Policy is a promise. Architecture is a guarantee.

Great point. 箴言集条目++

现在 Taalas 可以直接把模型写进硬件里达到 Llama 8B 17k token/s，以后本地运行一些智能助手我并不觉得会是问题

随着 OpenClaw 之类的 agent 能处理生活的越来越多方面，隐私问题也不得不提上日程了，就算 Anthropic 可以有 ZDR，但那是 policy，是 promise；本地模型的安全是 architecture，是 guarantee

https://juno-labs.com/blogs/every-company-building-your-ai-assistant-is-an-ad-company
Juno-Labs
Every Company Building Your AI Assistant Is Now an Ad Company
Every company building an AI assistant is funded by advertising. They are also building hardware that listens and watches 24/7. Local on-device inference is the only way out.
 
 
Back to Top