Turn your devices into a stake in the network
Lend spare compute, inference, personalization, or bandwidth from the hardware you already own, and earn ORB for verified work. Value comes before emissions, so rewards track real demand.
Every device earns for what it can actually do
Phones do not train models, and we do not pretend otherwise. Each tier is matched to work it can genuinely perform.
Serve private inference and contribute data. Targeted at the on device NPU and built ready for neuromorphic hardware. Runs while charging.
Heavier inference, on device personalization, and experimental training. Apple Silicon is especially capable here.
The real training tier. Stake ORB, pass hardware attestation, and take on distributed training work for the strongest rewards.
On the roadmap. Ultra low power always on inference through custom inference silicon, for devices that must sip milliwatts.
From install to claim
Add Hylon to your OrbVPN app, or install the desktop node on Windows, macOS, or Linux.
The runtime profiles your device: compute class, memory, bandwidth, and thermal headroom, and assigns suitable work.
Your device serves inference, personalizes, or trains. Every contribution is measured and verified, never paid for idle uptime.
You accrue rewards for verified work. Claim to a non custodial wallet when you choose. Rewards are backed by real demand, not inflation.
Rewards that do not death spiral
A decade of network economics taught a hard lesson: paying supply before demand exists collapses the token. Hylon rewards verified work, funds buybacks from genuine service revenue, and lets scarcity track demonstrated demand rather than headcount. You earn because the network sells real output, not because it prints more.
Wallets are non custodial. Rewards accrue for contribution and are claimed on your terms. Hylon is compute and data for AI, described plainly, never marketed as mining.
Join the network at launch
The node client ships with Gate 0. Read the token design, or dive into the whitepaper for the full mechanics.