Vision

Compute is the only scarce thing in this entire market. Everything else, supply, narrative, attention, can be conjured. GPU time cannot. p is built on that single fact: a token should mean a real claim on real hardware, not a story about one.

By bonding each token to a real inference node, we turn a speculative asset into something with a job. A launch that finds real demand for its compute becomes self-sustaining: it earns its place on the grid. A launch that doesn't, fades. We think that is the only honest tokenomics there is.

One node is fragile. A grid is not. By pooling every launch into a shared inference network, demand can be routed to whichever capacity can serve it cheapest, and no single node has to find its entire market alone. Each launch adds throughput the whole grid can draw on; each request served deepens the liquidity of compute underneath every token at once.

We do not pick winners. We provision the hardware and let real inference demand decide where the load goes. Capacity that serves useful work earns the fees to expand. Capacity that sits idle gets reclaimed. The market becomes a live census of which compute is actually doing something: priced, transparent, and impossible to fake with a chart.

The endgame is a decentralized inference layer that anyone can launch into and anyone can buy from, where the price of a token is the price of the work its hardware can do. We are building the market first. The grid follows the demand.