Cloud-based LLM gold rush is ending (automato.substack.com) AI

The article argues that the era of cloud-based LLMs for everyday tasks is declining, citing Apple’s WWDC direction toward running AI workflows locally on devices and using the cloud only when needed. It claims that LLM subscriptions and agentic systems face a “wall” due to escalating access costs and the real expense of validation, maintenance, and human oversight, while LLMs remain best as tools that amplify human work rather than provide deterministic execution. The author frames this shift as a move toward “tech sovereignty” and away from AI “arms-race” narratives, suggesting value will come more from practical local features than frontier benchmarks.