Vitalik Buterin, co-founder of Ethereum, says local AI on laptops has reached a practical turning point. On his Strix Halo laptop, the Qwen 3.8 Flash model handles a large share of AI tasks locally. Local inference protects privacy by sending only selected queries to remote models, while input-processing rates range from 109.82 to 373.22 tokens per second. Wallet control of crypto assets still requires fixed policies, transaction assertions and human approval beyond model judgment. The Ethereum ecosystem is funding Steward, a fully local macOS smart-account wallet with on-device AI assistant, and the draft EIP-7906 standard for post-transaction assertion frames.
Source: Read the original article
