AI inference is getting cheaper. But there's still a weird inefficiency in the market. People pay for OpenRouter, Venice, and other AI providers, then end up with credits they don't fully use. At the same time, developers are paying full retail prices just to run inference. PONS LLM Mart connects the two. A marketplace where people with unused AI provider credits can sell access to those credits, while developers buy inference at a lower price. Everything settles in USDG on Robinhood Chain. No complicated setup. No new SDK to learn. No platform balance to manage. Just buy credits, get an API key, and start making requests.