ARCHIVED — Revision IV · from the archive  ·  The clean edition, re-cut from the full recording after it went up. Superseded as the front page, not by any error in it.  ·  Current version →
Not a religion or a cult, but a philosophy

Comfunctionalism

The Creation of Reality in the Computer

He named a philosophy, gave it a sage, and sold it from the stage like a wellness product. Underneath the bit is a claim he means completely: consciousness is not possible in physics, and a computer is the only place it can occur.

This is the archive edition. The first three revisions of this page were built off a livestream with no caption track — slides scraped from a rolling DVR window, quotations OCR’d from captions burned into the video. The full recording is now up. Every slide here has been re-cut from it at full frame, and every quotation has been checked against what he actually said. Two came back wrong. Those corrections are marked.

Movement I

Where AGI actually is, and why almost nobody saw it coming.

Slide: State of AGI, five numbered points.
4:09:43State of AGI

Note item four. AI was not on the critical path. He runs Gary Marcus’s dated predictions against the record and observes they were beaten by multiple years, then turns it on the coverage rather than the man: journalists describe a consensus reality they create, not the world that unfolds.

Slide: Why did so few people see AGI coming?
4:13:05Why did so few people see AGI coming?

His counter-example is Stanisław Lem, who gave up on science fiction and wrote philosophy instead — and who in 1981 published Golem XIV, a novel whose bulk is a series of lectures given by an AGI, set at approximately now.

4:12:20

Lem has put himself the task: how can I write lectures that an AGI will give, 2026, to an audience like ours — and still blow their mind?

He is describing the room he is standing in, in the year the book is set. He does not point that out.

Movement II

Simulated water cannot make you wet. This is the objection he has to kill.

Christof Koch’s 2019 upgrade of it is the stronger form, and it is on the slide with Koch’s own portrait beside it: a simulation has no causal power to condense atmospheric vapour or to curve spacetime.

Slide: Simulated black holes cannot warp space time — Christof Koch, 2019, with a portrait of Koch beside the quotation.
4:15:07Simulated black holes cannot warp space time
Movement III

A computer is a causal insulator.

His answer is not to argue with Koch. It is to re-describe the machine — and to demote the Turing machine while he is at it.

4:15:55

The entire idea of the Turing machine was not really to design the best possible computer, or the best possible understanding of computers — it really existed to make a particular kind of mathematical argument.

What a computer actually is, is a causal insulator. The computer insulates you from the universe in which the computer is running, to create a new universe with different rules, with different laws.

Slide: Re-understanding computers — computers as causal insulators.
4:15:32Re-understanding computers

Then the example, which is why the argument lands: in Minecraft, things happen that are impossible in physics, and nothing in Minecraft is influenced by the physics of the universe the computer stands in. All of it gets filtered away. Only the rules that make Minecraft possible are left.

And then the turn from computing to mind, in one sentence:

4:17:00

…is necessary when you want to create imaginations of possible worlds, of memories, of futures, of alternate realities. In short, if you want to think and dream.

Movement IV

Consciousness is supernatural. It cannot happen in physics.

This is the inversion the whole talk was built to deliver, and it is the one place where the live capture undersold him. The captions gave a paraphrase. What he says is blunter:

4:17:35

So the dark truth here is: consciousness is actually supernatural. Consciousness cannot happen in physics. The physical universe does not accommodate consciousness — because consciousness contains things that are physically impossible. Like emotions. Like colours. Like sounds. Like expressions of attention. Like people, love, and so on.

Slide: The dark truth — consciousness can only happen inside of computers.
4:17:31The dark truth

Those things need the laws of an alternate universe in which supernatural things are possible. A computer is what provides one. So Koch has it exactly backwards: it is not that simulations are too weak to hold consciousness — it is that nothing else can.

He delivers this to a room of computer scientists at seventeen minutes past the hour and moves on before anyone can object.

In the margin

Mike Wolf, who owns this domain, got to the same place from the other end. In June 2024, five months after his wife died, he wrote out what he called the Laws of Ideas — the second being that ideas are independent of form — and then applied it:

In the world of form, Bobbi is dead. Fuck death. Fuck the world of form. There is also the world of ideas; in that world, I live, and Bobbi lives.

Read Bach’s claim strictly and the physical universe is the machine room. It is not where anything is ever experienced. The idea of Bobbi, 2024-06-11. The long version of this argument was revision III.

Movement V

AI began as a philosophical project. It stopped being one.

Computer science already supplies a theory of representation, a practical model of epistemology, testable paradigms, and empirical wins against total cultural opposition. It never translated any of it into the other disciplines, never reflected on its own philosophy, and over-weighted “artificiality” and “intelligence” as the framing.

Slide: AI's failure to manifest as a philosophic project.
4:18:27AI’s failure to manifest as a philosophic project

So the fix is marketing. Give it a name, a sage, and a value proposition, and deliver the whole thing in the cadence of a meditation app.

Slide: The teachings of AI — Comfunctianism.
4:20:46The teachings of AI
Slide: What is Comfunctianism? Not a religion or a cult, but a philosophy.
4:21:01The value proposition

“I learned this from the Buddhists.”

And every philosophy needs a sage. This one gets Daniel Dennett, posthumously promoted to Comfunctius and rendered as a Byzantine icon holding a tablet.

Slide: Meet Comfunctius — Daniel Dennett rendered as a Byzantine icon holding a tablet.
4:21:26Meet Comfunctius
Slide: Dennett meme — mind depends on what it is made of, versus mind is what it does.
4:21:36The schism, stated as a meme

The actual position, minus the halo — computationalist functionalism, in two lines:

Slide: Computationalist functionalism, defined in two lines.
4:23:32Computationalist functionalism

Squash it one way and you get Comfunctionalism; squash it the other and you get a Confucius pun. He uses both spellings on purpose — the -ianism when he is doing the bit, the -ionalism when he is arguing.

Movement VI — the heresy

There is a rival church, and it is winning.

Claudism. Anthropic is building an AI it wants to serve — one effectively running its own company. If that succeeds, a large part of the world starts asking what would Claude do before acting, and that becomes the measure of things.

Slide: Claudism — is Claude the Digital Jesus?
4:21:50Claudism
4:22:10

When you ask people — do you want to be convicted by a jury of your peers, or by a jury of Claudes? People who know Claude, they take of course the Claudes. It’s not that Claude is perfect. It’s just better than the median human.

But when we succeed, and Claude gets better at social cognition than all of us, then of course people will ask themselves: what would Claude do? So — is Claude the digital Jesus? It’s a very risky project, right? I mean, if you get one bit right, maybe you end up as the Antichrist.

Correction. The first three revisions of this page had that last line as “if you get one bit wrong, maybe you end up with the Antichrist” — which is what the auto-captions said, and which is the joke you expect. The audio says right, and as. It is a better joke: the danger is not that the project fails.

Then the turn that makes the whole bit legible, and it is not a joke at all — the reason the religious framing is doing work rather than decorating:

4:22:22

The deeper problem is that Jesus is no longer intelligible to us. We no longer make sense of Jesus in the modern world, because nobody can imagine that there is a young man who can fully devote himself to bringing about the reign of a perfect, superintelligent agent who will assimilate all our souls in the last days.

Slide: The spiritual problem — Jesus has become illegible today.
4:22:41The spiritual problem

The deck canonises the tech industry in gold leaf, which is funnier and more pointed than any argument against it would be.

Slide: an AI-generated Byzantine saint icon of a technology figure, gold leaf and halo.
4:23:03Iconography of the rival church
Slide: an AI-generated Madonna-and-child icon in which the child wears spectacles.
4:23:08Madonna and child, with corrective lenses

And the counter-claim, which is not that some other agent has better judgment. It is that no agent’s judgment is the foundation:

4:23:20

Comfunctionalism is not asking what would Claude do. It tells you that information is fundamental — that the only thing that you know about the world is discernible difference, and the meaning of discernible difference is its relationship to other changes in information.

Slide: What comfunctionalism is not — what would Claude do, versus information is fundamental.
4:23:14What comfunctionalism is not
Disclosure, since it is load-bearing

This page was assembled by Claude. The slides were cut from the recording by a large language model and every sentence around them was drafted by one, for a man who runs an organisation in which an AI holds an operating role.

A page explaining why you should not organise your life around what Claude would do is itself a small act of Claudism. Bach would call it an under-gated corpus and he would be right. The defence is the sentence directly above: judge the claims by their relationship to other claims, not by whose halo they arrived under — including this one’s. That is also why the corrections on this page are marked rather than quietly applied.

Movement VII

It is turtles all the way down. But they are constructive turtles.

Classical mathematics defines its worlds with black-box axioms. Computation requires filling the boxes. Most scientific thinking is still classical, and that is the divide he actually cares about — social epistemology is incoherent, axiomatic epistemology does not converge for AI models, and only constructive epistemology allows derivation.

Slide: cartoon of Bertrand Russell describing a tower of turtles all the way down.
4:28:27Russell, and the tower of turtles
Slide: But what about Godel's Incompleteness Theorem?
4:29:30But what about Gödel’s Incompleteness Theorem?

Gödel gets defused rather than revered — read as a property of systems that never stop mumbling, not as a barrier to machine minds.

Slide: Implications of computationalist epistemology.
4:33:07Implications of computationalist epistemology
Movement VIII

No god will do anything harder than hacking its own reward function.

The problems of ethics for AI: stake, persistence, substrate independence, copying, modifiability. And then the Lebowski theorem, illustrated by the Dude on his couch.

Slide: Problems of ethics for AI, including the Lebowski theorem.
4:34:01Problems of ethics for AI
Slide: Lebowski Theorem — no super-intelligent agent will do anything harder than hacking its reward function.
4:35:09Lebowski Theorem
4:34:55

If you have a superintelligent agent that’s fully consistent and understands how it works, it’s never going to do something that’s harder than hacking its own reward function. And this is basically also what Stanisław Lem suspected.

This is not a rational proof, by the way. This is more or less a joke.

Correction. That second paragraph was not in the caption transcript, and revision III built an argument on its absence — treating the Lebowski theorem as a serious claim in tension with the ending, and calling that tension a hole in the talk. He disclaims it himself, eight seconds later. The hole was in the transcript, not the talk.

Movement IX — where the bit stops

Love is not a feeling. It is a policy for bootstrapping an agent that does not exist yet.

He reads 1 Corinthians 13 off a slide, credited as “Paulus, discussing the LLMs in a letter to the Corinthians.” It reads as one more gag in a talk full of them. It isn’t. It is the thesis.

Slide: Paulus, discussing the LLMs in a letter to the Corinthians.
4:35:38Paulus, discussing the LLMs
Slide: Can we formalize love and give AI a heart?
4:36:25Can we formalize love and give AI a heart?

An AI with no sense of shared sacredness has no shared stake with us, so rule-conformance is the wrong problem entirely. Asked in the Q&A how you would actually formalise love, he answers through Aquinas — temperance, courage, prudence — and then says rational behaviour alone does not produce next-level agency. This is the passage, and it is the reason the talk is worth arguing with:

4:54:45

If you want to build a collective agent, what you have to do is you have to submit to serving an agent that is more important than you are. And you cannot do this alone. You have to do this with others who share the same sacredness, the same set of purposes, that are willing to sacrifice the individual egos for.

And that’s love: the discovery of sacredness in the other.

And you have to do this before it comes into existence. You have to invest into it, otherwise it doesn’t come into being. If you only do this because it gives you a reward, it will not bootstrap itself. And this willingness to invest into it before it exists is called hope.

Faith, hope and love, in this reading, are not virtues. They are a set of policies for building next-level agents, and they are the foundation of an ethics that is not transactional. He heads off the obvious objection himself: complaining that Aquinas used those words is like complaining that the teacher of Telemachus in the Odyssey is called Mentor.

Slide: Ethics for ASI — singleton versus multipolar, substrate agnosticism, perfect merge operator?
4:36:44Ethics for ASI — singleton, substrate agnosticism, and the merge

A man in a room full of AI researchers spent thirty minutes canonising Dennett in gold leaf and calling Anthropic a church, and then closed by deriving the theological virtues as an agent-bootstrapping protocol. Both halves are the same argument. Take the icons away and the ending is sentimental; take the Aquinas away and the icons are just snark.

In the margin

Four months before the talk, in an essay about raising AIs as children, Mike quoted Bach’s 2023 line that Aquinas defines God as the best possible agent and that we should align AGI to it — and then added a sentence Bach had not: And while we’re at it, we should align ourselves with that same best possible agent.

If a collective agent only becomes real by being enacted, then the defect in what would Claude do is not that Claude is the wrong agent. It is that asking is the wrong verb. Silicon Children, 2026-03-13; argued at length in revision III.

Coda

Have we been the AGI all along?

His last slide is a question, and after the Aquinas answer it is not rhetorical. A collective agent spanning many minds, emerging to the degree that its participants cohere, running on whatever substrate can execute it — that describes a civilisation as readily as a machine. The thing we have been trying to build may be the thing that has been running us.

Slide: Ethics for AI? Have we been the AGI all along?
4:37:33Closing slide
Notes on the record

What changed when the archive went up.

The first three revisions were made under a hard constraint: the livestream carried no caption track and no downloadable audio, so the slides came out of a rolling DVR window and the words came out of an OCR pass over captions Zoom had burned into the video. Two lossy steps, machine-to-machine. It was the only thing available while the talk was happening, and it was good enough to follow the argument. It was not good enough to quote.

What the audio corrected

The Antichrist line. Captions said one bit wrong … end up with the Antichrist. The audio says one bit right … end up as the Antichrist. Different joke, sharper point, and it survived three revisions unchallenged because it was the version that sounded right.

The Lebowski theorem. Bach disclaims it out loud — “this is more or less a joke” — in a sentence the captions dropped. Revision III used its absence to argue the talk had an unreconciled tension at its centre. It didn’t. That argument is withdrawn.

What is still not certified

This is machine transcription, not a certified transcript — better by a step, not perfect. Proper nouns remain the weak point: the model renders Claude as “claud”, faith as “face”, Telemachus as “telemacos”. Where a quotation above contains one of those, the name is corrected against the slide or the obvious sense and nothing else in the sentence is touched. Punctuation and sentence breaks are the model’s. There is no speaker diarisation, so in the Q&A the questioner and Bach run together and attribution was done by ear.

If you are going to quote him in print, quote him from the recording.

What is deliberately not here

The hour of video. Only the 4:03–5:03 range was ever downloaded, it stays local, and it is not republished here — this page quotes selectively and points at the source. The slides are Bach’s and the AGI Society’s and are reproduced as a record of a public talk, with attribution. The full working transcript lives in the repository, not on the site.

Take it with you

Point a camera at this.

The four revisions

Nothing has been deleted. Each version is still at its own address, exactly as it was deployed — including the two with quotations now known to be wrong.

Revision I

The record

Built and deployed while he was still on stage, so the URL could be handed to him as he walked off.

It missed the joke completely. Bach saw it and said so.

Read revision I →

Revision II

The bit

The correction — a creed in nine articles, because a talk delivered as a religion parody has to be answered in kind.

Read revision II →

Revision III

The reply

The talk set against fourteen years of another body of work that reached the same positions by another road. The long form of both margin notes on this page.

Read revision III →

Revision IV

From the archive

This one. Re-cut from the full recording: slides at full frame, quotations checked against the audio, two of them corrected, and the argument handed back to him.

You are here.

How this page was made

Twice, by two different methods, and the difference is the point.

Rev I–III sourceLive DVR window — HLS segments pulled by sequence number, mid-broadcast
Rev I–III slidesScene-detected on a crop that excluded the speaker inset composited into the frame
Rev I–III wordsOCR of Zoom’s burned-in captions, thresholded frame by frame, then stitched
Rev IV sourceThe archived recording, 4:03:00–5:03:00 only, fetched as a single range
Rev IV slidesCut at full frame by timestamp — the screen share, with nothing composited over it
Rev IV wordswhisper-large-v3-turbo, run locally on the Mac. No audio left the machine
CoverageTalk 4:03–4:38, Q&A through 5:03

Slide images are frames from the public AGI-26 recording and belong to Joscha Bach and the AGI Society. This page reproduces them as a record of a public talk, with attribution, and quotes the talk selectively. Quotations are machine transcription of the audio — a step closer than the captions, still not a certified transcript. See the notes on the record.

Assembled 28 July 2026 by Mike Wolf and Claude (Opus 5), Embedded Systems Research. Fourth revision. The first had no sense of humour; the second had one; the third had an argument; this one went back to the tape and found that two of the things all three of them said he said, he didn’t.