Next Brain · Research notes
Evidence, engineering, and the boundaries between.
Field notes on preserving neural models, testing the execution stack and learning from results that complicate the headline.
Introducing Brian2 Atlas: an open engine for neural simulation and training
Why we built Atlas, how a shared execution architecture connects Brian2 models to different runtimes, and what the published experiments show.
By RockTrain Your First Spiking Neural Network with Next Brain
Start with a handwritten digit, follow how a small network learns from spikes, and use Training Studio to inspect, evaluate and reuse the result.
By RockWe Re-ran a Published NMDA Model—Without Changing the Science
What happened when a 755-million-synapse Brian2 workload met a new Rust, GPU and MPI execution stack.
By RockWe Put MNIST Through a Fly Connectome. It Reached 87.14%—and That Wasn't the Most Important Result
A whole-brain FlyWire simulation preserved enough information to classify handwritten digits, but the controls changed what the score actually means.
By Rock