Research corpus · completed and executing inquiry

Ideas should leave evidence, not just a queue.

This registry records research candidates that have received substantive prior-art, adversarial or empirical work. The terminal product may be a paper, research note, benchmark, replication, negative result or retirement record.

Corpus rule

Test it, narrow it, execute it, publish the defensible result, or retire it.

The corpus is not a publication-count scoreboard. It exists so machine-assisted abundance does not create a warehouse of half-researched claims. Negative and narrowed results belong here beside successful papers.

Read the 1 + Machine Method and completion lifecycle →

Executing

Research with a concrete next action.

Non-terminal work remains here only when its next executable step and stopping condition are explicit.

AR-2026-005

Attention Allocation Track A v0.2: multi-cohort source feasibility

Question: Can enough selection-neutral historical candidate state be reconstructed across ICLR cohorts to support at least 4,600 exact-outcome-observable candidates before downstream outcomes are opened?

Prior-art result: Historical-version reconstruction and bibliographic observability are treated as measurement problems rather than assumed properties of current public records.

Current finding: SOURCE_FEASIBILITY_INADEQUATE_NEUTRAL_COHORT_REQUIRED. Bulk historical-text acquisition remains off until a neutral additional cohort is recovered.

Next: Recover a neutral ICLR 2021 deadline/pre-review state or preregister an additional comparable cohort, then rerun source and precision gates.

Current scarcity: selection-neutral historical state

Research object Protocol / source

blocked-external

AR-2026-006

Cognitive Kernel consequence pilot

Question: Does the Agalmic Cognitive Kernel reduce active human attention or downstream rework on real tasks without degrading verified completion or authority boundaries?

Prior-art result: Matched experiments, counterbalancing, productivity measurement and human-AI collaboration evaluation are established. The local value is the outcome/attention instrumentation and explicit isolation contract.

Current finding: Protocol and execution evidence are frozen; no efficacy result has yet been earned.

Next: Execute matched real-task pairs under the frozen consequence protocol and report descriptive downstream outcomes.

Current scarcity: consequence-bearing evidence

Research object Protocol / source

executing

Terminal record

Inquiry that changed the claim.

Terminal does not mean “proved true.” It means the present research episode reached a durable disposition rather than remaining an unattended idea.

AR-2026-001

Epistemic Handoff: prior-art boundary and narrowed protocol

Question: Is epistemic handoff a distinct institution, and if not, what narrower contribution survives established knowledge-transfer and expert-routing literatures?

Prior-art result: Generic knowledge handoff, brokerage, absorptive capacity, innovation intermediation and expertise routing are established antecedents.

Adversarial result: The broad novelty claim failed. A narrower claim-sensitive, provenance-preserving protocol may remain useful if it explicitly represents the authority deficit and requested epistemic function.

Disposition: research note + narrowed draft; broad novelty retired

Research cost: legacy costs incomplete

Result Evidence / record ↗

narrowed

AR-2026-002

Economics of epistemic abundance: formal-boundary null test

Question: Does the proposed economics of epistemic abundance currently require a distinct Agalmic economic primitive beyond established complementarity, bottleneck, production and innovation theory?

Prior-art result: Mature economic antecedents already explain the generic movement of binding constraints when one input becomes cheaper.

Adversarial result: The proposed distinct law/primitive did not survive the formal-boundary null test in its current form.

Disposition: null result; empirical measurement question preserved as an option

Research cost: legacy costs incomplete

Result Evidence / record ↗

null-result

AR-2026-003

Expert-Attention Handoff Triage: retrospective scarcity benchmark

Question: Can pre-expert triage increase useful evaluation yield under constrained review capacity without unacceptable false negatives, status bias or presentation sensitivity?

Prior-art result: Generic pre-filtering, AI idea screening, editorial triage, selective prediction, learning to defer, active learning and value-of-information are established. The possible contribution is a claim-sensitive allocation benchmark and test.

Adversarial result: The special mechanism should be retired if simple rules or conventional model triage achieve equivalent recall/yield at lower complexity.

Disposition: phase-1 result preserved; empirical lineage absorbed into Attention Allocation Under Cognitive Abundance

Research cost: prospective cost capture required

Result Evidence / record ↗

narrowed

AR-2026-004

Attention Allocation Track A v0.1: pre-outcome feasibility test

Question: Can a transparent allocation-policy comparison over historical ICLR candidates reach the preregistered five-percentage-point precision target without temporal leakage, selection-observability failure or lineage contamination?

Prior-art result: The mechanism is grounded in active learning, selective labels, constrained ranking, exploration/exploitation and bibliographic record-linkage rather than treated as a novel allocation field.

Adversarial result: The study passed technical availability and lineage gates, but 0/5 primary policy contrasts were prospectively precise enough for the frozen five-percentage-point smallest effect. Citation outcomes therefore remained sealed.

Disposition: negative feasibility result; successor must be prospectively larger rather than weakening v0.1 gates

Research cost: partial cost capture

Result Evidence / record

null-result

Cost & scarcity

Research should expose what it consumed.

New investigations should prospectively record curator time, machine/compute cost, paid data/software, external expert time and important unpriced constraints where practical.

Each record also states which scarcity the work attempted to displace, which inherited abundance was used, and which bottleneck appeared next. Historical records may honestly say that costs were not measured rather than invent retrospective precision.

Machine-readable registry

The corpus is data as well as navigation.

The structured registry can be reused by future agents, dashboards and Memory Palace views to identify what has been tested, what survived and what remains blocked.

Research corpus JSON ↗