The record
The Record
Preserve the failures as carefully as the successes.
The website is not intended to become a collection of favorable demonstrations. It is the public memory surrounding a bounded artifact: what was tested, under which version, what happened, what failed, what remained unresolved, what reconciliation later established, and what changed as a result. If the project develops, its history should remain capable of showing how it developed.
ATB / RECORD / 001 Artifact fixed Frozen protocol · owner access Preregistration pending Results empty
Scientific state must remain reconstructable. The organizing principle
The manuscript is the experimental artifact.The website is the experimental record.
Developmental history. Before the test edition was locked, the author and the developing manuscript formed a recursive process: reasoning produced the artifact, the artifact made that reasoning durable and inspectable, and reflection on the artifact helped reorganize later reasoning and revision. The manuscript participated as an external environment; it did not independently create the architecture or contain a complete model of the author's mind.
Experimental condition. Once an edition is designated for testing, that developmental loop is stopped. The printed manuscript becomes the bounded, version-locked artifact supplied without continued intervention from its author.
The website maintains the evolving scientific infrastructure around that artifact: preregistered protocols, controls, scoring methods, experiment versions, datasets, results, replications, and methodological revisions. The website may evolve. The artifact tested in a particular experiment must remain identifiable and fixed. A later website revision cannot retroactively alter the manuscript or the conditions of an earlier test.
Experimental exposure
Once I have encountered it, it is no longer genuinely novel to me.
Because I am presently the human reference against which reconstruction of the artifact can be evaluated, experimental exposure matters.
If I am shown a novel question, scenario, piece of evidence, interpretation, model output, or other information capable of affecting subsequent testing, that exposure becomes part of the experimental history. That exposure is documented. This requirement is not optional.
The scientifically relevant substance of qualifying exposure is preserved through the project's public research record, including its Open Science Framework record or designated successor. Later researchers must be able to distinguish a genuinely novel test from something I have already encountered.
01What the public record preserves
The record preserves enough information to maintain the interpretability and continuity of the research, including where applicable:
- artifact and architecture version;
- experimental period;
- relevant questions, scenarios, evidence, or exposures;
- the author's original response and reconstruction outputs;
- conditions required to interpret the comparison;
- identified discrepancies and their status;
- reconciliation findings; and
- resulting effects on the architecture or version history.
The purpose is contamination control, methodological memory, and an auditable history of how the architecture changed.
02Confidentiality
Confidentiality can protect the participant. It cannot erase the experiment.
Where appropriate and legally permissible, confidentiality may protect names, organizations, affiliations, contact information, internal project names, unrelated proprietary information, trade secrets, and other identifying or commercially sensitive information unnecessary to preserve the scientific meaning of the exposure.
My identity is known. Yours does not have to be. Your organization does not have to be identified. But my scientifically relevant experimental exposure must remain part of the record.
An engagement requiring the experimental exposure itself to disappear is incompatible with direct participation in this research program.
03Reconciliation
An observed discrepancy between the author and the artifact is not automatically treated as failure by either one. Experimental exposures accumulate before being considered through a separate reconciliation process designed to determine, as far as the evidence permits, what produced the discrepancy.
Reconciliation preserves the distinction between the original experimental observation and its later interpretation. A discrepancy may reflect corrected information, legitimate revision, an incomplete reconstruction, incorrect weighting, a procedural confound, architectural change, or a question that remains unresolved.
04Versioning and ESPS
The version under active testing remains bounded during its testing period. Reconciliation findings are evaluated collectively rather than used to rewrite the experimental object after individual trials.
ESPS serves as the formal architectural record of that reconciliation. Where findings establish something new about the distinctions, weighting structures, revision mechanisms, constraints, or other operations underlying the author's reasoning, the appropriate subsequent version records that architectural development.
The version history preserves what the earlier version reconstructed, what the author actually did, what later reconciliation determined, and what changed next. Earlier results are not rewritten to make a later version appear more successful.
Current state
Experiment Status
Accurate as of the Version 1.0 package release on September 4, 2026.
Artifact
The Acts of Erudition — fixed manuscript edition used for testing
Research package
Version 1.0 · owner-protected
Phase
Frozen v1.0 / Pre-Confirmatory Testing
Confirmatory question bank
Not publicly released
Formal results
Not yet published
Preregistration
Pending
SEALED
Deliberately absent
Why isn't the confirmatory question bank here?
A valid confirmatory test requires questions the author has not seen in advance. Publishing those questions before testing would contaminate the experiment.
The website explains the rules by which the question bank is created, screened, randomized, and preserved—but not the active confirmatory questions themselves before testing. After a completed experimental run is locked, the used questions may be released as part of that experiment's permanent record.
Artifact identity
What Exactly Is Being Tested?
Ask the Book does not test an abstract idea called “the author's philosophy.” It tests the identifiable surviving product of a recursive developmental process. For a formal experiment, that exact product becomes a bounded artifact.
Every formal experiment should preserve- Manuscript edition and publication date
- Cryptographic file hash
- Exact Ask the Book prompt version
- Model and model version
- Experimental protocol version
Developmental history and experimental condition are deliberately separated. The author and manuscript could influence one another while the work was being made; after version lock, the author no longer modifies the artifact or participates in the model's answer.
Later revisions to the book, ESPS, prompt, or experimental protocol are new experimental conditions. They must not be silently substituted for earlier versions.
This version ledger allows an experiment conducted years later to identify precisely what was tested—and lets a critic reconstruct the methodological state in which a result was produced.
Preserved revisions
Version History
Research Package v1.0
Initial formal confirmatory architecture.
Future versions will not overwrite earlier packages. Each revision must state what changed, why, whether it occurred before or after a run, and which experiments used it.
Results & Replications
No confirmatory result has been published yet.
Exploratory pilot testing informed Version 1.0 but is not being presented as confirmatory evidence.
What every future result record must preserve
- Experiment number and date
- Protocol and artifact versions
- Models tested and question count
- Structural Continuity Index results
- Austin-vs-Austin baseline
- Control-condition comparisons
- Inter-rater reliability
- Failed decision gates
- Full dataset and raw model outputs
- Analysis files and protocol deviations
- Replication attempts
- Limitations
Failures must be displayed as visibly as successes. This archive is an experimental record, not a collection of testimonials.
Chronological memory
The website preserves what happens around the artifact.
- 01Artifact
- 02Protocol
- 03Preregistration
- 04Test
- 05Data
- 06Result
- 07Criticism
- 08Replication
- 09Revision
- 10Next test
The current experiment asks whether a fixed artifact produced through recursive author–manuscript development retains recoverable, measurable aspects of the reasoning architecture that helped produce it. Possible future technologies remain speculation unless controlled evidence makes them more than that.
Researchers and investigators
The most valuable contribution is not endorsement. It is a better test.
If you see a weakness in the hypothesis, architecture, controls, scoring system, interpretation, or experimental design, test it. Propose adversarial questions, stronger controls, independent scoring, replication attempts, alternative explanations, or methods capable of showing that an apparent effect is generic, unstable, contaminated, or absent. A criticism capable of becoming a test is more valuable to this project than agreement that cannot.
Request Austin's participation