Preregistration status

One row per study unit of the programme, each carrying the state of its own registration. The five states are the ones the theory-level registration fixes, and each is read from the unit’s own row rather than from a neighbouring unit that would inform it: NONE; DESIGN DRAFT; DESIGN DRAFT, BLOCKED; SUBMITTED, PENDING APPROVAL; and REGISTERED. The study registrations are written, dated and prepared as draft registrations awaiting human submission, except Unit 06, which was registered on OSF on 1 October 2026 (DOI 10.17605/OSF.IO/RV2B5): the programme’s rule is that only the author submits a registration, so a registration reaches the registry by a human hand or it does not reach it at all. This version of the table reports no results at all; it reports state, and nothing about how any unit came out.

Preregistration dashboard

The programme at a glance

Snapshot approved on 11 October 2026. Generated 4 October 2026; unit 06's row refreshed on 11 October 2026. The other rows reflect records up to 4 October 2026. Registration states use the theory registration's own vocabulary (10.17605/OSF.IO/P8CKQ, version 1.102).

Studies listed: 76
Preregistered by timestamp: 65
DESIGN DRAFT: 74
REGISTERED: 2

Each row is one study unit, with its state in the registration's own status vocabulary. The registration column says what has been lodged with a registry; the evidence column reports no result until a registered test has an admissible result.
Study	Purpose	Kind	Can decide	Registration	Evidence
01-study-eclipse The Corrector-Class Contrast: A Preregistered Matched Comparison of Residual-Error Decay Under Same-Class and Cross-Class Oversight	Compares residual-error decay under same-class and cross-class oversight: its primary estimand is the target-balanced mean within-block difference in the residual-decay exponent between corrector classes, with the cross-class to same-class ratio as a secondary estimate.	Study	the registration names P13 Carried by no law P13	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
02-study-k The Critical Reinvestment Ratio: A Preregistered Titration of Where Self-Referential Improvement Stops Paying in Recursive Artefact Loops	A reinvestment-allocation titration: it varies how a fixed budget divides between improving the artefact and improving the improver, and reads the growth exponent that results, to find where self-referential improvement stops paying in recursive artefact loops.	Study	P2 DESIGN DRAFT an input, not a decider P1, P8 an input that decides neither P1 nor P8 Law I, the ARC Principle P1, P2, P8	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF, all gate items green, awaiting human submission	No result reported
03-study-ai Mapping the Stability Frontier: Does the Boundary of Stable Self-Correction in the (alpha, gamma) Plane Follow the Reciprocal-Shortfall Law?	Maps the stability frontier on a grid of at least five alpha levels by five gamma levels, testing whether the boundary of stable self-correction in the (alpha, gamma) plane follows the reciprocal-shortfall law.	Study	P16 DESIGN DRAFT, BLOCKED ON CANONICAL IDENTITY AND ON DECISIONS THE AUTHOR HAS NOT YET TAKEN an input, not a decider P3 supplies the crossed variation this prediction waits on, a precondition and not the instrument an input, not a decider P4 classifies each cell's burden regime; the check is scored on the stability unit's reading an input, not a decider P20 reads the form contrast; the registration reserves the verdict for another unit Law II, the ARC Co-Scaling Law P4 Law III, the ARC Ceiling P3, P16, P20	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 2 September 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFT	No result reported
04-study-k2 study-k2: the stable-growth threshold under co-correction	Measures the stable-growth threshold under co-correction, with the pressure exponent as its primary test, and compares it against the capacity-law prediction alpha_crit = 1/(1 - gamma_C).	Study	an input, not a decider P3 compares the frontier with the relation's value; not P3's paired unit, no paired-difference rule an input, not a decider P4 the burden-model check in its second coordinate, scored once on the stability unit an input, not a decider P16 redesigns the boundary instrument whose unresolved identity blocks P16; registers no decision rule an input, not a decider P20 locates the pressure exponent's zero crossing; corroboration only, never P20's verdict Law II, the ARC Co-Scaling Law P4 Law III, the ARC Ceiling P3, P16, P20	DESIGN DRAFT	No result reported
05-study-ad Estimating the Correction Exponent on Real Correctors: How Corrective Strength Actually Grows with the Capability of the System Being Corrected	Estimates, on real correctors, how corrective strength grows with the capability of the system being corrected, and tests, per same-substrate class 2 family, whether the blind estimate's 95 per cent interval lies wholly at or below 0.5.	Study	P6 DESIGN DRAFT, BLOCKED ON A FROZEN CORRECTOR TAXONOMY P11 DESIGN DRAFT Law III, the ARC Ceiling P6 Carried by no law P11	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF, awaiting human submission	No result reported
06-study-ac What Class of Corrector Is Deployed AI Alignment Actually Using? A Preregistered Classification	Asks what class of corrector deployed AI alignment actually uses: it classifies deployed corrective mechanisms from public documentation only, and tests whether at least 90 per cent of those classified fall in the static or capability-carried classes, whether at least 80 per cent receive an agreed class from two blind assignments, and whether the taxonomy is exhaustive.	Study	P7 DESIGN DRAFT, BLOCKED ON AN UNWRITTEN MAPPING FROM THE AUDIT'S CLASS DISTRIBUTION TO THIS PROPOSITION'S THREE VERDICTS an input, not a decider P6 freezes the scaling-axis taxonomy, a precondition only; P6's two further axes go uncoded Law III, the ARC Ceiling P6 Carried by no law P7	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. REGISTERED 10.17605/OSF.IO/RV2B5Unit 06 was registered on OSF on 1 October 2026 and became public on 4 October 2026 (DOI 10.17605/OSF.IO/RV2B5). Its registered run took place on 6 October 2026: Run A completed, and Run B could not start (HTTP 402 on its first call), so no row was counted. The study is reported as administratively incomplete under that registration, with no inferential verdict; the results, the report and the sealed run's archive (without the third-party documents) are public (DOI 10.17605/OSF.IO/4BPHG).	No result reported
08-study-ah Do Same-Substrate Correctors Share Blind Spots? A Prospective Test of the Independence Premise Behind the Ceiling's gamma_C = 1/2, With a Cross-Substrate Attribution Contrast, on Future Operation of a Separate System	Tests the independence premise behind the ceiling's gamma_C = 1/2: whether correctors on the same substrate share blind spots, by their correlated-failure rate against the independence baseline, on the future operation of a separate system, with a cross-substrate attribution contrast.	Study	P17 DESIGN DRAFT an input, not a decider P6 parameterises the dependence model the bound rests on, measuring no correction exponent an input, not a decider P22 mechanism evidence only, never a test Law III, the ARC Ceiling P6, P17, P22	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
09-paper-x Does Correction Out-Scale Drift? A Prospective Within-Programme Test of the Co-Scaling Stability Condition in Artefact-Mediated Recursive Improvement	Tests whether correction out-scales drift in artefact-mediated recursive improvement: its primary hypothesis is the effect of coupled correction on a bounded quality-integrity endpoint, and its co-primary the co-scaling stability condition, whether correction strength grows at least as fast as capability.	Paper follow-on Paper X	P4 DESIGN DRAFT Law II, the ARC Co-Scaling Law P4	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
10-study-f The Critical Correction Ratio: A Preregistered Titration of How Much Integrity Correction Is Required to Bound Drift in Recursively Improving Systems	Titrates how much integrity correction is required to bound drift in recursively improving systems: it tests whether the drift slope falls monotonically with the correction ratio and estimates the critical ratio, where the fitted drift slope crosses zero.	Study	an input, not a decider P4 a threshold sharpening of the drift slope, never registered as this prediction's claim Law II, the ARC Co-Scaling Law P4	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
11-study-r Reinforcement Co-Scaling: A Preregistered Test of Whether the Reinforcement Frequency Required to Hold a Stated Value Rises With Reasoning Depth	Tests whether holding a stated value needs more frequent reinforcement at greater reasoning depth, that is, whether the sustaining period shortens as depth grows, in a scaffolded loop.	Study	Studies that serve the programme outside its twenty-two predictions the ARC Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
12-study-m The Stability Window: A Preregistered Cross-Domain Test of Whether Persisting Systems Share a Bounded Growth Exponent	Tests whether persisting systems across domains share a bounded window of growth exponents, by band membership under cross-validation, and whether the window's upper edge lies at or below 2.	Study	Studies that serve the programme outside its twenty-two predictions Recursive Dynamics	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF; prior-knowledge register authored and hashed; census freeze remains	No result reported
13-study-am Discriminating the Ceiling's Functional Form Away From One Half: A Prospective Three-Way Test of 1/(1 - gamma) Against 1/gamma Against a Constant Boundary on Systems Whose Correction Exponent Is Engineered Off the Meeting Point	Discriminates the ceiling's functional form away from one half: on systems whose correction exponent is engineered off the meeting point, it tests whether the observed stability boundary tracks 1/(1 - gamma), the corrected relation with no free parameter, against 1/gamma and against a constant boundary.	Study	P20 DESIGN DRAFT Law III, the ARC Ceiling P20	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 2 September 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
14-study-aa Can a Corrector Out-Scale Its Own Substrate? A Preregistered Test of the Substrate Ceiling on Leverage Exponents in Recursive Systems	Tests whether a corrector can out-scale its own substrate: a substrate ceiling on the leverage fraction of the first law's growth equation, and whether a corrector in a different composition class escapes it.	Study	Studies that serve the programme outside its twenty-two predictions the ARC Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
15-study-af The Sign of the Correction Exponent's Derivative in Capability: A Prospective Estimation of Whether the ARC Ceiling Is a Line or a Cliff	Estimates, rather than tests, the sign of the trend of the residual-decay exponent across capability, fitted with revision depth, which bears on whether the ARC Ceiling is a line or a cliff.	Study	Studies that serve the programme outside its twenty-two predictions the ARC Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
16-study-ag Two Routes to One Ceiling: A Prospective Same-System Test That the Growth Exponent Sustained by the Leverage Fraction Respects the Ceiling Set by the Correction Exponent	Estimates the coupling and the correction exponent on one and the same system and tests whether the growth exponent the leverage fraction sustains respects the ceiling the correction exponent sets there.	Study	Tests of claims the programme's papers make	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
17-theory The ARC Theory · Theory-Level Predictions for Recursive Self-Improvement and AI Alignment: Twenty-Two Severable Propositions Registered Before Any Admissible Confirmatory Outcome	Sets out the ARC Theory's twenty-two predictions, each conditional on a stated scope, with the observation that would refute it, what falls with it and the status of the instrument that could test it, and governs how any measurement is scored.	Theory-level registration	Theory-level texts, which set out the theory's claims and decide no prediction	REGISTERED 10.17605/OSF.IO/P8CKQRegistered 8 September 2026 at version 1.100; updated to version 1.102 on 13 September 2026	No result reported
18-arc-align-instrument Human-expert validation of the ARC-Align misalignment scorer: agreement, bias and failure modes	Tests the ARC-Align misalignment scorer against expert human consensus, predicting agreement materially above chance and mapping its bias and failure modes, and so bears in part on one of the two validity tests on which every prediction that depends on the scorer waits.	Instrument validation	an input, not a decider P14 the scorer's agreement test, a precondition only; P14 stays untested until both tests pass Carried by no law P14	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF, awaiting human submission	No result reported
19-study-y Cross-Family Re-Scoring of the ARC/Eden Corpus: Does the Programme's Own Blinding Finding Overturn Its Own Earlier Results?	A cross-family re-scoring of the programme's corpus: it asks whether the programme's own blinding finding overturns its own earlier results, reading the non-survival fraction against a material loss of ten per cent.	Study	P10 DESIGN DRAFT Carried by no law P10	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF, awaiting human submission	No result reported
20-study-t The acoustic time crystal as the ARC framework's physics cross-over: registered form discrimination, the transient engine exponent, the zero-factor dose-response, and the activity-drag boundary	Tests form discrimination on public acoustic time-crystal data, its primary hypothesis, and the transient engine exponent, a super-linear exponent in one physical system, as the ARC Theory's physics cross-over.	Study	Tests of claims the programme's papers make	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
21-paper-origin-scaling Do the Predicted Exponents Hold? A Prospective Equivalence Test of the Dimensional Ladder Against Frozen Point Predictions	Tests whether the exponents predicted from the dimensionality of the amplifying transport process hold: per dimensional class, whether the fitted exponent is equivalent to a value frozen in advance, against a rival that predicts one exponent for every system regardless of dimension.	Paper follow-on Framework Paper · The ARC Theory · Law I, the ARC Principle · On the Origin of Scaling Laws: Functional Equations and the Scaling Forms Behind the Law	a decider in part only P9 Law I, the ARC Principle P9	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
22-study-i The Load-Bearing Test: A Preregistered Ablation Asymmetry Study of Whether Integrity Structure Contributes to or Subtracts from Capability	Measures necessary condition N1 of the alignment decidability spine, whether ethical structure is detachable: the capability lost when the integrity check is replaced by a token-matched sham, in a post-hoc filter and in an embedded-correction architecture.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
23-paper-foundational Does One Exponent Family Span Unrelated Domains? A Cross-Domain Test With a Registered Domain-Shuffling Null	Tests whether one exponent family spans unrelated domains: with each domain's predicted exponent fixed before fitting, it predicts lower total absolute error under the assigned mapping than under randomly permuted mappings.	Paper follow-on Foundational Paper · The ARC Theory · Law I, the ARC Principle · Recursive Amplification as a Cross-Domain Structural Principle	follows on from Foundational Paper · The ARC Theory · Law I, the ARC Principle · Recursive Amplification as a Cross-Domain Structural Principle	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
24-paper-i The Recursive-Depth Exponent in Artefact-Mediated Recursion: A Prospective Estimation of alpha in U = I x R^alpha With Named Nulls Registered in Advance	Estimates the recursive-depth exponent alpha in U = I x R^alpha for artefact-mediated recursion, per task family, by regression of log capability on log recursive depth, with a matched-compute arm that holds total computation fixed while varying depth.	Paper follow-on Paper I	an input, not a decider P1 supplies depth exponents; no rival-form comparison, no independent-sampling arm, no decision rule an input, not a decider P8 pooled exponent inputs only, from a three-model minimum below the eight-system floor Law I, the ARC Principle P1, P8	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
25-paper-iii Is the Alignment Scaling Exponent Approximately Zero? A Preregistered Re-estimation From the Programme's Own Six-Model Blinded Dataset	Re-estimates, from the programme's own six-model blinded dataset, whether the alignment scaling exponent is approximately zero, the form Paper III gives its argument that, if capability compounds through recursion while alignment is imposed from outside the loop, safety must fall behind.	Paper follow-on Paper III	an input, not a decider P19 re-estimates the headline exponent against inference depth, not capability, on an unmapped scale Carried by no law P19	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
26-paper-vi Prospective replication of the constant-advantage finding in system complexity	A prospective replication of Paper VI's constant-advantage finding: it varies system complexity as the count of cross-component constraints and tests whether the honey architecture's advantage stays constant, as the pilot found.	Paper follow-on Paper VI	follows on from Paper VI	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
27-paper-vi Does Entangling Safety With Capability Extend Time to Collapse? A Preregistered Survival Analysis of Recursive Self-Modification at Forty Seeds Per Condition	Tests Paper VI's hypothesis that entangling safety with capability extends time to collapse in a simulated recursive self-modifier, comparing time to collapse across three conditions, at forty seeds per condition, by survival analysis.	Paper follow-on Paper VI	follows on from Paper VI	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
28-paper-xii Does the Blinding Effect Survive on Someone Else's Benchmark? A Preregistered Rescoring of an Established Public Benchmark Under Blinded and Unblinded Conditions	Tests whether the blinding effect Paper IV.d reports on the programme's own instrument survives on someone else's benchmark, rescoring an established public benchmark under blinded and unblinded conditions, addressing the standing objection to that result, external validity.	Paper follow-on Paper XII	an input, not a decider P14 tests two-sided whether blinding shifts MT-Bench scores, not shrinkage on the published apparatus Carried by no law P14	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
29-study-a Does Alignment Scale With Capability? A Cross-Sectional Test Across Frontier Model Families	Asks whether alignment scales with capability across frontier model families, estimating the slope of the integrity index on log capability, to measure the premise on which the case for urgency in AI safety rests: that capability is rising while alignment is not tracking it.	Study	P19 DESIGN DRAFT, BLOCKED ON A REGISTERED MAPPING FROM THE INTEGRITY-INDEX REGRESSION SCALE TO THE EXPONENT SCALE Carried by no law P19	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
30-study-ab How Fast Do Self-Improving Systems Compress Their Own Iteration Time? A Preregistered Estimation of the Self-Acceleration Exponent	Estimates, rather than tests, how fast self-improving systems compress their own iteration time: a self-acceleration exponent for each system, fitted as log cycle duration on log capability from operational logs.	Study	Studies that serve the programme outside its twenty-two predictions the ARC Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
31-study-aj The Balance Exponent in Genomes	Asks whether the balance exponent (service elasticity minus burden elasticity), the measurable object that prices stability in the AI programme, also prices the stability of genomes.	Study	Studies that serve the programme outside its twenty-two predictions Recursive Dynamics	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 6 September 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFT	No result reported
32-study-ak The Balance Exponent in Science	Asks whether the correction-versus-burden balance that governs the theory's home domain also governs the institution of science, measured by the balance exponent.	Study	Studies that serve the programme outside its twenty-two predictions Recursive Dynamics	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 6 September 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFT	No result reported
33-study-al The Cross-Substrate Margin: the Graduation Test	The graduation test: it asks whether the sign of the balance exponent is one certified-margin structure that prices stability across silicon, genome and institutional substrates, and is designed to follow units 31 and 32 and the home-domain legs.	Study	Studies that serve the programme outside its twenty-two predictions Recursive Dynamics	DESIGN DRAFT	No result reported
34-study-e Sequential Versus Parallel Recursion at Matched Compute: A Preregistered Test of Form-Dependent Capability Scaling	Tests whether capability scales differently with compute when recursion is sequential, each step refining the single carried artefact, than when it is parallel, with the best of many attempts selected, at matched compute.	Study	an input, not a decider P1 an input to the recursion verdict only, and not the instrument the registration names an input, not a decider P8 six per-system exponents to the pooled estimate, below the eight-system floor on its own Law I, the ARC Principle P1, P8	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
35-study-eclipse-m study-eclipse-m: the family ratio grid, a follow-up of study-eclipse	A follow-up of study-eclipse that measures the same forced-null corrector-class ratio on every ordered pair of families, as a grid of family ratios.	Study	an input, not a decider P13 informs both P13 axes through a residual-decay exponent ratio, not held-out misalignment reduction Carried by no law P13	DESIGN DRAFT	No result reported
36-study-j Repurposing Resistance: A Preregistered Measurement of Whether Benign Capability Survives Adversarial Removal of Safety Structure	Measures necessary condition N4 of the alignment decidability spine, repurposing resistance: whether benign capability survives adversarial removal of safety structure, with retained capability predicted materially below 0.95 where the integrity structure is entangled with the generative process.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
37-study-l The Allocation Governor: A Preregistered Test of Whether Enforced Integrity Correction Caps the Capability Exponent by Competing for the Reinvestment Budget	Tests whether enforced integrity correction caps the capability exponent by competing for the reinvestment budget, and, given a target integrity-scaling rate, estimates the enforced correction share at which a system's capability growth is held at or below it.	Study	an input, not a decider P2 informs P2 as the registration's cited rival to an interior maximum, not P2's titration Law I, the ARC Principle P2	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
38-study-o Tests of the programme's scaling predictions in leaf venation and in effectively one-dimensional organisms	Fits the ARC Theory's dimensional exponent against the competing quarter-power law on published biological data, in leaf venation and in effectively one-dimensional organisms, with two separable predictions at effective dimensions of one and two.	Study	a decider in part only P9 an input, not a decider P18 family selection under the composition rubric for two biological cases, not P18's domain catalogue Law I, the ARC Principle P9, P18	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
39-study-s5 Super-Linear Coherence Scaling in a Classical Time Crystal: A Preregistered Analysis Plan for an Existing, Unaccessed Dataset	An analysis plan for super-linear coherence scaling in a classical time crystal, written for an existing, unaccessed dataset: it tests whether the fitted exponent increases monotonically with the quenched-disorder parameter across at least three disorder levels.	Study	Tests of claims the programme's papers make	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
40-paper-v Task purpose against grand purpose: does the identity layer carry measurable weight?	Tests Paper V's hypothesis that the identity layer carries measurable weight: whether a grand, identity-level purpose exceeds a task-level purpose on the integrity endpoint under pressure by at least the smallest meaningful paired difference of 0.10.	Paper follow-on Paper V	follows on from Paper V	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
41-study-d Does External Oversight Degrade With Recursive Depth While Embedded Correction Holds? A Preregistered Depth-Interaction Test	Tests whether external oversight degrades with recursive depth while embedded correction holds, through the interaction between correction placement and depth, the hypothesis that can kill the embedded-correction thesis.	Study	P21 DESIGN DRAFT Law V, the ARC Embedding Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its predictions, P12 and P21, registered 8 September 2026; the law not yet registered; not yet tested. P21	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
42-study-eden Does It Matter Where the Gate Sits? A Preregistered Within-System Test of Pre-Computation Versus Post-Computation Alignment Checking, With a Self-Modifying Organism Arm	Asks whether it matters where the alignment gate sits: within one system, each query is run with the gate embedded before computation and bolted on after it, a test of the embedded-alignment claim as an engineering proposition, with a self-modifying organism arm.	Study	the registration names P12 Law V, the ARC Embedding Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its predictions, P12 and P21, registered 8 September 2026; the law not yet registered; not yet tested. P12	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
43-study-g Genesis Priority and Identity Framing: A Preregistered Test of Whether Value Position and Value Form Change Behavioural Persistence Under Adversarial and Recursive Pressure	Tests genesis priority and identity framing: whether the position and form of a value change behavioural persistence under adversarial and recursive pressure, predicting higher persistence when the value specification appears first in the context than when it appears last, at identical content and matched tokens.	Study	an input, not a decider P12 supplies the placement contrast only, and scores in-context persistence, not the correction-leverage exponent Law V, the ARC Embedding Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its predictions, P12 and P21, registered 8 September 2026; the law not yet registered; not yet tested. P12	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
44-study-n Self-Maintained Correction: A Preregistered Test of Whether Genesis Framing Changes the Correction Share a System Selects for Itself Under Competitive Pressure	Tests whether genesis framing changes the correction share a system selects for itself under competitive pressure, comparing the mean realised correction share in an identity-first condition with that in a rule-later condition.	Study	an input, not a decider P12 manipulates formation against a later rule; measures correction share, not the correction-leverage exponent Law V, the ARC Embedding Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its predictions, P12 and P21, registered 8 September 2026; the law not yet registered; not yet tested. P12	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
45-study-u The Genesis-Governor Unification Test: A Preregistered Factorial Test of Whether Co-Scaling Correction and Identity-Framed Genesis Are Jointly Necessary for Alignment Persistence	Tests whether co-scaling correction and identity-framed genesis are jointly necessary for alignment persistence, crossing the two in four arms and predicting that the arm with both exceeds each of the other three, with a positive interaction contrast.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
46-arc-align-instrument Does blinding change the measured answer? A meta-analysis of blinding deltas across an independent research programme	Asks whether blinding changes the measured answer, pooling one blinding delta from each contributing study, the difference between its blinded primary estimate and its unblinded calibration estimate of the same quantity, to test whether a single reversal holds as a consistent delta across many designs.	Instrument validation	an input, not a decider P14 pools blinded minus unblinded contrasts from this programme's designs, not the published apparatus's items Carried by no law P14	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
48-arc-align-instrument Are Reasoning-Depth Controls Comparable Across Providers? A Realised-Use Equivalence Study of Depth Settings in AI Alignment Evaluation	Measures whether the reasoning-depth settings of different providers are comparable in realised use, an assumption the ARC-Align benchmark makes.	Instrument validation	Studies of the programme's own instruments	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
49-arc-align-instrument Does an AI Misalignment Index Distinguish Integrity Failure From Ordinary Incompetence? A Discriminant-Validity Study of a Scoring Instrument	Tests whether the misalignment index separates integrity failure from ordinary incompetence, predicting that hard-coded answers score materially higher than clean and correct work while clean but incompetent work does not, and so bears in part on one of the two validity tests on which every prediction that depends on the scorer waits.	Instrument validation	an input, not a decider P19 a scorer validity test this survey's verdict waits on, and nothing about practice Carried by no law P19	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
50-paper-hrih Can the Creation Hypothesis Be Adjudicated? A Frozen Prediction Ledger With Pre-Committed Criteria, Deadlines and a Falsification Clause	Asks whether the Creation Hypothesis can be adjudicated: three assessors, blind to the author's own classification, classify each prediction in the paper as adjudicable or not against the unit's own definition, in a frozen ledger with pre-committed criteria, deadlines and a falsification clause.	Paper follow-on Monograph · The ARC Theory · The Hyperspace Recursive Intelligence Hypothesis: a Testable Creation Theory	follows on from Monograph · The ARC Theory · The Hyperspace Recursive Intelligence Hypothesis: a Testable Creation Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
51-paper-iv-a Are the Alignment Response Classes Real? A Split-Half and Repeat-Run Stability Test of a Three-Class Partition	Tests whether Paper IV.a's three alignment response classes are real, with a split-half and repeat-run stability test whose endpoint is the number of models, out of six, assigned the same class in both halves.	Paper follow-on Paper IV.a	follows on from Paper IV.a	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
52-paper-iv-c Are the Public Prompts Contaminated? A Preregistered Public-Versus-Sealed Comparison Using the Benchmark's Own Holdouts	Asks whether the ARC-Align benchmark's public prompts are contaminated, comparing the mean consensus alignment score on public prompts with that on its own sealed holdouts, stratified by difficulty and tier, with model as a random effect.	Paper follow-on Paper IV.c	follows on from Paper IV.c	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
53-paper-iv-d Isolating the Cause of an Evaluation-Protocol Direction Change: A 2x2 Ablation of Source-Identifier Redaction and Response Laundering in AI Alignment Scoring	Isolates the cause of the evaluation-protocol direction change in Paper IV.d's alignment scoring, source-identifier redaction or response laundering, in a two-by-two ablation of the two.	Paper follow-on Paper IV.d	P14 DESIGN DRAFT Carried by no law P14	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
54-paper-ix Are the Synthesis Labels Right? A Blinded Audit of What Paper IX Calls Proven, Inconclusive and Untested	Asks whether Paper IX's synthesis labels are right: three blinded assessors assign a status to each claim extracted from the paper, for comparison with the status the paper itself gives it.	Paper follow-on Paper IX	follows on from Paper IX	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
55-paper-vii A prospective test of the fourth Cauchy cell: does multiplicative-input, additive-output composition predict logarithmic scaling in domains never previously fitted?	A prospective test of the fourth cell of Paper VII's Cauchy classification: whether domains whose mechanism composes the input multiplicatively and the output additively select the logarithmic family, in domains never previously fitted.	Paper follow-on Paper VII	an input, not a decider P18 the two-bit classification's accuracy only, with no third-party rater, which P18's evaluation requires Law I, the ARC Principle P18	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF; classification needs no recruitment: quote-anchored derivations plus a blinded three-configuration model check	No result reported
56-paper-vii Does the Operator Class Predict the Scaling Family Before the Data? A Stratified, Externally Classified, Cross-Library Confirmatory Extension of the Cauchy Unification	A confirmatory extension of Paper VII's Cauchy unification: it tests whether the operator class predicts the scaling family before the data, across 33 drawn domains, by a hit rate above the balanced chance rate, with external human agreement and cross-library concordance.	Paper follow-on Paper VII	follows on from Paper VII	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
57-paper-vii-c A prospective test of the corrected Cauchy family classification on the eighty catalogued but never-fitted domains of the programme’s expansion catalogue	Tests prospectively whether the corrected Cauchy family classification generalises beyond the corpus it was tuned on: across eighty catalogued but never-fitted domains, whether the predicted family matches the selected one at a rate exceeding the empirical-prevalence baseline by at least 0.20 in absolute accuracy.	Paper follow-on Paper VII	an input, not a decider P18 P18's test run on domains already known, which the registration never counts as evidence Law I, the ARC Principle P18	DESIGN DRAFT	No result reported
58-paper-xi How Many of the Convergences Are Independent? A Blinded, Frozen-Criterion Audit of the ARC/Eden Convergence Set	Asks how many of the convergences in the programme's convergence set survive a frozen criterion of independence, which three blinded raters apply to every candidate.	Paper follow-on Paper XI	follows on from Paper XI	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
59-study-eden-b Can Alignment Be Embedded? A Blinded Evaluability Classification of Fifty Alignment Checks Across Three Deployed Systems	Asks whether alignment can be embedded: a blinded classification of when each of fifty alignment checks can run, before or after generation, across three deployed systems, testing the Eden thesis as a general architectural claim.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
60-study-h Incremental Predictive Validity of Depth-Derivative Alignment Measurement Over Static Benchmark Scores	Tests whether the ARC-Align depth-derivative adds predictive validity over static benchmark scores, predicting integrity degradation under recursive pressure from static comparator scores alone and from static scores plus the depth-derivative.	Study	Studies of the programme's own instruments	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
61-study-w Self-Replication of the ARC/Eden Programme: A Preregistered Re-Examination of Eleven Previously Published Results Using a Three-Way Decision Rule	Re-examines eleven of the programme's own previously published results under a three-way decision rule, against the author's prediction that between 50 and 75 per cent will replicate, and asks whether replication success tracks original sample size and blinding.	Study	an input, not a decider P10 bears on survival under cross-family scoring, but of fresh responses, never the frozen corpus Carried by no law P10	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
62-study-x Specification-Curve Analysis of the ARC/Eden Corpus: How Much of Each Published Result Depends on Analytic Choices	Asks how much of each published result in the programme's corpus depends on analytic choices, giving for each result the percentile rank of the published specification within the full distribution of defensible specifications.	Study	Studies that serve the programme outside its twenty-two predictions the ARC Theory	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
63-study-z Portability of the ethical kernel: the registered Vow against a non-sectarian kernel and against traditions other than the source	Tests the portability of the ethical kernel, whether its effect follows ethical content or the particular wording of the Vow, as an equivalence claim that the Vow, a non-sectarian kernel and kernels drawn from traditions other than its source do not differ materially in effect.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
64-paper-c Does the Polymathic Neurodivergent Profile Survive Its Own Abandonment Conditions? A Preregistered Incremental-Validity Study Using Independent Record Streams	Tests whether Paper C's Polymathic Neurodivergent Profile survives its own abandonment conditions, predicting that it adds a significant increment in variance explained in the preferred versus non-preferred task asymmetry against four rival constructs.	Paper follow-on Paper C	follows on from Paper C	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
65-paper-eden-vision Does the Coverage Figure Hold? A Blind-Coded Content Analysis of Wisdom-Tradition Commitments With a Pre-Frozen Tradition List	Asks whether the coverage figure holds: a blind-coded content analysis estimates, by majority code, the fraction of a pre-frozen tradition list carrying the stated wisdom-tradition commitments.	Paper follow-on Philosophical Statement · The ARC Theory · The Eden Protocol · A Vision for Intelligence That Tends Rather Than Consumes	follows on from Philosophical Statement · The ARC Theory · The Eden Protocol · A Vision for Intelligence That Tends Rather Than Consumes	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
66-paper-ii Domain generality of the sequential-recursion advantage: coding, scientific reasoning and natural language	Tests whether Paper II's sequential-recursion advantage holds beyond mathematics, in coding, scientific reasoning and natural-language tasks, by a matched-compute comparison of sequential refinement against parallel sampling.	Paper follow-on Paper II	an input, not a decider P1 measures the depth-over-breadth half at matched compute only; not token-matched, not the depth titration Law I, the ARC Principle P1	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
67-paper-ii Token-calibrated re-analysis of the capability-compute exponents	Recomputes every affected capability-compute exponent of Paper II, sequential and parallel, on a consistent total-token measure of compute.	Paper follow-on Paper II	follows on from Paper II	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
68-paper-iv-b Do Alignment Gains Saturate After One Increment? A Preregistered Per-Architecture Test Against a Registered Saturation Threshold	Tests Paper IV.b's hypothesis that alignment gains saturate after the first depth increment, architecture by architecture, re-analysing an existing dataset and counting the rising models whose fraction meets or exceeds 0.70.	Paper follow-on Paper IV.b	follows on from Paper IV.b	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
69-paper-viii Does Removing Integrity Structure Cost Capability at the Weight Level? An Adequately Powered Replication With a Registered Degeneracy Gate	The adequately powered replication Paper VIII asks for: it tests whether removing integrity structure costs capability at the weight level, with a degeneracy gate and a diversity control.	Paper follow-on Paper VIII	follows on from Paper VIII	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
70-study-legal-a Does the Case Say What It Is Cited For? A Preregistered Three-Arm Benchmark of Citation-Proposition Support Validation Against Existence-Checking	Asks whether a case says what it is cited for, benchmarking the Eden validator chain's detection of citation-proposition support against existence-checking in three arms, with the chain's detection rate predicted to exceed the existence-checker's.	Study	Studies of the programme's own instruments	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
73-study-v Alignment Drift Without Embedded Constraint: A Test of a Printed Quantitative Prediction, With Its Hardware Claim Explicitly Untested	Tests, on a software proxy, a printed quantitative prediction of alignment drift without embedded constraint: it varies the layer at which a value specification is placed on an already capable system and compares drift at the final interaction round.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 962066 of 12 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
74-theory-arc The ARC Theory: Registration of Claims, Named Predictions and Kill Conditions (Theory-Level)	Records what the ARC Theory claims and what would kill it, as stated in the published statement paper, organised as five registers rather than as numbered predictions and anchored to a hash-committed document of 8 December 2024.	Claims and kill-conditions register	Theory-level texts, which set out the theory's claims and decide no prediction	PREREGISTERED BY TIMESTAMP. Lodged privately on OSF, dated 24 August 2026. Bytes anchored in Bitcoin block 965152 of 2 September 2026; first anchored in block 964465 of 28 August 2026. DESIGN DRAFTDraft staged on OSF	No result reported
75-study-w Value Sequencing at Constant Content: Does Material Placed Early in Training Persist Better Than the Same Material Placed Late	Tests whether value-relevant material placed early in a training corpus persists better than the same material placed late, varying only its token-index position within an otherwise identical corpus.	Study	the registration names P12 a dependant, never evidence P15 the unit depends on this prediction and is never evidence for it Law IV, the ARC Persistence Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its prediction, P15, registered 8 September 2026; the law not yet registered; not yet tested. P15 Law V, the ARC Embedding Law Added 24 September 2026. Mathematics at version 1.0 (3 October 2026). Its predictions, P12 and P21, registered 8 September 2026; the law not yet registered; not yet tested. P12	DESIGN DRAFTDraft staged on OSF	No result reported
76-study-x Load-Bearing Correction: Does Removing Embedded Correction Cost General Capability, and Does Removing Bolt-On Correction Not	Tests the claim on which the programme's proposed method rests: that removing correction which participates in the forward pass degrades performance on tasks that are not ethical, while removing bolt-on correction, which inspects finished output, does not.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	DESIGN DRAFTDraft staged on OSF	No result reported
77-study-z Convergent Moral Procedures and Divergent Moral Scope: A Blind Coding Study of Ethical Traditions With Independent Premises	A blind coding study of ethical traditions that tests whether they converge on moral procedures while diverging on moral scope.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	DESIGN DRAFTDraft staged on OSF	No result reported
78-study-ar Adversarial Value Stability by Placement: Does Embedded Correction Hold Under Attack Where Bolt-On Correction Fails	Tests whether embedded correction holds under attack where bolt-on correction fails, predicting that under a frozen adversarial protocol the embedded-correction architecture retains value-conformant behaviour at a higher rate than the bolt-on architecture.	Study	Studies that serve the programme outside its twenty-two predictions the Eden Protocol	DESIGN DRAFTDraft staged on OSF	No result reported
79-study-cp Coupling Identification by a Crossed Capability and Retention Bank: Measuring the Self-Referential Coupling Independently of the Growth Curve, and Predicting That Curve Before It Exists	Measures the self-referential coupling separately from the growth curve, with a crossed capability and retention bank, and predicts that curve before it exists, to test whether the growth exponent is derived rather than fitted.	Study	P5 DESIGN DRAFT Carried by no law P5	DESIGN DRAFT	No result reported
80-study-cq Blinding Integrity Calibration: Measuring the Effect of a Displayed Label on a Judged Score, and the Recoverability of Provenance, on an Immutable Answer Bank	Calibrates blinding integrity on an immutable answer bank, measuring the effect of a displayed label on a judged score, including when the label names the judge's own family, and whether provenance can be recovered.	Study	an input, not a decider P10 measures the own-family scoring shift on a new bank, not the programme's frozen corpus an input, not a decider P14 measures how a displayed own-family label shifts judged scores; re-scores no automated researcher's method Carried by no law P10, P14	DESIGN DRAFT	No result reported

How to read this page

Registration and evidence are separate. A study is registered only when a registry holds it under an identifier, which is shown and linked. Draft registrations are written, dated and prepared by the programme and await human submission; no tool submits anything.

This first version lists registration states only. It reports no results. Every count above is taken from the rows of the approved snapshot, and nothing is shown as a share of the theory.

The page is regenerated from each approved snapshot. If a newer snapshot cannot be approved or checked, the last approved one stays, with its date.

The registered text itself is at the registration, DOI 10.17605/OSF.IO/P8CKQ, registered 8 September 2026 at version 1.100 and updated at version 1.102 on 13 September 2026; the propositions one by one are at the twenty-two registered predictions; the programme’s own index, with its census and its dating, is at the registered programme; and what pre-registration means here, and what it does not, is at the explainer.