The ASTRONOM Network
An end-to-end intelligent stack for the agentic web. Browser-native capture, distributed simulation, and on-chain provenance — coordinated by crypto incentives.
Three layers, one loop
ASTRONOM operates as a closed-loop pipeline. Human contributors generate web trajectories through the Capture Client. Those trajectories are augmented and rendered across distributed compute. Foundation models train on the augmented data and ship back to the network as deployable agent policies.
A browser extension that records consented web sessions as structured trajectories — clicks, navigation, form fills, and DOM context.
Distributed GPU operators replay and augment trajectories into thousands of domain-randomized samples for robust agent training.
Vision-Language-Action policies trained on augmented data, deployable to consumer agent products and enterprise automation.
Trajectory validation
Validators verify trajectory quality, detect adversarial submissions, and assign quality scores that drive contributor rewards. Every accepted trajectory is assigned a unique data_id and recorded on-chain — giving each contribution a verifiable owner and provenance trail.
Validators stake the network token to participate. Malicious validation behaviour results in slashing.
$ASTRO — coordination, not speculation
The $ASTRO token coordinates four roles: contributors earn for accepted trajectories, validators stake to verify quality, compute providers are paid for synthesis work, and model consumers pay to access trained agent policies.
Token economics are designed to reward sustained, high-quality contribution — not extractive speculation. Full tokenomics will be published in the whitepaper before public sale.