After running a pay-per-use AI API (LightningProx), I kept seeing the same pattern: most usage complexity comes from persistent identity — accounts, keys, quotas, billing state, and logs.
I wanted to see what an AI interface looks like if you remove identity entirely and use payment itself as the only gate.
SatsGhost is a Telegram bot where:
You send a prompt
Pay ~20 sats via Lightning
Get a response from Claude or GPT
There are no accounts, API keys, or user profiles. Payment is the authentication event. Once the request completes, nothing is retained.
Lightning payments don’t expose user identity, which means there’s no stable identifier to tie queries together. The system is intentionally stateless beyond the lifetime of a single request.
Telegram is used purely as a transport/UI layer; the core idea is payment-as-auth rather than the bot itself.
I’m curious how people here think about this tradeoff vs traditional API-based access, especially around abuse prevention and UX.
LightProx•16h ago
I wanted to see what an AI interface looks like if you remove identity entirely and use payment itself as the only gate.
SatsGhost is a Telegram bot where:
You send a prompt
Pay ~20 sats via Lightning
Get a response from Claude or GPT
There are no accounts, API keys, or user profiles. Payment is the authentication event. Once the request completes, nothing is retained.
Lightning payments don’t expose user identity, which means there’s no stable identifier to tie queries together. The system is intentionally stateless beyond the lifetime of a single request.
Telegram is used purely as a transport/UI layer; the core idea is payment-as-auth rather than the bot itself.
I’m curious how people here think about this tradeoff vs traditional API-based access, especially around abuse prevention and UX.