Technodruid / Liminaut

Where the
threshold
fires.

Analog neurons built from salvaged and surplus parts, figured out in real time. Documented openly from the bench.

Neurons built from
discrete components.

Liminaut is a hobby-scale attempt to build a neuromorphic computing system from scratch, analog neurons made from salvaged and surplus components, built on recycled boards, figured out in real time.

The goal is something that can actually learn, using DIY memristive synapses and spike-timing-dependent plasticity built by hand. It's a slow project, openly documented, with no guarantee it works out the way I think it will.

  • Neuron modelAnalog LIF
  • ComparatorLM393
  • Op-ampMCP6002 / OPA2340
  • ScaleOpen-ended
  • Synapse pathMulti-substrate
  • Learning ruleDIY STDP
  • StatusActive build

Latest commits.

// fetching activity...

Collaboration, sponsorship, component samples, technical questions, or just curiosity, reach out. The work is open and so is the conversation.