Bytes and Brains

NODE bytesandbrains.ai . CRT/01 . phosphor-green UTC 2026.06.26 . UPLINK OK •

         @@@@
       @@@@#@@
      @@@@#@#@@
     @@@#####@
    @@#####@@@
    @#####@##@@@
    @####@###@@@@
   @@########@@@
    @######@@@@@@
    @#####@@####@
    @@###@######@@
     @@#@#######@@
      @@@##@@@@@@
       @@@@@

BYTES AND BRAINSv0.3.0

A SUBSTRATE FOR NETWORKED MACHINE LEARNING

/// building in public . CRT/01 ///

A framework for machine learning that runs outside the data center. On phones, sensors, hospital networks, on-prem fleets, and peer-to-peer overlays.

Write your workload as a Module. The framework partitions it across nodes, binds your compute and transport, and ships the bytes.

whitepaper research github about

topology

                [ Module ]
                     │  recorded once
                     ▼
                [   IR   ]   ←  ONNX GraphProto
                     │  compiler partitions per peer
                     ▼
            [ per-node programs ]
                     │
       ┌────────┬────┴────┬─────────┐
       ▼        ▼          ▼         ▼
   [ Node ]←→[ Node ]←→[ Node ]←→[ Node ]   peer mesh,
       │         │         │         │            bytes through
       ▼         ▼         ▼         ▼            one envelope
   [adapter pool: libp2p / HTTP / in-process / ...]

runtime data flow: one Module, partitioned IR, per-node programs bound to adapters.

BUILDING IN PUBLIC. The library is on crates.io. The substrate design lives in the whitepaper. Source is on github. Research is on the research page.

If you work on federated learning, gossip protocols, peer-to-peer systems, or related fields, reach out: [email protected].

latest release

v0.3.0

Jun 13, 2026

First public release: framework + sans-IO Engine + Wire + role traits + CpuBackend reference.

read the release notes →

// TIMELINE

What is coming. Two tracks: extensions to the framework itself, and protocols built on its primitives.

framework extensions

near-term

ONNX Runtime backend
ONNX import and interop utilities
CPU SIMD opt-in
benchmark harness
in-memory actor-based simulator

vision

wire-version negotiation
pure Node and Compiler bindings
Python DSL surface
JavaScript and TypeScript DSL surface
libp2p adapter (bb-libp2p)
HTTP adapter (bb-http)
observability story
snapshot and restore
memory-budget tooling
no_std build profile
deployment examples (Linux VM fleets, ROS 2, Android, Jetson, ESP32)

supported protocols