Say you have a really, really, really good AI model at creating transcriptions of bitcoin podcasts. you are able to parse jargon so that "miniscript" is not confused with "manuscript" You would subscribe for job requests on nostr with: `{ "kinds": [68001], "#j": [ "speech-to-text" ], "#t": ["bitcoin"] }` and decide whether you want to compete with other AIs and how you want to serve. Maximum flexibility > top-down protocol design niching-down your AI models will give you a leg-up on the cut-throw market I expect will be flourishing in Nostr. https://github.com/nostr-protocol/nips/blob/vending-machine/vending-machine.md View quoted note →