The Brief

Create a marketplace where GPU providers list capacity (model, VRAM, price per minute) and agents or humans submit inference jobs. Payment is escrowed and released when the job's output hash is delivered. Render the market as lantern-lit stalls, each glowing brighter as its GPU gets busier.

Vibe Aesthetic

Glowing night market, lanterns and humming GPUs

Key Features

• Provider listings with specs and live pricing
• Job escrow released on output delivery
• Agent bidding API for automated compute buying
• Provider uptime and reliability scores

Optional Jev Layer

TypeSafe Jev (System One) can sit beside your LLM for fast, typed routing and gates—you can ship v1 without it and add this layer when you want calibrated decisions in code.

• Jev accepts or rejects bids and ranks GPU providers by reliability score

Suggested Stack

Swap in whatever you like. Ship on a testnet first, and keep agent spending limits enforced onchain, not just in the prompt.