What do we do after we beat the Ender Dragon?

The dragon as a tutorial

On September 21, 2026, OpenAI announced¹ that an internal model had "resolved more than 100 long-standing open problems across most areas of mathematics."

I thought about Minecraft.

You can spend a long time preparing to defeat the Ender Dragon. You find materials, make equipment and learn to survive in unfamiliar places. The dragon gives that preparation a goal. Defeat it, and the credits can roll².

Then the world is still there.

I think of the route to the dragon as a tutorial. The things you learned along the way can also help you build a railway or a city with friends.

Minecraft actually starts after you have beaten it.

Creative mode

But suppose an automated player does all of that for me. When I come back the dragon is dead, and I still don't know anything about Minecraft as a game. Completing the objective didn't teach me what else I could do there.

A hundred reported solutions don't finish mathematics. But take automation to its extreme: suppose machines become excellent at finding answers, explaining them and suggesting worthwhile questions.

I build research tools and spend weekends at hackathons, and I'd gladly automate the setup that keeps me from trying an idea. Minecraft already has a version of this. Creative mode² hands you the resources, so you learn to build by building.

The instrument in front of me

Suppose I need a bass sound for a track. An assistant could generate a sample, or help me build a small synthesizer. With the synth, I turn a control, hear something unexpected and ask what changed. I'm learning to shape the sound without having to write the audio software.

A sample might be all I need to finish the track and go to bed. But even a generated sample could make me curious about how it was made, and then the evening goes to experimenting.

I asked for a sound; I did not yet know I would enjoy making sounds.

I can't always describe that interest in advance. Hidi and Renninger⁵ describe a passing interest growing through further engagement and support. A tool could get me far enough into sound design to find out whether I enjoy it, without deciding that for me or turning every request into a lesson.

AI tutoring studies separate getting the right answer with help from doing the next problem alone. Students with a less restricted tool⁴ did better on practice and worse on the exam without it. A carefully designed tutor⁴ beat the classroom lessons it was compared with. With the synth, I want to get better at shaping sounds, and continued help with the software lets me do that.

Widening the cone

Imagine a cone containing the questions I can realistically investigate with my current knowledge, resources and time. A useful tool could widen it across fields and deepen it within an unfamiliar one.

Some questions mean comparing thousands of papers, more than I could read in a useful timeframe. A system that organizes them and keeps their sources attached lets me follow a connection without holding the whole archive in my head. Don Swanson called what's waiting there undiscovered public knowledge⁶: findings published separately that support a discovery once someone connects them. The connections still need checking.

Something too difficult to try can become an interest once I have a way into it. So automation isn't a machine working through a fixed list of things humans want. Completing one request can give me a reason to make another I could not have thought of beforehand.

Daniel Litt³ distinguishes producing mathematical results from human understanding of them. A proof can be published and still be new to me. An assistant helping me understand it could hand me a question I want to investigate.

Chosen constraints

Not every difficulty is in the way.

On the drive to a trailhead, a bypass around the bends leaves more time for the walk. On a Group B rally stage, the bends are what the driver and co-driver came for; remove them and they are competing at something else. Bernard Suits⁷ calls these the traveler and the racer.

Some friction is part of an activity I chose, and some just gets in my way. I can enjoy experimenting with a recipe and still leave the whisking to a mixer.

Nor does every activity need to teach me something. I enjoy music more with friends, even a recording I could play alone; their reactions are part of it, and studies of live music⁸ link that shared feeling to enjoyment and meaning. I would call the playful part of this whimsy: dancing at a rave, a ridiculous game made up with friends, no need to get better at either. I do not want freedom from work to come with an obligation to spend every spare hour improving ourselves.

The world we actually have

Automating someone's job does not mean they can afford to stop working. Someone searching for replacement income is in a different position from someone whose hours fell without a pay cut.

The same goes for learning at work. If software takes over a report a junior employee used to prepare, their employer could have them dig into the results with an experienced colleague. Or the employer could decide the junior position is no longer needed.

There could be more projects worth trying than ever while someone has less money or time to try them. "People can still do it for fun" does not answer that.

I would love more time to explore. Someone else might mainly want fewer demands on their day. Pope Leo XIV's Magnifica Humanitas⁹ rejects measuring a person's dignity by their abilities or achievements. Nobody should have to prove they are still useful by replacing their old job with some impressive new pursuit.

Bounded infinities

Between two distinct rational numbers lies another.¹⁰ Take the midpoint, then subdivide again; there is always another number, and the interval never grows. Learning the basic rules of a subject works like that for me. It lets me look closer rather than meaning I'm done with it.

Jevons¹¹ argued that burning coal more efficiently could increase how much of it gets burned. Inquiry might go the same way as it gets cheaper: an afternoon instead of a month makes new questions worth asking. Who decides how the saved time gets spent is still open. I could attempt something new, or my employer could expect more of the same work.

After the dragon

After the dragon, I might build a railway that takes longer to construct than walking to its destination ever would. A machine could lay the track while a friend and I design the station. We might get carried away and start building a town around it instead. By the end of the afternoon, reaching the original destination could be the least interesting part.

When does getting distracted become finding something worth doing?