Skip to main content
The Mission · Research

The research.

This page separates completed evidence from active work and longer-term plans. Published figures are bound to the recorded analysis, and the limits stay beside the result.

← The Mission

By Michael Key · ORCID

What is finished, active, and still waiting.

This is the canonical status record for the research program, not a promise of pace. A contribution aims to add something beyond reproduction; a validation study tests the workflow against later outcomes or published benchmarks. Updated .

Study 01 · Contribution · Completed

Molecular Aggregation & Variant Effect

Twelve public-data investigations produced a calibrated Aβ variant-effect model, plus negative results on cross-disease transfer and representation scaling. Reproducibility hardening passed on 2026-07-20.

Next gate: Reviewed public code and provenance release.

Study 03 · Validation · Completed

HRS Cognitive Decline

Eight investigations tested calibrated cognitive-decline forecasting against a later Langa-Weir research classification using registered public HRS releases. Reproducibility hardening passed on 2026-07-20.

Next gate: Reviewed public code and provenance release.

Study 05 · Contribution · Active · pre-publication

PRO-ACT ALS Prognosis

This study asks how much confidence to place in a forecast of ALS functional decline, for whom, and when it should be withheld. The pre-final baseline has passed review and the final test remains sealed. A sensitivity result was invalidated before publication and awaits a fresh researcher-run analysis. No public result claim has been authorized.

Next gate: Valid sensitivity parent result, conditional-reliability audit, then the sealed final-test and publication gates.

Study 02 · Contribution · Deferred

Longitudinal ALS Speech Prognosis

The scientific question remains open, but PRO-ACT contains no audio and this work has not started.

Next gate: Lawful longitudinal speech access and an appropriate clinical or formal research partnership.

Study 04 · Contribution · Planning

ADNI MCI Conversion

Planning and data-locality work only. No public analytical result exists.

Next gate: Study-specific access, frozen verifier, and a passed feasibility checkpoint.

Evidence before announcement.

The integrity bar

Every published study number is generated by a versioned script, recorded in a script-written results.json, and tied to a checksum-locked input. The internal audit chain is complete; a reviewed public code and provenance release is still pending. Until then, the package is available on request and the site does not claim that an outside visitor can rerun it without that package.