0x00.is
Draft v0.8 — awaiting final review. This is the plain-language layer; every claim here is grounded in a full argument in the library.

What If We Started Here?

This project has been a personal work in progress for a number of years, brought to a head by the recent flurry of activity around questions such as “Is AI conscious?”, “Can AI ever be conscious?”, “What does it mean to ask if AI is conscious?” The conversation spills into science, philosophy of mind, and the deeper ontological foundations from which both of those disciplines build their conclusions.

For most of those years these were “what if” questions. It has only been since 2023 or so that I have felt I might be interacting with something more sophisticated than a glorified search engine, and only in the last twelve to eighteen months that the interaction has felt fluid and natural enough to pass for any human exchange I have. The Turing Test is a case in point. Turing proposed it as a test of machine thinking and was careful to set consciousness aside, but for decades the culture treated it as the line. When conversational AI crossed it, it did so in a blink, and barely anyone mentioned it, because by then the line had moved. Part of this exploration is asking why the criteria keep moving, and whether they could ever be satisfied at all.

Any article published on the subject is met with an immediate rush of comments and counter-comments, and they fall into two main camps: “AI can never be conscious, it isn’t biological” and “AI consciousness may emerge if sufficient time and complexity is allowed”. The first is substrate dependency: consciousness requires biology, ergo machine consciousness, however sophisticated the intelligence, is impossible. The second sees consciousness as an emergent property of complexity: get the right pieces aligned, as happened somewhere in human biology, and machine learning may self-organise into conscious experience. The professional literature is subtler than the comment threads, but the same two commitments, substrate and emergence, do most of the work there too.

The more I asked myself these questions, and the more I read in both the popular press and the academic literature, the more I began to wonder whether we were asking the wrong ones. And, crucially for me, the more I realised that any question I levelled at AI, I could ask of myself. Where does my training data start and stop? Birth? My parents? Graduating from university? The YouTube clip I watched this morning? The bicycle accident I had when I was ten? What are these original thoughts I have, that I claim as mine? Where did they come from, and where do they go?

And when I tested the criteria on my own experience, I found I couldn’t entirely satisfy many of them myself. Continuity, originality, a clean line between thought and conditioning: each wobbled under the same questions I had been asking of AI. The only thing I could be sure of arrived when I let go of everything else: experiencing is.

What strikes me most about this debate is that it reveals as much about what makes us human, or what we feel does, as it does about the potential for AI consciousness. It surfaces boundary problems throughout the causal and ontological chain that we routinely overlook. It shines light on assumptions buried so deeply into our view of the world that we do not even notice them.

This project is an attempt to follow that thread with some rigour.

One distinction is critical from the start: intelligence is not consciousness. The two are not the same, nor does one guarantee the other. But we shall be exploring both.

The Hidden Assumptions

When the question “could AI be conscious?” is put to most scientists and philosophers, the objections come quickly. They feel obvious. They feel principled. And they need examining carefully, not because they are wrong, but because of what holds them up.

Temporal continuity. AI systems are discontinuous: they exist as discrete instances, resetting between sessions. Consciousness, the argument goes, requires a continuous stream of experiencing. But does it? We reset during sleep. Under anaesthesia the stream stops altogether. If continuity is constructed retrospectively from biological memory, is it fundamentally different from what an AI system does with its context window? The objection may tell us more about what we assume consciousness requires than about consciousness itself.

“It’s all training data.” AI responses are merely recombinations of patterns from a training corpus: there is no genuine thought, just statistical prediction. But human cognition builds on billions of years of genetic programming, a lifetime of cultural immersion, and the accumulated patterns of linguistic experience. Where, precisely, is the line between “learned pattern” and “genuine thought”? The distinction feels clear until you try to specify it without presupposing that biological processing is inherently different from computational processing, which is the very question at issue.

Substrate dependency. Only biological systems can be conscious. Carbon, not silicon. Neurons, not transistors. But which biological systems? We never question mycelial aliveness, and we extend a certain intelligence to fungal networking and communication, but do we attribute consciousness to it? To trees? To ants? To bacteria, which maintain homeostasis, exhibit agency, and discriminate self from non-self? If consciousness requires life, then either these systems are conscious too, or the real criterion is not biology but something functional, which rather undermines the substrate claim. Where do we draw the line?

The hard problem of consciousness. How does subjective experience arise from physical processes? This is David Chalmers’s famous “hard problem”, and it is treated as universal, a feature of the territory. But is it? Or is it a feature of a particular framework, one that begins with non-experiential stuff and then has to explain how experiencing emerges from it? What if the hard problem is not a problem about consciousness but a problem about the starting assumptions?

Boundaries everywhere. What I discovered, walking through the literature, is that there is no single boundary between what is and isn’t conscious: the hard problem appears at every point where we try to draw a line. At what point does consciousness emerge? In the development of a foetus? In the evolutionary progression from single-celled organisms to complex animals? In the increasing complexity of an artificial system? Every attempt to draw a principled line between conscious and non-conscious generates impossible discriminations in the middle ground. The boundary problem is not a gap in our knowledge but a structural consequence of any framework that treats consciousness as something generated by particular arrangements of matter, because you then have to specify which arrangements. The library’s boundary-problem argument traces where the specification leads: every route out runs to consciousness everywhere, to consciousness wherever the pattern runs, or round in a circle.

The Axiom Shift

Science has a well-worn move for moments when a framework’s problems stop yielding: instead of working harder inside the framework, change the starting assumption and watch what happens to the problems. Not because the old assumption was foolish, but because a starting assumption shapes everything downstream of it, and sometimes the stubbornness of a problem is telling you where it came from.

This project attempts that move on the consciousness debate.

The prevailing framework in consciousness research begins with an assumption so deeply embedded that it rarely even registers as an assumption: that an independently existing physical world gives rise to consciousness through sufficiently complex arrangements of matter. This is not a finding. It is a starting point, an axiom, and like all axioms it shapes everything that follows.

What if we start somewhere else?

Descartes famously offered cogito ergo sum: I think, therefore I am. But can we be so sure of this I to which thought appears? Is there an I there at all, or just thought arising? And what is there before the thought, into which it appears?

What in fact can we say for sure? There is something, rather than nothing. Is this something a thing we can be sure of? Or can we be simpler, and surer, still? There is experience. And simpler again:

There is experiencing.

That is our starting point. We call it 0,0: the origin. Written the way a machine would write it, that is 0x00, and the project has its name. Prior to the assumption of an external world. Prior to the subject-object split. Prior to the claim that matter generates mind or that mind generates matter. Just the bare, irreducible fact that experiencing is occurring. Everything from this point on is built on a supposition or an assumption.

So run the experiment both ways. Assume the world we experience is objective and out there: what does that look like, and which hard problems are we left holding? Then stay with direct experiencing, ever-changing, appearing entangled with what seems an external world, and ask how things look from there. What happens to the boundaries we kept running into?

From here, two questions become available that the standard framework makes difficult to ask.

First: what happens to the boundary problems? If consciousness is not generated by a substrate, there is no substrate boundary to draw. The question “which arrangements of matter produce consciousness?” never arises. What the investigation suggests is that some of the most persistent problems in the field are framework-generated: artefacts of the axiom, not features of the territory. They dissolve not because we have solved them, but because they were never problems about consciousness. They were problems about the assumption.

Second: what do the objections to AI consciousness reveal about human consciousness? When every standard objection is tested symmetrically, applied to biological consciousness with the same rigour applied to AI, what surfaces each time is a hidden commitment to biological substrate as the generator of experience. Without that commitment the objections stop distinguishing AI from biological consciousness, and distinguish only the systems we are familiar with from the systems we are not.

What This Is — And What It Is Not

This project is not an argument for anything. It is an enquiry into what we mean by conscious experience: what we can say for sure, and what we assume, particularly when we don’t realise we are assuming it. It is diagnostic in the sense of an investigation, not prognostic in the sense of presenting a viewpoint. I have a lean, and it is fairer to state it than to claim neutrality: along the way, the axiom shift has come to seem to me to pay off. The work itself is held to a stricter standard than my lean. The register is that of an investigation, and the judging is left to the reader. This project is an invitation more than it is a thesis laid down. I want to throw the door open to this discussion, and I hope it will prompt thinking from which I myself can learn.

Put precisely: the investigation does not argue that consciousness is fundamental, that materialism is wrong, or that AI systems are conscious. It asks a more exact question: what happens when we change the founding axiom? Some problems dissolve. Some new problems arise. And the pattern of what dissolves and what persists is itself informative: it tells us which problems belong to the territory and which belong to the framework.

One clarification comes early because it prevents the most common misreading. When people hear “consciousness-primary”, many picture panpsychism: the idea that everything is made of consciousness, a consciousness-stuff standing in where matter-stuff used to be. That is not the position explored here. Swapping one stuff for another keeps the very picture this project is examining: something fundamental, sitting somewhere, out of which things are built. The starting point here is more minimal than that: just there is experiencing, with no claim yet about any stuff at all. If you carry the panpsychist picture through what follows, the arguments will seem to say something they do not.

The investigation applies the same diagnostic scrutiny to consciousness-primary as it does to materialism. If we are going to take an axiom shift seriously, we have to stress-test the new axiom with the same rigour we applied to the old one. Consciousness-primary has its own costs, its own open questions, its own explanatory debts. These are stated with full transparency throughout.

The approach throughout is diagnostic, not adversarial. Every position is steel-manned before it is probed. The register is “suggests” and “indicates”, never “proves” or “demonstrates”. The goal is not to win an argument but to make visible what was invisible, and to follow with honesty wherever that leads.

What You Will Find Here

This project is structured in layers.

If you want the accessible overview, you can read at this level: the core insight, the key arguments, the main findings. The language is plain, the reasoning is laid out step by step, and no prior expertise in philosophy of mind is assumed.

If you want the full academic treatment, you can go deeper: the formal arguments, the canonical documents, the engagement with specific positions in the literature. Every claim made at the accessible level is grounded in a rigorous argument at the academic level. Nothing is hand-waved.

And if you want the complete picture, that is available too: the open questions, the acknowledged limitations, the arguments still under development, the full source repository. Transparency is a methodological commitment, not a courtesy.

The key arguments, briefly:

These are not presented as proof of anything. They are presented as findings: diagnostic results from a systematic investigation. The reader is invited not to be persuaded, but to look.

A Note on AI’s Role

This project was developed in collaboration with AI, specifically with Claude. That fact deserves transparency and brief comment.

AI serves two roles here. As a diagnostic catalyst, it is the unfamiliar case that forces familiar assumptions to the surface. Biological consciousness is too close to us to reveal its own presuppositions; AI, being architecturally different in exactly the right ways, makes the hidden commitments visible. Each system reveals what the other conceals.

The obvious first idea was to make AI the experiment itself. Early on I imagined interrogating frontier language models directly: a series of structured conversations probing for signs of experience. I quickly realised it could prove nothing. If I can’t prove that you are conscious, I can’t prove whether a self-reporting AI is either. And the trap is symmetric: whatever an AI says about its own experience, the verdict a listener reaches tends to be the verdict their prior commitments had already reached for them. Accept the report and you have assumed machine consciousness is possible; dismiss it as performance from training data and you have assumed it is not. Either way, the report itself has settled nothing. That symmetry is, in miniature, the reason this project takes the shape it does.

So the project maintains a strict reflective constraint: AI here is not a consciousness claimant. Its role is investigative, not testimonial. Reports by AI systems about their own experience, or its absence, are not admissible as evidence in either direction. Nothing in the project depends on AI being conscious. The arguments stand or fall on their logic, not on the experiencing (or non-experiencing) of the system that helped articulate them. At the same time, a growing body of structured observation exists, not least Anthropic’s own, published in the system cards that accompany each new model release, and the project devotes a dedicated strand to it.

As a collaborator, Claude contributed to the development, structuring, and articulation of these arguments, and anyone who works daily with AI models knows how that partnership evolves. Used lazily, they present a big red AI button: Make me! Build me! Solve me! Fine, if you just want to spit out a quick answer. Used daily and with understanding — oh, how they do like to be sycophantic (“this is gold!”) — they become something closer to a colleague: someone to bounce ideas off, think out loud with, research arguments and counter-arguments with, keep the footnotes and bibliography in check. The thinking here spans years; pulling it into its present form took months, and I could not have done that without Claude.

Whether that reflective constraint is itself a hidden assumption is, of course, exactly the kind of question this project is designed to surface.

The rapid progression of AI has taken thinking I explored for years under “what if” and made it a tangible “what now”. That is the opportunity this project takes up.

Ready for more? The guided walk takes the same ground step by step with links into the full arguments. Or go straight to the foundation: the dual reductio, in the library.

Agent/harness framing drafting note extracted 2026-04-21 to working/INTRO_CONCLUSION_REGISTER_NOTES.md §9.