.@VitalikButerin says local AI is getting fast enough to handle much more work without sending your full personal context to the cloud. His latest tests showed Qwen 3.8 Flash running at up to 373 tokens per second on a laptop. But for crypto wallets, he argues speed and privacy are not enough: transaction authority still needs deterministic limits and human approval. https://cryptoslate.com/ethereum-cofounder-vitalik-says-local-ai-is-nearly-ready-but-crypto-wallets-still-cannot-trust-it-alone/