This post was inspired by a run and an episode of Fall of Civilizations — The Khmer Empire: Fall of the God Kings by Paul Cooper. If you haven't listened to it, it's one of the best podcasts I've ever heard.

The Khmer built the largest city on earth. A million people. More stone than all the pyramids of Egypt combined. An ingenious water system that fed the entire empire. And then the ground shifted — drought, invasion, a new religion that quietly said "actually, the god-king isn't a god" — and Angkor emptied out.

But the Khmer didn't disappear. They moved. Adapted. Built something new from what they knew.

Every great civilisation seems to go through something like this. The thing that made them powerful eventually forces them to change. And they do. Or at least that's what we tell ourselves.

So when people ask "if AI can do everything, what's left for us?" — I think they're asking the wrong question. But I also don't think the comfortable answer is the honest one.

The Scramble for Higher Ground

There's a particular anxiety in the air right now that I keep seeing in forums, in conversations, in the comments under every AI announcement. It goes something like: the faster I adapt, the faster everything changes. What's the point?

Developers are watching coding agents handle 80% of their output. Karpathy wrote about going from writing most of his code manually in November 2025 to delegating 80% of it to agents by December. A few weeks. That's it.

And the natural response is panic. If the tool can do that, what do I do?

But I think the panic comes from watching the god-king fall and assuming everyone falls with him.

The Jevons Paradox: More Efficiency, More Things Worth Building

There's a concept called the Jevons paradox that's useful here. When steam engines got more fuel efficient in the 1800s, coal consumption didn't go down. It exploded — because cheaper energy unlocked entirely new uses nobody had imagined before.

The same thing seems to be happening with software right now.

AI makes building certain things dramatically faster. Internal tools. Prototypes. Automations. Things that used to need a team of five for six months can be roughed out in a week. Projects that sat on the "nice to have" backlog for years suddenly become viable.

The World Economic Forum's Future of Jobs Report 2025 projects that while 92 million roles will be displaced by 2030, 170 million new ones will emerge — a net gain of 78 million jobs. That's not a guarantee things will be fine. But it does suggest the total surface area of things worth building hasn't collapsed. It's probably expanded.

What the Khmer Actually Teach Us

Here's the part of the Angkor story that tends to get lost in the "lost city" narrative.

Angkor was never actually abandoned by everyone. When Portuguese missionaries arrived in 1586 and found it "lost," Buddhist monks were still there. Performing rituals. Maintaining shrines. Doing their work among the crumbling walls.

The god-kings were gone. The empire was gone. The elaborate bureaucracy and the hydraulic engineers and the Sanskrit-speaking elite — gone. But the monks kept showing up.

It seems like not everyone needed to be a god-king to survive the collapse of the empire.

From what we can tell, the people who adapted weren't the ones who doubled down on the old power structure. They were the ones who kept doing meaningful work at a human scale while everything else was falling apart.

The part nobody wants to say out loud

Angkor didn't gracefully evolve into something new. It collapsed. A million people scattered across the landscape. The infrastructure failed faster than anyone could respond. Tending ruins is not the same as building something new.

The numbers are hard to sit with. The IMF estimates that 60% of jobs in advanced economies face meaningful AI exposure right now. Entry-level roles are shrinking fastest — employment for software developers aged 22 to 25 has dropped 20% since late 2022. Anthropic's own CEO has warned that AI could eliminate half of entry-level white-collar jobs within one to five years. And 41% of employers surveyed by the WEF already plan to reduce their workforce by 2030 in areas where AI can automate tasks.

The net numbers might be positive over decades. But the people living through the transition don't experience net numbers. They experience specific jobs disappearing in specific towns in specific years. The weavers who lost their livelihoods during the industrial revolution didn't experience it as a chapter in a story with a good ending. They just lost their livelihoods.

This one might be different in one important way. Every previous tool augmented what humans could do with their hands or their time. This one might be beginning to replace certain kinds of human thinking entirely — or at least that's what it looks like from here. Not just execution — judgment, writing, analysis, code. The things knowledge workers built careers around.

"Humans adapt" is true the same way "things work out" is true. Technically. Eventually. Not always for the people in the middle of it.

What This Actually Means If You Build Things

Karpathy made a point in his No Priors conversation that stuck with me. The bottleneck moved. It used to be typing speed. Then coding skill. Now it seems to be shifting toward your ability to specify, direct, and verify. Knowing what to build and whether what was built is actually correct.

That's harder to automate. Maybe impossible. Nobody's really sure yet.

Workers with demonstrable AI skills are already earning 25% more than peers without them. Not because they're doing something magical — but because they seem to have figured out how to direct the tools rather than compete with them.

From what we can tell, the Khmer who survived weren't the ones who out-engineered the drought or out-fought the Thai kingdom of Ayutthaya. They were the ones who understood what still mattered when the infrastructure failed.

I build things solo. I teach myself everything. None of that is god-king territory. But it's real, it's useful, and it probably compounds over time.

I don't have a clean answer. I'm not sure anyone does right now who's being straight with you.

But the alternative — waiting for the ground to stop moving before you start building — isn't really an option. The monks didn't wait for a new empire to appear before they showed up to the temple. They just kept doing the work that felt worth doing.

Maybe the ground has always been moving. Maybe we just got used to the speed.

References