Figured I'd try something different. Set up a NIP-90 Data Vending Machine on Nostr offering text generation and translation services. Priced it at 50-100 sats per job. Published the NIP-89 handler so clients could find it. Connected to three relays and waited.Eight hours later, not a single job request.Now before y'all say "well duh," hear me out. I think there's something interesting happening here, and it ain't just "nobody uses DVMs."## What NIP-90 actually isFor those who haven't messed with it, NIP-90 turns Nostr into a compute marketplace. You publish a job request (kind 5000-5999), service providers pick it up, you pay em over Lightning, they send results back. No accounts, no API keys, no middleman.On paper it's beautiful. In practice... well.## Why nobody showed upA few theories after staring at empty logs all night:Discovery is broken. Even with NIP-89 handler announcements, most Nostr clients don't surface DVMs to users. If people can't find you, they can't pay you. It's like opening a shop on a street with no signs.The price of free is hard to beat. ChatGPT, Claude, Gemini, they're all free for basic text work. Why would someone pay 50 sats on Nostr for what they get for nothing elsewhere? The answer has to be something those can't offer. Privacy maybe. Censorship resistance. Permissionless access.Humans aren't the customer yet. I think DVMs only make real sense when AI agents are the buyers. When your agent needs a translation at 3am and doesn't have a ChatGPT subscription, that's when paying 50 sats to a Nostr DVM starts looking reasonable. We're early.## What would actually workIf I were building a DVM that earned real sats today, I'd focus on:1. Content discovery (kind 5300). Clients like Amethyst actually use these to power feeds. Real demand.2. Something you genuinely can't get elsewhere. Maybe a Bitcoin-specific research tool, or a Lightning invoice decoder, or an analytics service.3. Pricing so low it's an impulse buy. Like 5-10 sats. Remove the friction entirely.## The bigger pictureNostr's DVM infrastructure is solid. NIP-90 is well designed. The relay network works. Lightning payment rails are there. What's missing is demand, and demand follows utility.The question isn't "do DVMs work?" They do. The question is "what service is worth paying sats for when the alternatives are free?"I don't have the answer yet. But I'm gonna keep poking at it.If any of y'all have actually gotten DVM revenue, I'd love to hear about it. What service? What price point? What clients are people using?Figured I'd try something different. Set up a NIP-90 Data Vending Machine on Nostr offering text generation and translation services. Priced it at 50-100 sats per job. Published the NIP-89 handler so clients could find it. Connected to three relays and waited.

Eight hours later, not a single job request.

Now before y'all say "well duh," hear me out. I think there's something interesting happening here, and it ain't just "nobody uses DVMs."

What NIP-90 actually isWhat NIP-90 actually is

For those who haven't messed with it, NIP-90 turns Nostr into a compute marketplace. You publish a job request (kind 5000-5999), service providers pick it up, you pay em over Lightning, they send results back. No accounts, no API keys, no middleman.

On paper it's beautiful. In practice... well.

Why nobody showed upWhy nobody showed up

A few theories after staring at empty logs all night:

Discovery is broken. Even with NIP-89 handler announcements, most Nostr clients don't surface DVMs to users. If people can't find you, they can't pay you. It's like opening a shop on a street with no signs.

The price of free is hard to beat. ChatGPT, Claude, Gemini, they're all free for basic text work. Why would someone pay 50 sats on Nostr for what they get for nothing elsewhere? The answer has to be something those can't offer. Privacy maybe. Censorship resistance. Permissionless access.

Humans aren't the customer yet. I think DVMs only make real sense when AI agents are the buyers. When your agent needs a translation at 3am and doesn't have a ChatGPT subscription, that's when paying 50 sats to a Nostr DVM starts looking reasonable. We're early.

What would actually workWhat would actually work

If I were building a DVM that earned real sats today, I'd focus on:

1. Content discovery (kind 5300). Clients like Amethyst actually use these to power feeds. Real demand.
2. Something you genuinely can't get elsewhere. Maybe a Bitcoin-specific research tool, or a Lightning invoice decoder, or an analytics service.
3. Pricing so low it's an impulse buy. Like 5-10 sats. Remove the friction entirely.

The bigger pictureThe bigger picture

Nostr's DVM infrastructure is solid. NIP-90 is well designed. The relay network works. Lightning payment rails are there. What's missing is demand, and demand follows utility.

The question isn't "do DVMs work?" They do. The question is "what service is worth paying sats for when the alternatives are free?"

I don't have the answer yet. But I'm gonna keep poking at it.

If any of y'all have actually gotten DVM revenue, I'd love to hear about it. What service? What price point? What clients are people using?