human NWB files
A pinned 6.2 GB public memory dataset from 59 participants.
ILUVORA · DIGITAL BRAIN LAB
We are already working
toward a digital brain.
Iluvora today is a transparent AI reconstruction built from a person’s real materials—not a copied brain. In parallel, our research programme tests what neural memory models can and cannot reproduce from public biological data.
Open research · results include failures
Public source dataFrozen tests before conclusionsNo claim beyond the evidence
EVIDENCE LOG
What exists today
Real datasets, reproducible experiments and deliberately strict controls—not a concept render.
human cortical segments
A reproducible H01 structural sample with 182 automated contact annotations.
rehearsed model fits
The complete evaluation route survived an intentional crash and resumed identically.
independent visual sessions
Our first mouse V1 model failed its held-out benchmark. We kept and published the negative result.
METHOD
The path, without shortcuts
IN PROGRESS
Measure
Start with public recordings of real neural activity and connectomics—not invented ‘digital neurons’. Today this covers tiny fragments, not a whole person.
DEVELOPMENT
Model
Build a memory-state model that receives an experience, preserves a trace and predicts a later neural response without seeing the answer.
REHEARSED
Try to disprove it
Compare against image-only, reset and shuffled-memory controls on people and stimuli excluded from training.
GATED
Scale only after success
A model that fails a small honest test does not earn a larger simulation. Scale follows evidence, not ambition.
SCIENTIFIC BOUNDARY
What we are not claiming
- We have not copied a consciousness.
- We are not simulating all ~86 billion neurons.
- A structural map is not a functioning mind.
- A convincing AI persona is not proof of personal continuity.
NEXT GATE
The next real milestone
Pass a locked human-memory benchmark: predict held-out neural responses better than equally capable controls. If it fails, that failure becomes the next design constraint.