Next Brain Research

Build the evidence
before the claim.

We use the execution platform to study which neural details must remain explicit, which can be approximated, and how results survive changes in runtime and scale.

Read research notes Meet Next Brain Lab
01 · ModelEquations and structure
02 · InterventionChange one boundary
03 · EvidenceRecord result and limits

Featured publication · arXiv v1

A Unified Intermediate Representation and Execution Architecture for Heterogeneous and Distributed Neural Simulation

A shared model-to-execution architecture, with measurements across CPU, GPU, browser and distributed systems. Xinjun Li · Next Brain · 10 October 2026.

Explore the engine
4.13M / 24.1Bneurons / recurrent synapsesMacaque multi-area engineering run
139,255FlyWire neurons15.1M weighted connectome edges
87.14%MNIST test accuracyFixed dynamics + external readout
5 pathsCPU · Metal · CUDA · WASM · MPIExplicitly bounded support

Research map

Three directions.
Connected through NeuroAI.

Read the Lab research programme for the questions, methods and collaboration opportunities behind these directions.

01

Foundation available · Brian2 Atlas

Neural Simulation

How can we simulate neural systems faithfully, at scale, across computing architectures?

02

Research agenda

Neural Circuits & Motifs

What are the smallest functional circuits that can learn, adapt and support evolution?

03

Research agenda

Memory Mechanisms

How can neural systems store, retrieve and update useful information over time?

Evidence ledger

What we know.
What remains open.

Every public claim should carry its model identity, code version, environment, result and boundary.

ProgrammeCurrent evidenceBoundary / next test
Engineering verifiedRuntime substrate

A shared model identity reaches native Rust, GPU, browser and MPI execution paths.

Broader model compatibility and independent replication remain open.

Boundary explicitMulti-area model

Full-scale 4.13M-neuron / 24.1B-synapse bounded engineering runs have completed.

Scientific equivalence, repeatability and speed / cost advantage are not yet closed.

Scientific evidenceFlyWire vision

A fixed real-connectome network carries task information readable by an external classifier.

Matched random rewiring, multiple full seeds and internal plasticity are still needed.

Boundary foundApproximate dynamics

Richer proxies and selective retention expose a measurable fidelity–cost frontier.

The E1 plateau and 393 Hz artefact motivate stateful proxies and automatic motif selection.

Long-term blueprint

The roadmap is a sequence of testable questions—not a promise of inevitable scale.

  1. 01
    Verified execution substrateModel identity · portable plans · bounded runtimes
  2. 02
    Large-scale explicit modelsIndependent references · multiple seeds · cost accounting
  3. 03
    Selective structureProcedural connectivity · motif retention · fidelity frontier
  4. 04
    Stateful local proxiesShort memory · adaptive dynamics · boundary recovery
  5. 05
    Cross-scale memoryPersistent state · route-based reactivation · continual adaptation
Research principleEngineering results and scientific interpretation stay separate.

A faster runtime is an engineering result. A model preserving the same scientific meaning is a separate claim—and needs its own evidence.

Follow the work

Methods, measurements and limitations stay together.

Research notesRead the methods and findings.ExploreDiscover interactive neural systems.