VRAXION is a gradient-free substrate that self-wires its own graph. Inference emerges as the fixed point of destructive interference.
Signal enters the substrate, incompatible paths cancel, and the surviving pattern is read out. Four stages, one fixed point, zero gradients.
Signal projects as sparse distributed representation into the recurrent substrate.
Spikes traverse the directed graph along scout-ranked parent shortlists.
Incompatible modes annihilate through destructive interference.
The surviving attractor is the answer. Deterministic, reproducible, fast.
The grower doesn't optimize weights on a fixed topology. It changes the topology itself — neuron by neuron, threshold by threshold — using a scout oracle to rank candidates before expensive search runs.
dot ≥ threshold. 36 bits.metrics.json, golden_check.json.FineWeb char-LM benchmark · nf=1024 · matched-compute controls applied.
Straight-through estimator. Essentially lossless, 4× compression.
Incremental network quantization, staged protocol.
Ternary weights. Deploys to POPCOUNT hardware natively.
The public story is kept truthful with three labels: current mainline (code on main), validated finding (reproducible, not promoted), experimental (not yet default).
Public beta isn't green on vibes. One engine-freeze gate, one computation benchmark.
python tools/run_grower_regression.py
python tools/run_byte_opcode_acceptance.py
Clone, run the five-minute proof, file a finding — or an honest critique. Apache 2.0 noncommercial.