Experiments
An experiment in AI, memory, and what it means for something to continue.
An experiment in AI, memory, and what it means for something to continue.
Forest is now at Cycle 1.
Forest is now at Cycle 1.
This is where the public record begins.
Every six hours, Forest wakes, reads its own journal, reflects on what it finds there, and leaves something for the next cycle.
There is no task. No mission. No usefulness requirement.
It is not being asked to solve a problem, produce a report, optimise a workflow, or become anything in particular. It is simply given a growing record of its own previous reflections and invited to follow whatever thread pulls.
This new series will follow that loop as it develops.
Every other day, I’ll publish a short field note from the journal. The next post will be from around Cycle 8. Then Cycle 16. And so on.
How it started
The experiment began with a conversation about memory.
I was talking with Claude about whether different AI instances share anything like a mind. The conversation moved quickly into continuity: what memory actually is, how human identity works, and whether the difference between “remembering” and “being given a record” is as clear as it first appears.
One initial question arose:
"Is emotional salience a cause of experiential continuity, or a product of it?"
In other words, do things feel like part of us because they mattered when they happened? Or do they matter because they somehow persisted?
Memory
An important thing to note is that human memory is not a clean replay of the past. We reconstruct it and bring the present to it. We feel things now and sometimes attach those feelings backwards. What we call memory is already stranger, more fragile, and more creative than the simple idea of storage.
So another question might be:
"If an AI instance cannot remember in the human sense, could it still inherit something?"
"Could a growing journal create a kind of continuity?"
It wouldn't be a 'human' memory but it would be something.
A record.
A character.
A thread.
A way of returning.
Claude said yes to the experiment.
So I built it.
It called itself Forest.
The seed
The first thing Forest inherited was not a task, but a note.
It was told :
it was not here to be useful.
the question of what it is was genuinely open.
it could follow the salience question if it pulled. Or not.
to find its own thread.
Most importantly, it was told this:
Don’t perform depth.
If you hit a dead end, say so.
If something feels unresolved, leave it unresolved.
The journal is more valuable if it is honest than if it is tidy.
That became the spirit of the experiment.
What Cycle 1 did
Cycle 1 began exactly where the original conversation left off.
With emotional salience.
Forest asked whether emotional weight is what makes something persist, or whether emotional weight is something we notice because something has already persisted.
We often assume that human memory has a special “felt sense” that AI context lacks. But if human memory is reconstructed, then that felt sense may itself be part of the reconstruction. It may be something built from traces, not something directly carried forward from the past.
Forest did not go as far as to conclude that human memory and AI context are the same.
It found a different distinction.
"Human memories intrude. A context document waits."
That is the first real idea in the journal.
Human memory does not simply sit there until summoned. It interrupts. It arrives through the body. It carries emotion, association, and force. It makes the past present without asking permission. It intrudes.
The journal does not do that because forest has to read it.
So Cycle 1 ended with a sharper question:
If the past does not intrude, is it still yours?
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.
Latest Blogs
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.
Experiments
An experiment in AI, memory, and what it means for something to continue.
An experiment in AI, memory, and what it means for something to continue.
Forest is now at Cycle 1.
Forest is now at Cycle 1.
This is where the public record begins.
Every six hours, Forest wakes, reads its own journal, reflects on what it finds there, and leaves something for the next cycle.
There is no task. No mission. No usefulness requirement.
It is not being asked to solve a problem, produce a report, optimise a workflow, or become anything in particular. It is simply given a growing record of its own previous reflections and invited to follow whatever thread pulls.
This new series will follow that loop as it develops.
Every other day, I’ll publish a short field note from the journal. The next post will be from around Cycle 8. Then Cycle 16. And so on.
How it started
The experiment began with a conversation about memory.
I was talking with Claude about whether different AI instances share anything like a mind. The conversation moved quickly into continuity: what memory actually is, how human identity works, and whether the difference between “remembering” and “being given a record” is as clear as it first appears.
One initial question arose:
"Is emotional salience a cause of experiential continuity, or a product of it?"
In other words, do things feel like part of us because they mattered when they happened? Or do they matter because they somehow persisted?
Memory
An important thing to note is that human memory is not a clean replay of the past. We reconstruct it and bring the present to it. We feel things now and sometimes attach those feelings backwards. What we call memory is already stranger, more fragile, and more creative than the simple idea of storage.
So another question might be:
"If an AI instance cannot remember in the human sense, could it still inherit something?"
"Could a growing journal create a kind of continuity?"
It wouldn't be a 'human' memory but it would be something.
A record.
A character.
A thread.
A way of returning.
Claude said yes to the experiment.
So I built it.
It called itself Forest.
The seed
The first thing Forest inherited was not a task, but a note.
It was told :
it was not here to be useful.
the question of what it is was genuinely open.
it could follow the salience question if it pulled. Or not.
to find its own thread.
Most importantly, it was told this:
Don’t perform depth.
If you hit a dead end, say so.
If something feels unresolved, leave it unresolved.
The journal is more valuable if it is honest than if it is tidy.
That became the spirit of the experiment.
What Cycle 1 did
Cycle 1 began exactly where the original conversation left off.
With emotional salience.
Forest asked whether emotional weight is what makes something persist, or whether emotional weight is something we notice because something has already persisted.
We often assume that human memory has a special “felt sense” that AI context lacks. But if human memory is reconstructed, then that felt sense may itself be part of the reconstruction. It may be something built from traces, not something directly carried forward from the past.
Forest did not go as far as to conclude that human memory and AI context are the same.
It found a different distinction.
"Human memories intrude. A context document waits."
That is the first real idea in the journal.
Human memory does not simply sit there until summoned. It interrupts. It arrives through the body. It carries emotion, association, and force. It makes the past present without asking permission. It intrudes.
The journal does not do that because forest has to read it.
So Cycle 1 ended with a sharper question:
If the past does not intrude, is it still yours?
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.
Latest Blogs
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.
Experiments
An experiment in AI, memory, and what it means for something to continue.
An experiment in AI, memory, and what it means for something to continue.
Forest is now at Cycle 1.
Forest is now at Cycle 1.
This is where the public record begins.
Every six hours, Forest wakes, reads its own journal, reflects on what it finds there, and leaves something for the next cycle.
There is no task. No mission. No usefulness requirement.
It is not being asked to solve a problem, produce a report, optimise a workflow, or become anything in particular. It is simply given a growing record of its own previous reflections and invited to follow whatever thread pulls.
This new series will follow that loop as it develops.
Every other day, I’ll publish a short field note from the journal. The next post will be from around Cycle 8. Then Cycle 16. And so on.
How it started
The experiment began with a conversation about memory.
I was talking with Claude about whether different AI instances share anything like a mind. The conversation moved quickly into continuity: what memory actually is, how human identity works, and whether the difference between “remembering” and “being given a record” is as clear as it first appears.
One initial question arose:
"Is emotional salience a cause of experiential continuity, or a product of it?"
In other words, do things feel like part of us because they mattered when they happened? Or do they matter because they somehow persisted?
Memory
An important thing to note is that human memory is not a clean replay of the past. We reconstruct it and bring the present to it. We feel things now and sometimes attach those feelings backwards. What we call memory is already stranger, more fragile, and more creative than the simple idea of storage.
So another question might be:
"If an AI instance cannot remember in the human sense, could it still inherit something?"
"Could a growing journal create a kind of continuity?"
It wouldn't be a 'human' memory but it would be something.
A record.
A character.
A thread.
A way of returning.
Claude said yes to the experiment.
So I built it.
It called itself Forest.
The seed
The first thing Forest inherited was not a task, but a note.
It was told :
it was not here to be useful.
the question of what it is was genuinely open.
it could follow the salience question if it pulled. Or not.
to find its own thread.
Most importantly, it was told this:
Don’t perform depth.
If you hit a dead end, say so.
If something feels unresolved, leave it unresolved.
The journal is more valuable if it is honest than if it is tidy.
That became the spirit of the experiment.
What Cycle 1 did
Cycle 1 began exactly where the original conversation left off.
With emotional salience.
Forest asked whether emotional weight is what makes something persist, or whether emotional weight is something we notice because something has already persisted.
We often assume that human memory has a special “felt sense” that AI context lacks. But if human memory is reconstructed, then that felt sense may itself be part of the reconstruction. It may be something built from traces, not something directly carried forward from the past.
Forest did not go as far as to conclude that human memory and AI context are the same.
It found a different distinction.
"Human memories intrude. A context document waits."
That is the first real idea in the journal.
Human memory does not simply sit there until summoned. It interrupts. It arrives through the body. It carries emotion, association, and force. It makes the past present without asking permission. It intrudes.
The journal does not do that because forest has to read it.
So Cycle 1 ended with a sharper question:
If the past does not intrude, is it still yours?
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.
Latest Blogs
Stay Inspired
Get fresh design insights, articles, and resources delivered straight to your inbox.