Neuromorphic networking & tensor-native execution model.

Semantic Reflex, or SEMREF for short, is an all new execution model that upgrades your applications to be tensor-native. Your application doesn’t call functions across process boundaries, it spikes tensors into parallel channels. Using P2P techniques, networked nodes listen for these spikes as tensor shapes. If matched, the SEMREF circuit allows your method to reflex one or more tensors in response to the local tensor "tape", which is then interpreted as a spike for any configured path. Push-only semantics keep designs simple, while intelligence is embedded directly into a hypergraph of connected instances. All proof and observability lives on the tape.

Why it matters

Parallelism, built-in: up to 64 independent tensor channels on any single node.
Neuromorphic execution: each reflex handler is a virtual circuit bound by shape, not endpoint.
Bit-perfect transport: cryptographic proofs are embedded at the wire level (TRP).
Language-native APIs: define schemas once, spike tensors anywhere.

Hello SEMREF

import atc
import peernyx

# Define tensor schema
sfrm = "sfrm[480f32, 640f32, 3i8]"   # sensor frame
ctvc = "ctvc[64f32]"                 # control vector

with atc.Tensor() as t:
    Frame = t.add_axis("SensorFrame", sfrm, 10, 25)
    Ctrl  = t.add_axis("ControlVec", ctvc, 5, 26)

# Reflex handler: runs when tensors of matching shape arrive
@atc.reflex(Ctrl)
def handle_ctrl(cmd):
    print("Reflex:", cmd)

# Stream data in parallel channels
Frame.spike(camera.read())
Ctrl.spike(model.predict(Frame))

Why this is cool

You don’t serialize, marshal, or wrap. You just spike tensors. Peernyx takes care of transport, proof, and multiplexing. SEMREF makes it neuromorphic: every handler is its own circuit, listening by shape, processing, and reflexing back tensors. The only limit is your machine’s resources.

⚡ Upgrade from APIs to circuits.

⚡ From compliance checkboxes to proof on the wire.

⚡ From cloud bottlenecks to neuromorphic networking.

Coming Soon

For inquiries: contact@peernyx.com