Introducing AQAL v0.1

The Foundation Model for the Neurodiverse Brain

AQAL is the world's first AI foundation model that predicts how autistic minds experience sight, sound, and language — mapping 20,484 cortical points in real time.

1,545
brain scans
20,484
vertices
1,002
connections

Understanding Neurodiversity at Scale

Understanding how autistic brains process the world has required individual brain scans — expensive, slow, and impossible for many.

AQAL changes this. A foundation model that predicts how any neurodiverse brain responds to any stimulus — transforming months of lab work into seconds of computation.

Neurotypical
NT brain
Neurodiverse
ND brain
Same stimulus, different neural activation patterns
Live Demo

See AQAL in Action

Stimulus
Neurotypical
Neurodiverse
Door opens
12% divergencelow
Peak: Visual at 18% divergence
t = 0s · Overall 12% NT–ND divergence
Visual18%
Auditory8%
Salience10%
Motor5%
Social7%
Default Mode6%

Low sensory load. The visual cortex registers the door opening — both NT and ND brains respond similarly at this point. Minimal divergence.

VisualAuditorySalienceMotorSocialDefault Mode

AQAL Architecture

A proprietary tri-modal pipeline purpose-built for predicting neurodiverse brain activity.

Video
Audio
Text
Encoders
Transformer
ND Mapper
Brain output
01

Multi-Modal Encoding

Specialized encoders process sight, sound, and language independently.

02

Universal Integration

A deep transformer fuses modalities into brain-aligned representations.

03

Neurodiverse Mapping

Transforms predictions to neurodiverse patterns from real brain recordings.

Connectivity Maps

Heatmaps showing where brain wiring differs between autistic and neurotypical individuals. Generated from 1,545 real fMRI brain scans across 36 clinical sites.

Brain Region

One of 100 parcels the brain is divided into. Each handles a specific function — vision, movement, language, emotion, etc.

Connection

A pair of brain regions that communicate with each other. With 100 regions, there are 4,950 possible pairs to test.

Connectivity

How strongly two regions are linked. Measured by how closely their activity patterns correlate over time in an fMRI scan.

t-statistic

A number measuring how different ASD and TD groups are for a given connection. Larger = bigger difference. Can be positive (ASD stronger) or negative (TD stronger).

FDR Correction

A noise filter. When you test 4,950 connections, ~248 will look significant by pure chance. FDR ensures at most 5% of your findings are false alarms.

Network

A group of brain regions that work together. The 7 major networks: Visual, Somatomotor, Dorsal Attention, Salience, Limbic, Control, and Default Mode.

FDR-Corrected

1,002 Significant Connections

q < 0.05
FDR-corrected connectivity heatmap

Each bright pixel represents a statistically reliable wiring difference between ASD and TD brains. Red = stronger in ASD, blue = stronger in TD. Only connections surviving FDR correction are shown.

How to read this map

Each row and column is one of 100 brain regions, grouped by network (labels on axes).

A bright red pixel at row X, column Y means that connection is significantly stronger in autistic brains.

A bright blue pixel means it's significantly stronger in neurotypical brains.

A dark/black pixel means no reliable difference was found for that connection.

The purple grid lines separate the 7 brain networks so you can see which networks are most affected.

Network Breakdown

Most Affected Brain Networks

Network-level connectivity differences

Default Mode Network shows the most affected connections (511), followed by Somatomotor (367) and Dorsal Attention (300). This aligns with known autism neuroscience — DMN alterations affect self-referential processing and social cognition.

Default Mode511

Self-reflection, mind-wandering, social thinking. Most affected — autistic brains show altered internal processing.

Somatomotor367

Movement and body awareness. Differences here relate to motor planning, stimming, and physical coordination.

Dorsal Attention300

Focus and concentration. Altered wiring can explain hyperfocus on interests or difficulty shifting attention.

Visual265

Processing what you see. Autistic brains often notice more detail but can be overwhelmed by busy scenes.

Salience252

Filtering what matters vs background. Altered salience = difficulty ignoring the AC hum when someone is talking.

Control155

Planning and decision-making. Relates to preference for routine and difficulty with unexpected changes.

Limbic154

Emotion and reward. Differences here affect emotional regulation — feeling things more intensely or differently.

Uncorrected (all 4,950)

Full Connectivity Landscape

reference
All connectivity differences

All 4,950 connections before FDR filtering. Many of these are noise — the FDR map above shows only the reliable 1,002.

AQAL Scaling Laws

AQAL follows a scaling law: accuracy increases log-linearly with more brain data. No plateau reached.

50
100
200
400
871
1,545
10K
100K
400K
Subjects in training

What AQAL Understands

AQAL maps how sensory processing differs across six key brain networks.

Visual

55% divergence

Auditory

72% divergence

Social

91% divergence

Default Mode

48% divergence

Salience

63% divergence

Motor

85% divergence

Powered by AQAL

The foundation model enables practical tools for autism accessibility.

Sensory Audit

Live

Upload video of any space. Get second-by-second sensory analysis, accessibility score, and actionable recommendations.

Brain Comparison

Live

Side-by-side neurotypical vs neurodiverse brain activation maps for any stimulus input.

Sensory Passport

Coming Soon

Personalized sensory profile through 5-minute calibration. Portable document for schools and clinics.

Neurotrack

In Development

Therapy progress tracking with dashboards, developmental milestones, and sensory-linked reporting.

1,545
Brain recordings
1,002
Neural connections
20,484
Cortical points
7
Networks profiled

Roadmap

AQAL v0.1Seed

Foundation model trained. Statistical brain transform. Live API.

AQAL v0.2Root

FDR-corrected. Site harmonization. Age stratification. Bootstrap uncertainty. Single consortium (871 subjects).

AQAL v0.3TrunkCurrent

Dual-consortium (1,545 subjects, 36 sites). 1,002 FDR connections. Deployed.

AQAL v0.5Sprout

GPU fine-tuning. Sensory subtypes. Video input. Age-specific models.

AQAL v1.0Bloom

Clinically validated. 10K+ subjects. Sensory Passport. Published metrics.

AQAL v2.0Canopy

Real-time processing. EEG integration. Wearable support. 100K+ subjects.

AQAL v3.0Forest

Foundation model for all neurodiversity. ADHD, SPD, anxiety. 400K+ subjects.

Build with AQAL

API access for researchers, clinics, schools, and companies building accessible products.