test of technology; it’s a test of what problems we see, and whether we dare to take on a problem we once felt we had no right to touch. We should be doing things that can truly change the world—think bigger.

EP0113 · 2026-10-03 · 03:58. The picture uses a digital human, and the voice uses my cloned voice. This is an English translation of the approved Chinese script used to create the narration.

We should be doing things that can truly change the world—think bigger.

The world’s very best programmers are spending every day making slide decks and writing reports for us. I’m talking about agents like Claude Code and Codex. We’ve got agents this good in our hands, on call twenty-four hours a day, for just a thousand-odd yuan a month, and yet what they’re doing now is all stuff boxed in by our own thinking. We’re still having them make slide decks, still having them send emails. What are we doing? It’s a sheer waste.

If our goal is to make more money in a short time, then we should be figuring out which ways make more money, and then have AI do those things—not have AI make more slide decks.

And calling them the world’s top programmers is no exaggeration at all. This February, Anthropic ran an experiment: sixteen Claudes worked together for two weeks, without a single line of code written by a human, and built a hundred-thousand-line compiler from scratch. It can even compile the Linux operating system, and the whole thing cost under twenty thousand dollars. Last September, at the World Finals of the International Collegiate Programming Contest, there were twelve problems in all. OpenAI’s model got all twelve right. And the best human team? They got eleven.

So how do people actually use it day to day? This June, Anthropic put out a report. On weekdays, what people used Claude for was writing business emails, writing marketing copy, and making slide decks; only on weekends did it shift to talking about relationships, asking about health problems, asking about investments. Across all conversations, the ones where the output was code made up only about one-sixth. It’s the same on OpenAI’s side. Last September, OpenAI researchers and several economists published a report with the National Bureau of Economic Research (NBER). They sampled over a million ChatGPT conversations. Among work uses, writing came first, at forty percent. Across all conversations, the ones related to coding were only four percent.

So let’s ask it a different way. If the best programmer in the world were working at our company today, ready to work for us anytime, and for free, what would we have this programmer do? For example, take the owner of an antique-watch business. At first they’d think, what does AI have to do with me? I just sell watches. But ask it this way, and suddenly there’s so much more they can do: build a separate page for every single watch, pull together each watch’s era, its design, and the story behind it, redo the whole website, study the customers, and even serve those small customers they never had time for before.

A lot of things didn’t get done in the past not because they weren’t valuable, but because they were too expensive: no programmers, no designers, no research team, and no time. Now we have all of that, right? In the past, many people lost because they couldn’t do it; from now on, more and more people will lose because it never even occurred to them that this can already be done. AI has raised what every one of us is able to do. This April, OpenAI co-founder Greg Brockman said on the Big Technology podcast that what AI will raise next is how big a problem people dare to take on. But how big we dare to think—AI won’t raise that for us automatically.

At the in-person sharing session in Chongli this time, there were a few women from our group—a director, a teacher, and someone in advertising and marketing—and in just half a year, they honestly did wonders with AI. They used AI to coordinate the large-scale events of the Chongli 168 ultra-trail race, and the Chongli 168 was a trail race that more than twelve thousand people took part in this July. They also used AI to build all kinds of interesting platforms and apps. I’ve said before that people who aren’t programmers can build AI products too, right?

Us guys, on the other hand, spend every day just using AI for consulting and making slide decks.

And none of them are programmers—they’re professionals in their own fields. Professionals in every field are the same: in the past, they were close to the problems but far from building a product. Now that distance has gotten shorter. We don’t need to become the best programmer in the world. What we need to know is which problem in our own industry is actually worth having the best programmer in the world solve.

Take healthcare. Doing gene-level health management used to take an entire team. Blueprint, the program run by American tech tycoon Bryan Johnson, was reported on by Fortune in 2023: Johnson was spending over two million dollars a year, with thirty doctors revolving around one person. Now, with AI, the cost of what that whole team did is almost gone. Anyone with AI’s support can get to that level, and on top of that, there’s so much open-source data to use. I’ve done it myself with Claude Code: I hooked up my own genetic data and checkup reports to public medical literature databases and came up with a half-year follow-up plan.

Think bigger doesn’t mean shouting out bigger goals, and it doesn’t mean we have to build a ten-billion-dollar company. It means looking back at the things that, in the past, we couldn’t even afford to touch because there were no people, no money, no time—and today, for the first time, they’re worth thinking about seriously. So this question isn’t testing technology. It’s testing what problems we see, and whether we dare to take on a problem we once felt we had no right to touch.

We should be doing things that can truly change the world—think bigger. That’s it for today. See you tomorrow, bye.