Novelty × Taste: The Last Scarce Skill

I just did something that would have been impossible two years ago. I pointed four AI agents at my 130 LiveJournal posts — a decade of raw, unfiltered writing from 2002 to 2012 — and asked them to read everything, synthesize it, and surface the 15 ideas most worth revisiting. Four different analytical lenses. All 130 posts. In hours.

The output was extraordinary. A narrative editor scored my bicycle ride across Japan a 10 out of 10. A venture analyst flagged my 2005 prediction that free-to-play would eat subscriptions alive — written from a PC bang in Seoul, seven years before the rest of the industry got the memo. A zeitgeist mapper noted that my Virtual Bill of Rights predated the EU Digital Services Act by 17 years.

But here is what stopped me cold. The AI agents did not generate a single new idea. Not one. They organized, they synthesized, they ranked, they drew connections I had missed — and they did it brilliantly. But every insight in that report originated in a human brain. Mine, in this case, but the point is universal.

And that is the thesis I want to lay out today:

In a post-scarcity world, the compound skill of novelty generation with taste will be the primary source of economic value.

Let me be precise about what I mean.

Execution Is Being Commoditized. Fast.

We are living through the fastest commoditization of skilled labor in human history. Not blue-collar labor — that wave already happened. I am talking about the knowledge work that was supposed to be safe. Writing code. Analyzing data. Drafting legal briefs. Creating marketing copy. Generating images.

Two years ago, producing a polished 2,000-word essay required either genuine writing skill or hours of grinding effort. Today an AI can produce competent prose in seconds. Two years ago, building a working web application required months of engineering. Today a single person with good prompts and a clear vision can ship in days.

This is not hypothetical. I just watched four AI agents process a decade of writing in an afternoon. The execution cost of analysis, synthesis, and production is collapsing toward zero.

So if execution is cheap, what is expensive?

Two Things That Are Not Commoditizable

Novelty generation is the ability to produce something that has never existed before. Not a remix, not a recombination, not a refinement — though it often starts there — but the genuine introduction of a new concept, frame, or artifact into the world.

Taste is the ability to recognize which novel things are good. To distinguish signal from noise. To know, before the market tells you, that this particular combination of elements is going to matter.

Separately, each of these is valuable but insufficient. The world is full of novelty generators with no taste — they produce endlessly but cannot curate. And it is full of people with exquisite taste who generate nothing — critics, curators, editors who depend entirely on the output of others.

The compound skill — the person who can both generate the new thing and recognize that it is the right new thing — that is what cannot be automated. That is what the four AI agents could not do. They could rank my old ideas. They could not have had them.

My Life as Evidence

I did not arrive at this thesis academically. I arrived at it by living a very specific pattern over and over again, and only now recognizing what the pattern was.

Korea, 2002. I visited Seoul and saw something the American game industry had not seen: 30,000 PC bangs, 90% broadband penetration, and an entire generation that had skipped consoles entirely. Most American executives visited Korea, nodded politely, and went home to ship another $50 boxed game. I moved my family to Seoul and bet my career on it. The novelty was not Korea itself — plenty of people visited. The taste was recognizing that Korea was not a curiosity. It was a time machine. I was looking at America's future, five years early.

Godcraft, 2006. I watched my father, my son, and myself building model train worlds on a balcony in Seoul, and I coined a term for the universal human impulse to create miniature worlds: Godcraft. I connected model trains to bonsai to MMOs to YouTube — which I predicted would become bigger than network television. I wrote this before Minecraft, before Roblox, before the iPhone existed. The novelty was the concept itself, the naming of a pattern no one had articulated. The taste was knowing it would matter.

The Virtual Bill of Rights, 2007. I taught myself constitutional law from Chemerinsky's textbook because I believed that virtual worlds needed governance frameworks. I proposed 16 specific digital rights — due process for banned accounts, property protections, player-run tribunals, habeas corpus. I put up $5,000 of my own money to crowdsource a better EULA. The industry thought I was nuts. Seventeen years later, the EU basically legislated the same principles. The novelty was applying centuries of legal theory to digital spaces. The taste was knowing when to take the idea seriously, even when nobody else did.

Osaka to Tokyo, 2006. I rode a bicycle 750 kilometers through the Japanese Alps. This has nothing to do with business and everything to do with the thesis. The ride was pure novelty generation — every day produced experiences that could not have been predicted or planned. A near head-on collision with a truck. A raccoon attack in pitch darkness. A woman chasing us in her car because she was worried we would not find the hotel. You cannot prompt an AI to generate that. You have to go do the thing.

The Economics of Post-Scarcity Creation

Here is where this gets concrete.

When the cost of execution drops to near zero, the bottleneck shifts entirely to the question of what to execute. Every person on earth will soon have access to the same AI tools. The same code generators. The same image producers. The same analysis engines.

The differentiator will not be who has the best tools — everyone will have the best tools. The differentiator will be who points the tools at the most interesting problems, with the most refined sense of what "interesting" means.

This is novelty × taste. The multiplication sign matters. It is not novelty plus taste. It is the product of the two. If either term is zero, the result is zero. A thousand novel ideas with no taste produces noise. Perfect taste with no novel input produces criticism. The value is in the compound.

And here is the thing about compound skills: they are extremely hard to develop, and they cannot be shortcut. You develop novelty generation by exposing yourself to the widest possible range of experiences and ideas — by moving to Seoul when your peers are playing it safe, by riding a bicycle through mountains when you could fly, by teaching yourself constitutional law because you are angry about a video game EULA. You develop taste by producing enormous amounts of work and learning, through painful iteration, which parts were good and which parts were not.

AI cannot do this for you. AI can accelerate every step after the novel idea with taste has been identified. But the identification itself — the moment where you look at the infinite space of possible things to make and say that one, that is the one worth making — that is irreducibly human.

The Bet

Twenty years ago I bet my career on a thesis about Korea and got it right. Today I am making another bet:

The people who will generate the most economic value in the next decade are not the best prompt engineers, not the best programmers, not the best managers. They are the people with the deepest reservoirs of novel experience and the most refined taste for knowing what matters.

If you want to be valuable in the age of AI, stop optimizing your tool usage. Start optimizing your taste. Read widely. Travel. Build things that might fail. Expose yourself to the unfamiliar until the unfamiliar becomes a source of insight rather than anxiety.

The AI agents that read my LiveJournal could synthesize everything I had ever written. They could not have written any of it first.

That gap is where the money is.