Skip to main content
Author a model. Compile to a sealed .thx artifact. Run on the behavioral simulator or neuromorphic hardware — same command, same transcript, same conformance contract.

Deterministic

Seeded, reproducible, CRC32-sealed. The artifact is the spec and the implementation in one file.

Energy-proportional

Energy scales with spike count, not model size. Silent inputs cost nothing.

Hardware-portable

Compile once. Validate on the simulator. Deploy the same artifact to neuromorphic silicon.

Author, compile, run

Built for systems where physics decides the stack

Robotics

Sub-millisecond sensor-to-actuator loops. Event-driven — quiet when the arm is still.

Drones & UAVs

Energy scales with information rate. Onboard inference without GPU-class power.

Aerospace & Space

Finite power for years. No cloud. Verifiable before launch.

Defense

Reproducible builds, offline verification, evidence packages auditors can review.

THRINDEX Certified

Hardware backends earn the badge by passing the conformance harness against CONFORMANCE_ENVELOPE_V0 — a derived bound on divergence from the float reference simulator. Read the conformance spec →

What THRINDEX is not

  • Not a general-purpose framework. Spiking neural networks on neuromorphic hardware — specifically.
  • Not a black box. Every run prints a structured transcript: spike counts, syn-ops, energy estimate.
  • Not an abstraction over hardware. A deterministic model of what happens on the target, with a conformance bound on divergence.

Get started

Install THRINDEX and run your first model in under five minutes.

Learn about SNNs

Spiking neural networks from first principles — biology, mathematics, and code.