We're building a new kind of computer to beat GPUs on optimization

GPUs have taken us far, but they’re still not good at a core class of problems: reasoning, association, and optimization. These show up everywhere — multi-object tracking, data association, graph labeling, scheduling, Max-Cut — and they don’t parallelize well on GPU architectures.

We’re working on a different approach: a Resonant Phase Computer. Instead of crunching digital instructions, it uses coupled oscillators that naturally settle into low-energy states of the problem you encode. You load the problem into physics, let it evolve, and read out the synchronized phases as solutions.

Where we’re at today: • We’ve simulated systems from 64–512 nodes, showing 2–3× lower latency and energy compared to GPU baselines, with comparable or better solution quality. • We’ve released methods and CSVs publicly for transparency (Benchmarks page). • We’re now taking pre-orders to fund and deliver the first DevKits (resonantcomputer.com): small USB-C/Thunderbolt boards that researchers and developers can plug into a laptop or workstation.

Why this matters: • For AI/ML, it accelerates the reasoning layer GPUs can’t. • For robotics, it could cut decision latency in half, improving safety and responsiveness. • For optimization at scale, it’s a room-temperature alternative to quantum annealing.

We haven’t built hardware yet — this is about proving the architecture and getting early users on board. We’re looking for feedback from the HN community: skepticism, prior art, interest, or partners who want to experiment once hardware ships.

Benchmarks: https://www.resonantcomputer.com/benchmarks Pre-orders: https://www.resonantcomputer.com/