iq19zero
Show HN: Invariant – pre-execution control layer for agentic workflows
Show HN: Physical constants from 2 integers – MIT, 1225 tests, falsifiable
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/