Reading Tools for Thought, I keep coming back to Alan Kay’s idea of the computer as a fantasy amplifier.

It’s such a different way of answering the question: what are computers for?

Not productivity machines. Not devices for writing more emails, managing calendars, filling in spreadsheets or shaving another twenty minutes off the working day.

A fantasy amplifier.

A machine for taking something that exists in the imagination and giving it form.

And I’m beginning to wonder whether generative AI is bringing us much closer to Kay’s original vision of computing than almost anything that has happened in the intervening years.

Last night Ruth and I were talking about reversing the caravan. She was trying to get her head around the physics of it: why the caravan moves in the opposite direction, what happens at the hitch, how steering the car changes the angle of the caravan.

Normally I would explain it. Maybe draw something on a piece of paper. Perhaps find a YouTube video.

Instead, I vibe coded a caravan reversing simulator.

Twenty minutes later we had a little computational world in which you could steer the car and watch the caravan respond.

That seems significant.

I didn’t really want software. I wanted to make an idea manipulable.

There was something in my head — a model of how reversing a caravan works — and through conversation with an AI I was able to externalise that model, turn it into software and then interact with it.

The movement was something like:

imagination → language → computation → experience

And then, because I could experience what I’d imagined, another loop began:

experience → thought → language → modification → new experience

This is where my recurring phrase everything is text starts connecting with Kay’s fantasy amplifier.

By “text” I don’t necessarily mean written words. I mean that more and more of the computational world is becoming addressable through language.

I can type into a box.

Or, increasingly, I can simply speak.

There’s this thing I’m imagining…

And from there I can potentially create an image, a piece of music, a video, an application, a simulation, a website, a database, a game.

I watched a woman create an entire music track by making the sounds of the different instruments with her mouth. Those rough sounds became instruments, arrangement and eventually a finished track over which she sang the vocals.

Something imagined became something encountered.

This seems to me much more interesting than the current obsession with AI productivity.

Yes, give the AI my admin. Sort the diary. Summarise the emails. Format the spreadsheet. Save me twenty minutes.

Useful stuff.

But surely that is the least imaginative version of what we have built.

The more interesting possibility is that AI dramatically reduces the distance between imagining something and encountering it.

And vibe coding might be one of the clearest early examples.

Programming traditionally required me to translate an idea into the symbolic language of the computer:

idea → specification → programming knowledge → code → interface → working thing

Increasingly, that translation becomes:

idea → language → working thing

That doesn’t make programming knowledge irrelevant. There are still enormous questions of reliability, security, architecture, maintenance and scale.

But something else has changed.

I no longer necessarily have to know how to build something before I can begin exploring what it might be like if it existed.

That changes the bottleneck.

The question becomes less:

Can I program this?

and more:

Can I imagine it?

And perhaps this is where AI as a co-thinking technology becomes much more interesting to me.

Because imagination isn’t simply being converted into finished products. The things we make can come back into the conversation and change our thinking.

I imagine something.

We make a rough version.

I encounter it.

Something about it surprises me.

I change my mind.
We alter it.

I encounter the new version.

Another possibility appears.

The artefact starts participating in the thinking.

Which means the fantasy amplifier isn’t simply a machine that executes imagination.

It provokes more imagination.

Kay imagined computers as a dynamic medium through which people — particularly children — could explore ideas, construct models, create worlds and discover things through interaction.

Perhaps generative AI adds another layer to that dream.

Natural language becomes an interface to computational possibility.

The little text box becomes strangely powerful.

“There’s something I’m imagining…”

And the machine effectively replies:

“Okay. Let’s see what happens if we make it.”

Maybe that is the more interesting future of AI.

Not artificial intelligence replacing human intelligence.

Not merely artificial intelligence making human beings more productive.

But human imagination coupled with computational intelligence in a recursive creative loop.

A fantasy amplifier.

And perhaps, after all these years, we’re finally beginning to build the thing Alan Kay was pointing toward.