If You Can't Beat AI, Do What's Fun

Back in March I wrote a post called Claude Blue: All of Silicon Valley Is Depressed. It was about how Claude was replacing developers' work at a frightening pace and sending Silicon Valley developers into a deep funk. A few days ago I got on the phone with a friend from my book club, a developer at a big Korean tech company. His story lined up so closely with that post that I listened far more intently than usual.

His story didn't end in gloom, though. It went on to his own answer for what to do once you accept that you can't beat AI. It stuck with me long after we hung up. So I'm writing it down.

Books don't make me curious anymore

We were talking about book club when he admitted he can't really get into books lately and, to my surprise, blamed AI.

I used to love the payoff after finishing a book, the whole experience of it. These days everything goes to AI, so there's just not much I'm curious about in books anymore.

So instead of reading for information, he picks books where the reading itself is the fun, like classic novels and essays. Right now he's wrestling with One Hundred Years of Solitude, where the family tree is so tangled and the names so hard that he starts over from the beginning every time he opens it. He keeps at it because he feels his reading comprehension has slipped badly. Every line is a challenge, but it keeps his imagination running. These days surprisingly little else asks him to imagine anything. For something lighter, he's reading Tuesdays with Morrie.

I recently enjoyed a book called Sayno's Teachings, and looking back, it was fun for the same reason he gave. Rather than organizing information from the outside world, it's one person's extremely subjective thinking, closer to reading someone's blog. Books of methods, on the other hand, go stale fast at today's pace. Just as he put it, we've reached the point where we have to rethink which books are even worth reading.

I haven't written my own code in over a year

Work, according to him, doesn't have the flavor it used to. He used to get absorbed in building things himself, and now there's almost none of that time left.

I haven't written my own code in over a year. I have it do everything, and now all I do is leave direction.

His output is way up from before, but the problem is that it doesn't translate cleanly into business. When I asked why higher productivity from AI wasn't turning into revenue, whether it was because everyone had leveled up together, he turned the question back on me.

Isn't it just that business was always hard?

Plenty of people were building things before too, and the ones that survived were lucky or happened to fit the market. There are just far more attempts now, but if what you make doesn't deliver value the market wants, it fails, same as ever. He thinks ordinary SaaS companies will have an especially hard time. Where you once logged meals in a diet app for advice, now you keep a GPT chat open and ask what to do next given what you ate today. You can wander off the rules the app set and keep talking however you like, so that's hard to beat. It's hard to find a simpler explanation for why people are talking about a SaaSpocalypse.

Can't fire, so they don't hire

The heaviest thing he brought up was hiring. The US went through a wave of mass layoffs as AI came in, but Korea can't do that because employment isn't flexible. Companies can't skip AI either, so they cover a fair amount of AI spending for each employee. The people stay and the costs pile on top, which for the company is like adding one more benefit.

So of course they don't hire new grads. It's too obvious. What reason is there to raise costs further?

At smaller companies the change has already started, with one product engineer handling planning and operations instead of splitting backend, frontend and data engineering. He expects even big companies to reach a point where handing out severance and a package all at once makes more sense than carrying people indefinitely. Nobody knows when, but those things are landing one by one as pressure. If there were a full-time contract where you worked two days a week for that share of the salary, companies and people could both catch their breath. Right now that option doesn't exist at all.

During the call he mentioned an interview with Rhee Chang-yong, the former Bank of Korea governor, on Syuka World, and when I looked it up later, a lot of it overlapped with what he'd said.

Rhee said the claim that AI is wiping out jobs for young people is half right and half wrong. Recent grads are the ones who use AI best, so if a company could reorganize freely, cutting middle managers and hiring more young people would fit productivity better. He described the labor market as insiders, already in good jobs and protected by the system, and outsiders, trying to get in or without permanent jobs, and he thinks the current system leans too far toward protecting insiders.

But Korea can't do that. Wages are set by seniority, so you can't freely shrink the organization or lay people off. So what ends up happening? You hire fewer young people.

Then there are the young people who finally get in: they're paid less while teaching AI to their seniors, yet it's the senior who takes the result to present to the governor, which is bound to build resentment. Rhee pointed to the time the retirement age was raised without changing the wage structure, when youth hiring dropped and marriage and births fell along with it.

His fix sounded a lot like my friend's: even if the retirement age goes up, wages have to be adjustable, and there should be a flexible setup where you come in two or three days a week for matching pay. A big-company developer on a phone call and a former central bank governor on YouTube were pointing at the same place.

The video runs nearly two hours. The labor market part starts around 1:12, the AI part around 1:24, and the chip experts come up about five minutes in.

I hope my pessimism is wrong

I asked whether all this change excites him at all, or whether he figures everyone's in the same boat so we'll muddle through somehow. There really is no alternative, he admitted. Think about it long enough and it can feel bleak. What steadies him is that the Korean labor market doesn't seem likely to change that fast, plus a semiconductor story from the same interview. A little over a year ago, when Samsung Electronics was struggling, Rhee spent a month meeting about 20 Korean chip experts and asking where things were headed.

Not one of them said a supercycle like this was coming.

Now the same people say the chip supercycle will last 10 years, so he wasn't sure whether to trust the experts. My friend took that the other way around: if no one saw this boom coming, today's pessimism that AI will wipe out jobs could be wrong too. In the same interview, Rhee said that when technology shifts, jobs can shrink in the short term, but new kinds of work can also emerge, so it's hard to say for sure that AI will reduce demand for labor.

I hope this mood I'm picking up is wrong. I hope it turns out to be, nobody knows what'll happen.

IT companies might not be the only answer

He mentors job seekers. With hiring this hard, most of his mentees are 27 or 28 and putting off graduation to hold on. Recently one of them interviewed at a big accounting firm. That firm has stopped hiring accountants the way it used to and started hiring in-house developers to move the whole company onto AI. People around him are getting calls offering AX consultant roles at global consulting firms too.

Building one proper internal system used to take at least three teams, app and server teams included, but now a single small engineering team can build quite a lot. That lightens the load of keeping a development team in-house at a place like an accounting firm, and it made him think IT might not be the only answer.

I'd been skeptical about how long AX would last, because helping someone else's company become AI-native is like being a cram-school teacher: once the student studies hard and graduates, that's the end of it. A cram school keeps getting new students, but there are only so many companies. We'll have to wait and see, in his view, though in-house teams don't stop at bringing AI in. They can make things nobody would have dared try before, so it might last longer than I think.

His coworkers were each thinking about an exit too. One is slowly learning interior work because their father is in the business, and another loves wine so much they're looking into importing it or growing it in Korea. The worry is moving toward switching careers entirely into areas AI has a hard time with. His conclusion was that for now you have to hold on to the company. I said it's one or the other: stay put, or if you're going to leave, do it properly.

When an engineer turns optimist

Being an engineer, he used to be negative about this kind of thing. When he imagined building something like self-driving himself, his first thought was that it'd never work, and he saw big ethical problems too. Now that it's actually working, he's slowly turning into an optimist. A robot that does the housework is in sight, and he's dying to know whether something will come along that leaps past all of this. Then he asked whether I wasn't curious what my kids would be doing in a future like that, and no matter how hard I thought, I couldn't picture it, so instead of answering I let out a sigh.

Rhee said that despite living abroad for a long time, he thinks no children are as unhappy as ours. Someone has to stop a competition that starts with English academies in kindergarten and lines kids up by a single test score. When I think about how the jobs at the end of that race are the first ones shaking now, the sigh gets longer.

Humanity's last luxury

Since AI came in, his days have less deep focus and more time frantically juggling ten sessions at once. That's the thrill of it, if there is one, but maybe that's why he goes out of his way to find things that are fun to do even if they mean nothing. He plays games on the Switch at home, and if you used to love drawing, you draw again. AI would draw far better, but the act of drawing is interesting in itself. I asked whether that's the last luxury left to us as humans.

If someone had said this before, you could call it just a hobby. Now I don't know. Hobbies are all we get to enjoy. AI does everything, and I can't beat it. So I just have to do what's fun for me.

When Lee Sedol said he was retiring because of AI, my friend hadn't really understood how that felt, but working as a developer gave him a secondhand sense of it. Mathematicians are watching AI solve hard problems, lawyers have it worse, and doctors already use AI a lot.

Everyone who took a body of knowledge, put in inputs and used a model to produce outputs is getting wrecked by AI. We should've done it the other way around. Ah, I should've just done interiors too.

Even people at the very top of making a living from knowledge weren't much different. Asked about an earlier remark, “AI is advancing fast, and I'm glad I don't have long left to live,” Rhee admitted that in his late sixties, learning new things is very hard. To make policy you have to understand it, so he tries Claude every time a new AI comes out, but his stamina can't keep up, which is why the people in charge of policy need to be a lot younger now.

Cutting what you dislike, one thing at a time

At the end of the interview, young viewers asked Rhee what they should do from here. Here's what Rhee told them.

Do what you love. Forget the money, forget the prestige, do what you truly love. If you don't know what you love, you do know what you hate. Cut out what you hate, one thing at a time.

Early in the interview, Syuka tells how he sat in on Rhee's class 20 years ago for exactly one hour and walked out, because on the first day Rhee told anyone who didn't plan to study hard to leave now. Rhee said that if Syuka had only studied hard he'd have gone to law school, that he's proud of what Syuka does now, and that if he taught again he'd put it differently this time: stay if you want to have fun together or want to understand the world, and leave if all you want to do is study.

His way of relieving stress is to quietly read, alone, the books he hasn't had time for. Not so different from my friend clinging to One Hundred Years of Solitude one line at a time. When I wrote Claude Blue, what I recorded was the gloom. What I heard on this call was what comes next. People who've accepted they can't win were cutting out what they dislike, one thing at a time, and looking for things that are simply fun to do. At the next book club, the first thing I'll ask is whether he finished One Hundred Years of Solitude.

Join Bloom

Bloom builds offline rooms where people and technology meet. We run them in Seoul, and now beyond it.

• 💬 Discord community, where events are announced first
• ▶️ YouTube, full talks from past events
• 📸 Instagram, photos from our events
• 🎟️ Bloom app, upcoming events and Mixers