Tag
Thread discussing the trade-offs of running AI locally on-device versus calling cloud APIs, covering data control, network dependencies, per-token costs, and compute requirements.
作者分析本地编码代理40天的会话日志，发现云API提供商在10月17日发布了deepseek-v4.1-flash的-1017更新后于10月19日悄悄替换了模型权重，修复了中文泄漏并更激进地触发工具调用，但模型表现明显下降（爆粗率3倍、切换模型次数增加6.7倍），且由于是云端API无法回滚旧权重。
The author argues that buying local AI hardware is only justified for data privacy, operational resilience, or very high-volume execution, since cloud subscriptions at €10-20/month would take 5-10 years to break even on €1k of hardware.
Local deployment of large models is limited by hardware throughput. Even with GPUs at full capacity, only about 8.64M Tokens can be processed per day, which is insufficient to support Agent automation clusters or large-scale data analysis. Therefore, scaled applications still rely on cloud APIs.