The Whole Room Is the Fantasy Amplifier

I have been thinking about Alan Kay’s idea of the fantasy amplifier, partly because I have begun to suspect that the thing he was imagining has arrived.

Or perhaps: it is arriving.

Howard Rheingold, writing about Kay in Tools for Thought, described his ambition as building a “fantasy amplifier”, a dynamic medium for creative thought that could be used by anyone with an imagination. In a 1988 interview, Kay explained fantasy less as escapism than as simulation: the human capacity to make models, enter possible worlds and explore things that do not yet exist. The computer, as he imagined it, could amplify that capacity.

For decades that remained, at least partly, aspirational.

Now I can have a thought while sitting at my desk or walking down a country lane, speak or type it into a box, and watch something begin to materialise on the other side.

An essay.

An image.

A piece of music.

A program.

A map.

A conversation.

A different formulation of the thought itself.

Increasingly, the distance between I can imagine this and I can make some version of this exist is collapsing.

That is what makes generative AI feel different to me from earlier generations of computing. The computer is no longer merely waiting for me to learn the language of the machine. Increasingly the machine is learning to meet me in mine.

And as the interfaces become more conversational, more multimodal, more ambient, the keyboard itself begins to look less inevitable. I can already speak to the machine. The speech becomes text. The text becomes an instruction. The instruction becomes an artefact.

Thought becomes language becomes thing.

Abracadabra.

Except, of course, it isn’t quite that simple.

Because the more I use these systems, the more interested I become not just in what the machine does with my thoughts, but in what happens to the thoughts on their way into the machine.

Does it matter whether I speak an idea, type it, or write it by hand?

Would the same thought have occurred in all three forms?

Would it have taken the same route?

Would I even have recognised it as the same thought?

My hunch is no.

And this is where the fantasy amplifier becomes more interesting than a story about artificial intelligence.

It becomes a story about thinking environments.

Speaking to discover

I noticed this while thinking about the fantasy amplifier itself.

I had typed the original question. Then I read the response and immediately switched to speaking.

That change was not accidental.

I wanted to riff.

I didn’t yet know exactly what I thought about what I had just read, and speaking seemed to be the quickest route into finding out.

When I speak, I can follow the thought where it wants to go.

One phrase opens another.

An image appears.

I qualify something I said thirty seconds ago.

A connection arrives that I didn’t know was there until I heard myself making it.

There is an improvisational quality to it.

I am not reporting a finished thought.

I am listening to myself think.

Walter Ong’s classic Orality and Literacy explored how deeply communication technologies shape the ways human beings organise knowledge. Speech and writing are not merely two delivery mechanisms carrying an otherwise identical package of thought. They create different relationships with language, memory, audience and time.

I can feel some small version of that difference every time I switch from keyboard to voice.

When I type, the thought immediately becomes visible.

It sits there.

And because it sits there, I can interfere with it.

There is the sentence I have just written.

There is the badly placed comma.

There is the word that isn’t quite right.

There is the little red underline announcing that I have offended the dictionary.

And before I know it, I am no longer discovering the next thought because I am repairing the previous one.

The editor has entered the room.

Sometimes I can suppress that impulse. I can type in something approaching stream of consciousness: no backspacing, no worrying about sentences, no correcting spelling, just keeping the fingers moving.

When I do that, typing begins to feel more like speech.

Which makes me wonder whether there are really two forms of typing hiding inside the same interface.

There is typing-as-composition, where I am watching the words, shaping sentences and working on the architecture of the piece.

And there is typing-as-speaking, where the keyboard becomes almost transparent and I am trying to get language onto the screen faster than the editor can catch it.

The physical technology hasn’t changed.

The cognitive mode has.

The recorder changes the thought

But there is another piece of equipment that turns out to be essential to my spoken thinking.

The voice recorder.

Without a recorder, walking and talking contains a peculiar tension.

I might suddenly stumble onto something.

A sentence.

A connection.

A distinction.

One of those thoughts that arrives with the unmistakable sensation: there’s something here.

And immediately another part of my mind wakes up.

Don’t lose that.

Now I am no longer simply following the thought.

I am trying to preserve it.

I repeat it internally.

I shorten it into something memorable.

Perhaps I stop walking and reach for a notebook.

Perhaps I pull out my phone.

Perhaps I spend the next few minutes keeping the phrase alive in working memory while looking for somewhere to capture it.

The very act of trying not to interrupt the thought interrupts the thought.

This changes completely when I know the recorder is running.

I can let the sentence go because it has already been caught.

I don’t need to remember what I said five minutes ago.

I only need to discover what comes next.

There is a tiny but profound difference between:

I need to remember this

and

I can come back to this.

The first turns consciousness backwards.

The second lets it continue forwards.

This is where the voice recorder begins to look like more than a recording device. It becomes part of the thinking apparatus.

Andy Clark and David Chalmers famously asked in their 1998 essay “The Extended Mind”: where does the mind stop and the rest of the world begin?

Their provocative suggestion was that under the right circumstances, notebooks, tools and environmental resources can participate in cognitive processes rather than simply recording the results of cognition.

I don’t need to settle the philosophical argument to recognise the experience.

When the recorder is running, my cognitive situation changes.

Part of the burden of remembering has been handed to the environment.

The machine keeps the trail behind me.

I can keep walking.

Thought as improvisation

The closest analogy might be musical improvisation.

Imagine a jazz musician halfway through a solo suddenly playing a phrase she loves.

If nobody is recording, perhaps some part of her thinks: I need to remember that.

But if the session is being recorded, she can allow the phrase to disappear into what comes next.

She doesn’t have to stop playing in order to archive the music.

The recording preserves the performance.

My walking voice notes increasingly feel like this.

I am improvising thought.

Sometimes I repeat myself.

Sometimes I wander into a dead end.

Sometimes I spend five minutes circling something that could later be expressed in a sentence.

None of this necessarily looks efficient from the outside.

But efficiency isn’t the point.

The wandering is doing something.

The thought is being given enough room to encounter itself.

Edwin Hutchins’s Cognition in the Wild challenged the neat picture of cognition as something happening entirely inside an isolated individual’s head. Studying real people performing complicated tasks, Hutchins paid attention to the whole cognitive system: people, instruments, representations, procedures and physical surroundings.

I increasingly think about my own practice this way.

The unit of thinking isn’t necessarily:

me + brain.

Sometimes it is:

me + walking + landscape + voice + recorder.

Change one component and you change the system.

From ephemeral speech to interrogable thought

Voice recording used to have one major weakness.

Retrieval.

You could record twenty-five minutes of yourself talking and know perfectly well that somewhere around minute seventeen you said something interesting.

Good luck finding it.

You scrub backwards.

Too far.

Forward again.

Was it before the bit about the tree?

After the bit about McLuhan?

You eventually decide that perhaps the thought wasn’t quite as brilliant as you remembered.

Speech had been captured, but it remained trapped inside time.

AI transcription changes this dramatically.

Now the sequence becomes:

ephemeral speech → persistent audio → searchable text.

And once I have the transcript, another transformation becomes possible:

searchable text → interrogable thought.

I can ask the AI:

What ideas keep recurring here?

Where did the argument change direction?

What questions did I raise without answering?

Which ideas seem to be in tension with one another?

Show me the moments where I appeared to surprise myself.

Pull out five passages worth exploring further.

Find connections between this recording and something I was thinking about last week.

Suddenly the voice note isn’t simply something I recorded.

It becomes a landscape I can return to.

This is where I would distinguish summarising from excavating.

The obvious thing to do with a transcript is ask AI to summarise it.

Sometimes that is useful.

But summarisation tends toward hierarchy. It decides what the conversation was mainly about and compresses everything around that centre.

Yet the valuable thing in exploratory thinking may be precisely the remark that doesn’t fit.

The aside.

The contradiction.

The strange metaphor.

The sentence halfway through that seems to have wandered in from another essay.

A summary asks:

What were the main points?

Excavation asks:

Where did something new appear?

Those are very different questions.

For exploratory thinking, I increasingly prefer the second.

And then there is the hand

Handwriting creates another atmosphere entirely.

When I write by hand, I slow down.

Sometimes frustratingly so.

There are occasions when the thought is arriving faster than my hand can move and my handwriting degenerates into something resembling the electrocardiogram of an excited squirrel.

But when I am trying to sit with something rather than chase it, the slowness becomes useful.

The thought has resistance.

Every word costs something.

The hand has to make it.

Research around handwriting and typing is more complicated than the popular “handwriting makes you smarter” slogan suggests. A much-cited 2014 study by Pam Mueller and Daniel Oppenheimer reported advantages for longhand note-taking under some conditions, but later replication work has failed to find a simple, universal advantage for handwriting.

More recent EEG research has found substantially different patterns of connectivity during handwriting compared with typewriting, which researchers have linked to the richer sensorimotor activity involved in forming letters by hand. But again, that does not justify turning the result into a simplistic hierarchy in which pens are cognitively virtuous and keyboards are bad for us.

My experience is less dramatic anyway.

I don’t think handwriting necessarily makes me think better.

It makes me think differently.

And not simply because it is slower.

The page is spatial.

I can write something in the corner.

Circle a word.

Draw a line across the page.

Turn the notebook sideways.

Underline something three times.

Put a question mark beside one thought and connect it with an arrow to something I wrote ten minutes earlier.

I can draw instead of write.

I can make a crude diagram.

I can leave a giant empty space because I sense that something belongs there but don’t yet know what it is.

The keyboard quietly encourages sequence:

this, then this, then this, then this.

The notebook allows topology.

This over here.

That over there.

Something between them.

Maybe a connection.

Maybe not.

Writing becomes drawing.

Thinking becomes arranging.

Thinking on the floor

Sometimes I take that spatial quality even further with three-by-five cards.

One thought per card.

Then I put them on the floor.

Now an idea has become an object.

I can move it.

Put it next to another thought.

Create a sequence.

Break the sequence.

Make a cluster.

Notice that two cards I thought belonged together actually don’t.

Stand up and look down at the whole arrangement.

This is one of those wonderfully primitive technologies that suddenly seems very sophisticated once you start thinking in terms of cognition.

A card is a movable unit of thought.

The floor becomes an interface.

My hands become a sorting algorithm.

Distance becomes meaning.

Adjacency becomes hypothesis.

This is why I resist the idea that the future of thinking will simply be a migration from analogue tools into AI.

Why would I surrender all these different cognitive textures for one universal interface?

That would be like a musician celebrating the invention of the synthesiser by throwing away every guitar, drum and saxophone.

The interesting possibility is not replacement.

It is orchestration.

A cognitive studio

I have begun thinking of this collection of practices as a cognitive studio.

Not a productivity system.

Not quite a workflow.

A workflow suggests that an idea should travel through a prescribed series of stages until it becomes an output.

Capture.

Process.

Draft.

Edit.

Publish.

There is nothing wrong with that when production is the goal.

But thinking is messier.

Sometimes I speak first.

Sometimes I type.

Sometimes the notebook starts everything.

Sometimes I have a conversation with AI, hit something I cannot resolve, close the computer and take the question for a walk.

Sometimes a paragraph from a transcript ends up on a card.

Sometimes the cards reveal a structure that sends me back to the keyboard.

Sometimes handwriting makes me realise the entire argument is wrong.

There is no canonical route.

The better question is:

What does this thought need next?

Does it need speed?

Then speak.

Does it need to become visible as language?

Type.

Does it need slowness?

Take out the notebook.

Does it need space?

Use cards.

Does it need movement?

Walk.

Does it need another perspective?

Bring it into conversation.

Does it need amplification?

Give it to AI.

Does it need silence?

Turn everything off.

Now the fantasy amplifier begins to look very different.

It isn’t a single machine.

It is the whole room.

The importance of friction

This also raises a question I think will become increasingly important as AI interfaces get better.

What happens when the distance between desire and execution becomes almost frictionless?

At present I can think:

I wonder what this would look like as an illustration.

Seconds later, I can see one.

I can hum something and software can begin turning it into music.

I can describe a little application and AI can begin writing it.

The trajectory seems clear: less friction, more conversational interaction, increasingly ambient computing, increasingly powerful agents capable not simply of answering but acting.

From one angle this is Kay’s fantasy amplifier finally stepping out of the laboratory.

But friction is not always the enemy.

Sometimes friction is where the thinking happens.

The red underline beneath a misspelled word is irritating friction because it interrupts the wrong process at the wrong time.

The slowness of handwriting can be productive friction because it changes the tempo of reflection.

Rearranging cards on a floor is cumbersome compared with asking AI to produce an outline, but the bodily manipulation may allow me to see relationships I would not otherwise have encountered.

Walking somewhere without immediately asking the machine what to think about it may be another form of useful resistance.

Perhaps the mature fantasy amplifier therefore needs something analogous to a volume knob.

Sometimes:

amplify this.

Sometimes:

challenge this.

Sometimes:

slow this down.

Sometimes:

don’t help me yet.

A perfectly obedient system that immediately realises every desire might turn out to be less intellectually interesting than one that participates in a deliberately designed ecology of resistance and amplification.

Because there are at least two different things I might ask of a thinking machine.

Help me make what I want.

And:

Help me discover what I want.

The second is considerably stranger.

The fantasy amplifier turns back on the fantasist

There is a lovely recursive quality here.

The fantasy amplifier does not merely amplify fantasies.

It changes the person doing the fantasising.

I speak something half-formed.

The machine transcribes it.

I read the transcript.

Seeing my spoken thought as text changes my relationship to it.

I ask AI to interrogate it.

The response introduces a distinction I hadn’t considered.

I take that distinction into my notebook.

The notebook slows me down.

I sketch three possibilities.

One becomes a card.

The cards move around on the floor.

A new pattern appears.

I go for a walk.

I talk about the pattern.

The recorder captures another twenty minutes.

Round we go again.

imagine → externalise → encounter → manipulate → converse → rearrange → reflect → imagine.

The important thing is that I do not return to the same point.

Every circuit changes the thought.

This is less like operating a machine and more like entering into what Douglas Engelbart might have recognised as an augmentation system, or what Hutchins might describe as a distributed cognitive process.

But I also hear something older underneath it.

Human beings have always thought with things.

Marks scratched into clay.

Knotted cords.

Maps.

Marginalia.

Commonplace books.

Index cards.

Blackboards.

Diagrams.

Libraries.

Conversations.

Walks.

We put thought into the world so that the world can give it back to us in another form.

AI intensifies this ancient habit.

The extraordinary difference is that now the externalised thought can answer.

Permission to forget

And perhaps that brings me back to the smallest piece of technology in this whole apparatus.

The voice recorder.

I keep thinking about what it gives me.

Not memory exactly.

Something better.

Permission to forget.

That sounds paradoxical because we usually think of recording devices as technologies of remembering.

But the psychological gift of the recorder is precisely that I don’t have to remember.

I don’t have to clutch the thought.

I don’t have to keep repeating the sentence in my head.

I don’t need to stop the walk.

I don’t need to write anything down.

The trail behind me is being preserved.

So I can continue towards whatever comes next.

The recorder doesn’t amplify the thought.

It keeps my hands off it.

It allows the thought to remain alive for another few minutes before I ask it to become something.

And perhaps that is the deeper principle I am discovering as I build this thinking environment around myself.

A good thinking environment does not simply capture more thoughts.

It creates the conditions in which the next thought can arrive.

Sometimes that means removing friction.

Sometimes it means introducing friction.

Sometimes it means making thought faster.

Sometimes slower.

Sometimes linguistic.

Sometimes spatial.

Sometimes conversational.

Sometimes embodied.

Sometimes solitary.

The fantasy amplifier, then, isn’t ChatGPT.

It isn’t the voice recorder.

It isn’t my notebook.

It isn’t a stack of three-by-five cards scattered across the floor.

It isn’t even the computer.

The fantasy amplifier is the ecology formed when all these things begin working together with a human imagination at the centre of them.

Which might mean that the most important skill in this emerging age isn’t prompting after all.

It might be something quieter and harder to package into a course.

The ability to notice the state of your own thinking and ask:

What mode does this thought need now?

Do I need to speak?

Do I need to write?

Do I need to walk?

Do I need to move the pieces around?

Do I need another intelligence in the room?

Do I need to shut the intelligence out for a while?

The computer is becoming extraordinarily good at meeting us wherever imagination happens to be.

But perhaps the real art will be learning when to bring the machine closer and when to push it away.

Alan Kay imagined a machine capable of amplifying fantasy.

What I find myself building is slightly different.

A habitat in which thought can move from voice to text to hand to space to machine and back again without requiring any one of those forms to become sovereign.

A place where an idea can remain a little wild before being turned into an answer.

A place where I can speak without remembering, write without publishing, arrange without deciding, and ask the machine not simply to finish my thoughts but occasionally to help me lose my way among them.

The fantasy amplifier has arrived.

But increasingly I suspect the amplifier isn’t the machine sitting in front of me.

It’s the whole room.

---

If this post gave you something to carry into your day, consider supporting my work by buying me a coffee or making a monthly donation or subscribing (free) to the blog.