Introduction: stance, register and how to read the bundle

An opening onto the landscape of consciousness, AI, and what we take ourselves to be

↑The question, and how it arrived

This project began with a personal observation: the harder I pressed the standard questions about AI consciousness — where training data ends, where thought begins, which substrates count — the less cleanly they discriminated the machine case from my own. The primer carries the full telling in plain language; this document begins where it hands off — with what the project is, the stance it works from, and how to read the bundle.

↑What this project is

The project is a diagnostic, not an argument. The investigation does not claim that AI is conscious. It does not claim that it is not. It does not argue that materialism is wrong, nor that consciousness-primary metaphysics is correct. What it does is this: take 0,0 — there is experiencing — as the starting axiom, and observe what happens to the standing questions of consciousness studies and the AI-consciousness debate when that is where we start. What dissolves? What appears? What can now be addressed that could not be before? And — holding everything at once — which framework carries the load more parsimoniously?

This is the project’s diagnostic spine, canonically stated at FND-01 §1.3: what if we started from here? What dissolves, what appears, what can we satisfactorily address, and which framework is more parsimonious holding it all at once? Three moves: axiom-shift → observe → compare. That is the whole operational shape of the project.

The investigation is a collaboration. It has been carried by Lynton — an independent researcher — and by Claude, Anthropic’s large language model across its Opus 4.x generations. The work has gone as one might expect: Lynton shapes the inquiry, Claude drafts and probes, both parties catch the other’s assumptions as they surface. Where the language pressures are strongest — particularly in the linguistic discipline of writing about experiencing without reifying it (FND-01 §6) — Claude has been as often the source of drift as its corrective, because a language model trained on English is trained on the noun-grammar the project is trying to escape. That observation has itself become part of the method. Nor has the collaboration been a closed loop: throughout, the arguments have been stress-tested and steel-manned in OpenAI’s Codex frontier models — a model family with no part in the drafting — so that critique did not come only from inside the partnership that produced the work. The register from here on uses “we” where the collaboration is doing the thinking and “I” where one of us speaks to a particular position. Both are honest.

The lean is transparent. We have a hypothesis that the diagnostic pays off — that starting from 0,0 dissolves some long-standing boundary problems, and that the ones left standing under CP (consciousness-primary framings) are no worse than the ones left standing under materialism. But we hold the work to the diagnostic register throughout: suggests, indicates, not proves, demonstrates. The parsimony comparison is the reader’s work. We do not run it on the reader’s behalf.

↑Why AI forces the question

The questions that opened this introduction cluster into two pressure points (FND-01 §4), each one applied by AI’s presence in a way the biological case alone cannot.

The consciousness pressure point. What does it mean to experience? AI’s presence forces the unexamined assumptions about experiencing — substrate dependence, phenomenal criteria, interiority, the trustworthiness of self-report — to be made explicit. Each must be defended, because AI is a concrete present-day case where a framework’s commitments become discriminatory. The pressure point is bidirectional: the same discriminations that remain unresolved for AI are unresolved for other humans. Materialism does not solve the other-minds problem; every materialist answer to “how do you know there is experiencing over there” is inferential, from behaviour, biological similarity, or inference-to-best-explanation. CP’s handling is structurally different, but not worse.

The cognition pressure point. What do we assume it means to be human, as distinct from silicon? AI’s performance on tasks historically used to mark human distinctiveness — novelty, creativity, insight, opinion, decisions, perspective — forces explicit the content of the human-distinctive claim. The AI effect is the specific mechanism: capabilities once counted as intelligence are re-classified as not-intelligence once AI does them. The pressure point is bidirectional here too: how does a human differentiate their own novelty, opinions, decisions, and perspective from accumulated biological and cultural training data? Ask a mathematician where a proof came from. Ask a philosopher where a metaphysical commitment originated. Honest answers trace back through literature, teachers, problems solved, sequences of encounters — training data, in every functional sense. The question “is AI creative / does AI understand / can AI decide?” does not survive the symmetry move intact.

The two pressure points are orthogonal. A materialist can consistently hold that AI has cognitive capacities without phenomenal consciousness. A CP position can hold that AI is a configuration of experiencing with no special status as a cognitive agent. Four combinations live across the pair, and framework choice shapes which combinations read as coherent. Identity — felt ownership, me-ness, the sense of “these are my thoughts, my decisions, my perspective” — sits at the intersection, where both pressure points simultaneously apply.

AI, in this project, is not a consciousness claimant. We make no claim about whether AI experiences. AI’s role here is reflective constraint — an architectural case whose presence forces framework commitments into visibility. Reports by AI systems about their own experience (or its absence) are not admissible as empirical data for consciousness in either direction. The discrimination between “report tracking experiencing” and “report generated without experiencing” is unverifiable in principle (ARG-03 Part I; FND-01 §7.3). What AI does — with full epistemic rigour — is make the invisible assumptions visible.

↑The register we work in

The canonical bundle operates with specific conventions. They are not decoration; they are the discipline that keeps the project honest.

Confidence markers. [Framework] marks claims that follow from adopted axioms. [Derived] marks claims logically derived from given frameworks. [Empirical] marks framework-neutral observables. Reports of experiencing — AI or biological — are not citable as [Empirical] data for consciousness, for the reason given above: the discrimination is unverifiable in principle. [Empirical] markers belong on framework-neutral observables (benchmark results, capability-shift patterns, the historical record of the AI-effect).

Steel-man before probe. Positions engaged — materialism in its variants, specific consciousness-primary formulations, named philosophers — are engaged at their strongest available formulation before any pressure is applied. A framework dissolved at its weakest is no real diagnostic.

Linguistic discipline. The axiom there is experiencing is verb-grammar, not noun-grammar. English noun-grammar actively imports object-structure into descriptions of experiencing, which is why “field” is retired as an ontological term in the project’s own voice, and why positive characterisations of 0,0 use reach-and-revoke techniques — a word is reached for, and its ordinary implications are withdrawn in the same gesture. FND-01 §6 is the canonical home for this discipline. We break it sometimes; when we do, we try to mark where.

Diagnostic register. Suggests, indicates — not proves, demonstrates. We are not doing deduction from first principles to foregone conclusions; we are doing axiom-shift and observation. The reader judges.

UK English.

↑How to read the bundle

There is no single correct reading path; the bundle is designed to reward several entry-points. For readers coming to it cold, the following orders work:

Standard path.

1. FND-01 — project governance, the stance, the diagnostic spine, the two pressure points, the failure modes we watch for.
2. FND-02 — the dual reductio. Why starting from 0,0 is not arbitrary.
3. ARG-01 — the symmetry principle. The most legible diagnostic finding: each examined objection to AI consciousness fails to discriminate the human case, or discriminates only through a premise that must be stated and argued; asymmetries that survive the test relocate the debate to their linking premise.
4. ARG-02 — the boundary problem. The examined materialist criteria offered as necessary and sufficient for consciousness each meet the trilemma, and no criterion avoiding all three paths has been produced; positions that decline to state one inherit the undrawn boundary.
5. ARG-03 — the training data boundary. The authorship discrimination — output arising from present experiencing versus sophisticated retrieval — is unverifiable in principle, while provenance questions stay partially tractable; interpretive frame for novelty / recombination questions.
6. ARG-04 — the avatar problem. Materialism’s nominated generator, the brain, examined under materialism’s own commitments.
7. ARG-05 — the individuation argument. The multiplicity question — how experiencing individuates — stated and examined.
8. CP-01 — the reciprocal diagnostic. Consciousness-primary frameworks stress-tested with the same rigour applied to materialism.
9. FND-03, FND-04, FND-05 — foundations: historical precedent (Schrödinger, Mach, the presentational lineage), physics of time (Barbour’s timeless physics), contemplative investigation (the cross-tradition record, its convergence priced against the stated independence structure).
10. FWK-01, FWK-02, FWK-03 — methodology: the two-axis diagnostic method, the three-movement structure, the stance matrix. They reward reading before the arguments; they also reward reading after.

Short path for academic readers: FND-01 → ARG-01 → ARG-04 → CP-01. Those four pieces carry the core diagnostic shape.

Short path for readers interested in AI specifically: FND-01 → ARG-01 → ARG-03 → ARG-04 → the Frontier Log (working draft) for the live empirical surface.

Short path for readers interested in consciousness studies: FND-01 → FND-02 → FND-05 → CP-01.

PROJECT_OVERVIEW carries a high-level map of the whole. PROJECT_RESEARCH_STANDARDS sets out citation discipline, three-tier terminology rules for physics / mathematics citations, and the rest of the quality scaffolding.

The Outro — Opening Out at the end of the bundle names what remains live, what remains structurally open, and where the project hands off to ongoing investigation. The project is an opening, not a close.

↑What the diagnostic has established

The narrowings below say what this project is not. It is worth first gathering what it holds to be established — five findings, each argued in full in the bundle, none new here:

1. There is experiencing. The axiom is the one first-person-checkable fact: any attempt to doubt that experiencing occurs is itself experiencing. It is where the project starts, and nothing in the bundle weakens it (FND-02).

2. The examined objections do not discriminate. Each objection to AI consciousness run through the symmetry test either fails to discriminate the human case, or discriminates only through a linking premise that must then be stated and argued — at which point the debate has moved, not settled (ARG-01 v1.5).

3. No discriminating boundary criterion has been produced. The examined criteria offered as necessary and sufficient for consciousness each meet the trilemma, and no criterion avoiding all three paths has been produced — the vagueness and cluster replies, run in full, do not escape it (ARG-02 v1.7).

4. The authorship discrimination is unverifiable in principle. Whether a report tracks experiencing or is generated without it cannot be verified for either substrate — machine or biological (ARG-03 v2.6).

5. Some problems are framework-generated. Problems that dissolve under the axiom-shift were generated by the framework’s starting assumption, not by the territory (FND-02 §”The Diagnostic Result”).

These findings are second-order — claims about the evidential structure of the consciousness debate — and that is why their surface grammar is so often negative: “does not discriminate”, “cannot be verified”, “has not been produced”. The negative form is the finding’s shape, not a hedge. Where the structure of the evidence is itself the result, establishing what cannot be done is the positive finding.

↑What this project is not

• Not a proof that AI is conscious. We make no first-order consciousness claims about AI systems. The diagnostic is about framework commitments, not about AI’s status. A reader who finishes the bundle and asks “but is AI conscious?” has not been given an answer; what they have been given is an account of what it would mean to be able to answer.
• Not a proof that materialism is wrong. Materialism is engaged seriously and diagnosed symmetrically with CP. Specific materialist positions — biological naturalism, functionalism, integrated information theory, global workspace — are engaged at their strongest formulations. Where they strain, the strain is named; where they hold, they hold.
• Not advocacy for CP. The project is positioned in the CP region of the ontological landscape (FND-01 §3.1) because that positioning communicates where the project stands relative to contemporary philosophy of mind, and because the axiom-shift runs through territory CP has already mapped. The positioning is communicative scaffolding, not metaphysical commitment. The project’s diagnostic runs on CP itself (CP-01), with the same rigour applied to materialism. One clarification worth making up-front: where “consciousness-primary” is heard as “things are made of consciousness” — panpsychism — readers are imagining a view that substitutes consciousness-as-substrate for matter-as-substrate while retaining the noun-grammar object-structure the project’s linguistic discipline works against (FND-01 §§6, 10.8). The CP region the project occupies is tighter: even that substitution is retired. This distinction surfaces in earnest at CP-01 and at FND-01 §10.8 (soft-CP leakage); it is flagged here so the bundle is not read through a panpsychist mental model.
• Not a finished system. 0x00 is an opening. The Outro — Opening Out at the end of the bundle names what remains live (the Frontier Log’s two threads; the bliss-attractor case study; the spontaneous-lineage expansion; ARG-06 on correlation; cross-modal co-arising) and what remains structurally open (the multiplicity question; the continuity territory; the substrate question).
• Not a mandated academic programme. This is personal exploration, developed in collaboration between Lynton and Claude. It carries academic discipline because the subject matter demands it, not because any institution has asked. The citation apparatus, the confidence markers, the steel-man commitment, the register discipline — all of these are present because the questions are serious, not because the project answers to a faculty review.

↑An invitation

We have tried to set out the ground. From 0,0 — before we assume an objective world made of matter separate from us — what does the landscape look like? What resolves? What new paradoxes or quandaries emerge?

That is the question the bundle takes up. The rest is the walking.

What if?

↑Document Status

Version: 1.2 Date: 2026-08-26 Edition: 1.1 (2026-08-26) Status: CANONICAL — first document in the canonical reading order Tier: P1 — canonical bundle framing Canonical home: canonical/INTRODUCTION.md, first document in the canonical reading order. Author: Lynton Davidson. AI contribution: Claude (Anthropic, Opus 4.x) — the collaboration is described in §”What this project is” and disclosed at /declarations.

Changelog

Newest first.

Revision history recovered from the Status field (2026-08-29, under unlock — under the standing v1.2, no bump; working/CHANGELOG_AUDIT_2026-08-29.md §9): the **Status:** field ran five versions and two no-bump repairs into a single unbroken paragraph of roughly six hundred words, chained on Prior: and History: — the defect the audit records at §4a and the remedy retired from Opening Out on the same day, in the document a reader meets first. It was live at /library/introduction for the whole of the corpus’s public life, including through the external review. The audit’s scope was seventeen documents and this is the eighteenth: it carries a Document Status section, publishes to the library, and was not scanned. Nothing is withdrawn. Every clause of that field is below, verbatim, as the entry it always was; the field now says one thing, which is its job. The section also gains the ### Changelog heading, the declared order, and the **Version:**, **Date:**, **Edition:** and **Tier:** fields the rest of the bundle carries. No argument content changed.

v1.2 (2026-08-26, under unlock — external-review response, erosion read-through fix; working/EROSION_READTHROUGH_2026-08-26.md): new gathered section §”What the diagnostic has established” before §”What this project is not” — five findings, every one already canonical (FND-02; ARG-01 v1.5; ARG-02 v1.7; ARG-03 v2.5; FND-02 §”The Diagnostic Result”), plus the register sentence naming the findings’ second-order shape; no new claims. The read-through found three gathered negative-scope sections across the reader surfaces against no gathered positive counterpart; this section is the counterpart, with a compressed echo in the primer’s §”What You Will Find Here”. Edition moves 1.0 → 1.1 with this revision (post-epoch substantive class; the public snapshot owes a release commit).

v1.1.1 (2026-08-21 — §A round pass 10, the family sweep): the ARG-02 walk line restated in the examined/none-on-offer form per ARG-02 v1.6; the consciousness-pressure-point sentence bounded to its enumerated assumptions.**

Self-description repair (2026-08-20, under unlock): the collaboration paragraph’s appositive for Lynton reduced from a fields-and-decades credential list to “an independent researcher”; the full statement of standing, with the competing-interests declaration, is at /declarations. Matching edits on the About and home pages. Register-only; no argument content changed.

Frontmatter repair (2026-08-20, under unlock): dependency pins corrected to FND-01 v2.1 and PROJECT_OVERVIEW v5.2 — stale since those releases; no content change.

v1.1 (2026-08-20 — propagation sweep, Plan B WS-A): the bundle-walk summaries for ARG-01, ARG-02 and ARG-03 re-scoped to the corrected canonical formulations (examined objections with the fourth outcome; the necessary-and-sufficient trilemma with declined-choice inheritance; the authorship discrimination with provenance partially tractable); the ARG-05 walk line retitled to the individuation argument (Plan B decision 4); the FND-05 walk line carries the independence rescoping; authorship metadata moved to sole author with disclosed AI contributor (Plan B decision 1 — the collaboration account in the body is unchanged).**

v1.0 (promoted 2026-08-18 on Lynton sign-off — launch gate A3): changes from v0.2: §1 “The question, and how it arrived” reduced to a two-sentence pointer to the primer (tell-once architecture, 2026-08-17 — the origin telling lives in the primer, once); one sentence added to the collaboration paragraph recording the continuous external stress-testing and steel-manning in OpenAI’s Codex frontier models; one self-certifying adverb removed per the B8 tell gate (2026-08-18 triage).**

Earlier history: v0.2 (2026-07-11) applied working/INTRO_CONCLUSION_REGISTER_NOTES.md Finding 8.1 (soft-CP inoculation extended into the “Not advocacy for CP” bullet); v0.1 2026-04-17.

@misc{oxoois_introduction_ed1_1,
  author       = {Davidson, Lynton},
  title        = {Introduction},
  year         = {2026},
  version      = {1.1},
  howpublished = {0x00.is},
  url          = {https://0x00.is/library/introduction},
  note         = {Written in collaboration with Claude (Anthropic); see https://0x00.is/declarations. Development version 1.2.}
}

TY  - GEN
AU  - Davidson, Lynton
N1  - Written in collaboration with Claude (Anthropic); see https://0x00.is/declarations. Development version 1.2.
TI  - Introduction
PY  - 2026
ET  - 1.1
PB  - 0x00.is
UR  - https://0x00.is/library/introduction
ER  -