SparseRead: token-efficient reading for AI agents. 读得更少,tok…
Run DeepSeek-powered Codex sidecars as cost-effective worke…