Consolidated research catalog · v0.10.0

What the evidence supports

Independent catalog rebuild and content review accepted.

EEGT describes recurring numerical patterns in public around-ear voltage recordings. Different representations can agree on some measurements while disagreeing on others. The work has not established a universal token vocabulary, semantic meaning, a brain-state diagnosis or thought decoding.

The source catalog

89 candidate recordings; 88 qualified recordings; 476.053296 qualified source hours. 0 qualified recordings have unknown duration. Reusing a recording in another experiment adds no people, nights or source hours.

Dataset	Status	Recordings	Source-local people	Sessions	Known hours	Unknown durations
ds004015	QUALIFIED	36	36	36	47.289732	0
ds005178	QUALIFIED	32	10	32	238.232252	0
ds005185	QUALIFIED	2	1	2	14.350478	0
ds005207	QUALIFIED	18	18	18	176.180833	0
ds005207	QUARANTINED	1	1	1	UNKNOWN	1

People counts are local to each dataset and status group; they are not globally unique identities and must not be summed as such. Source hours, quality windows, selected support, event rows and independent people are separate denominators. Missing and quarantined records remain visible.

The final fixed validation

Experiment 013 censused 9,600 candidates and selected 174 blocks. Complete people: 6 in new sessions (READY); 2 in new people (INSUFFICIENT_PARTICIPANTS, fixed minimum 3). All 8 endpoints remain visible.

Cohort	Model	Metric	Status	People	Mean effect	Adjusted p
new_people	cbramod	change	NOT_ESTIMABLE	0	not estimable	not estimable
new_people	cbramod	geometry	NOT_ESTIMABLE	0	not estimable	not estimable
new_people	codebrain	change	NOT_ESTIMABLE	0	not estimable	not estimable
new_people	codebrain	geometry	NOT_ESTIMABLE	0	not estimable	not estimable
new_sessions	cbramod	change	COMPARED	6	0.007455065359477121	1.0
new_sessions	cbramod	geometry	COMPARED	6	0.03799173178043146	0.5
new_sessions	codebrain	change	COMPARED	6	-0.061683006535947715	0.25
new_sessions	codebrain	geometry	COMPARED	6	0.012695927653357994	1.0

0 of 4 estimable comparisons clear their adjusted threshold. Signed effects and model differences remain visible; no between-model significance is inferred. Continuous encoders are not discrete language models.

What earlier results mean

The original 11.4% was 393 accepted assignments among 3,458 external scalp windows, a coverage measure. It was not diagnostic accuracy or a fraction of a person's brain activity. Later source, preparation and representation changes prevent a controlled improvement claim. Experiments 011 and 012 also retained inconclusive adjusted results. Dense landmarks can match after phase randomization, and cycles occur in noise; matching alone does not prove shared biological state.

Inspect and rebuild

Independent review and exact hashes

Versioned catalog and checksum downloads · Offline rebuild instructions · Example SQL queries · Source-linked research synthesis

The small package includes the database builder and every hash-bound curator input. The larger model/event archives stay in their original releases. Rebuilding the catalog runs no model or detector and downloads no waves.

Remaining boundaries

Physiological calibration, exact reference/gain/clock compatibility, model-pretraining overlap and physical Neurable equivalence remain UNKNOWN. Private NoticingMind waves, personal history and semantic/task labels did not enter discovery. More independent people and acquisition setups, with a method frozen before access, are needed for stronger replication. Meaning requires a separate designed study. DOI metadata is prepared; no issued DOI is claimed.