Skip to main content
THE DECODE INSTITUTEPrecision Neuroscience

We use AI to read minds medicine misread.

For 100 years, the frameworks were built on the wrong data. Incomplete. Biased. Deficit-first. We're rebuilding them — condition by condition, evidence by evidence.

1 in 6
People are neurodivergent globally
Global prevalence
91.8%
AI detection sensitivity — autism
Multimodal neuroimaging
6 papers
Publication pipeline, 30 months
DAI research arc
4 conditions
Autism · Schizophrenia · OCD · ADHD
Research pipeline

The problem, the method, the goal

THE PROBLEM

100 years of wrong frameworks.

Diagnostic models built on incomplete data. Conditions treated as deficits. People harmed by the tools meant to help them. The frameworks are broken. We are here to replace them.

THE METHOD

AI as code-breaker.

Multi-agent synthesis. Evidence-only research. Built in public. Every claim cited. Every model stress-tested by 15 specialist agents before it goes anywhere near a paper.

THE GOAL

Frameworks that actually serve people.

Precision models that replace blunt instruments. Detection tools that work across race, gender, and age. Research built with the people it studies — not on them.

1 in 6
People are neurodivergent globally
91.8%
AI detection sensitivity — autism (multimodal neuroimaging)
6 papers
Publication pipeline, 30 months
4 conditions
Autism · Schizophrenia · OCD · ADHD

PROJECT 01 — DIMENSIONAL AUTISM INTELLIGENCE

Questioning the spectrum

A framework in development — drawing on published neuroscience, emerging AI detection data, and lived experience.

01

Wrong Ruler

70 years of linear spectrum. Built on 1970s data from white boys. Flattens extraordinary complexity into a single axis.

02

Evolutionary Insight

2025 Stanford: autism genes concentrate in L2/3 IT neurons — the fastest-evolving neurons in the human brain. Not deviation. Human evolution at full intensity.

03

A Different Proposal

Dimensional Autism Intelligence proposes replacing severity scoring with a cognitive fingerprint — 8 independent axes instead of one line. Systemising. Sensory. Monotropism. Communication. Pattern Recognition. Executive. Language. Emotional Regulation.

04

AI Shows Promise

Multimodal neuroimaging: 91.8% sensitivity reported. EEG deep learning: AUC 0.937. Eye-tracking: ~88%. Detection tools exist in the literature. The diagnostic framework may need to catch up.

ACTIVEPROJECT 01

Dimensional Autism Intelligence

The world measured autism on a single line. We replaced the line with 8 independent axes. Not a spectrum. A cognitive fingerprint.

Enter Project DAI

Hover over the nodes to explore the 8 dimensions of the DAI Model.

PROJECT 01 — DIMENSIONAL AUTISM INTELLIGENCE

Eight independent cognitive dimensions

Not necessarily a single spectrum — possibly a cognitive fingerprint. Hover each proposed dimension to explore.

  • Systemising Intelligence: Drive to analyse, build systems, and understand rules. Pattern creation and technical design.
  • Sensory Processing Fidelity: Higher-resolution sensory perception. Not dysfunction — different operating parameters.
  • Monotropism Index: Depth of focus architecture. Flow states as default. Attention optimised for depth, not breadth.
  • Social Communication Style: Directness, literal processing, authenticity-driven interaction. Not deficit — different language.
  • Pattern Recognition Speed: Ability to detect connections, anomalies, and structures others miss. Data archaeology.
  • Executive Function Profile: Task initiation, working memory, temporal awareness. Neurotype-specific support structures.
  • Language Processing Architecture: Different input/output modes. Internal language precision vs. external expression timing.
  • Emotional Regulation Strategy: Need for predictability, sensory regulation, and autonomy in managing arousal. Not fragility.

THE RESEARCH PIPELINE

Four conditions. One method. Built in sequence.

Decode does not pick conditions randomly. Each project targets a condition where the current diagnostic model is provably inadequate, where AI detection tools already exist but are under-deployed, and where the people diagnosed have been harmed by the frameworks meant to help them.

ACTIVE01

Dimensional Autism Intelligence

The world measured autism on a single line built from 1970s data. We replaced the line with 8 independent axes. Not a spectrum — a cognitive fingerprint.

Papers 1–6 · 2026–2028
ENTER PROJECT
IN PREPARATION02

Schizophrenia: Decoding the Signal

The most misunderstood condition in psychiatry. What if the so-called hallucinations are a different kind of signal processing — not a malfunction, but a mismatch?

Opens 2027–2028
FOLLOW PROGRESS
PLANNED03

OCD: Precision, Not Pathology

When the threat-detection system never gets to rest. A rigorous look at whether the deficit framing has obscured what is actually a hyperactive protective system.

Opens 2028–2029
Coming soon
PLANNED04

ADHD: Attention Optimised for Urgency

Built for a world that no longer exists. The question is not why people with ADHD struggle in standard environments — it is why we still call that struggle a disorder.

Opens 2029–2030
Coming soon
View all projects →
BUILT IN PUBLIC

We show our work.

Most research institutions publish results. We publish the process. Every agent debate transcript. Every methodology decision. Every time we change our mind because the evidence said so.

This is not a PR strategy. It is an accountability mechanism. If the work is sound, it survives scrutiny. If it is not, we want to know.

Read the latest research updateSee how the agent team worksFollow the publication pipeline