APUS AI于10月13日开源Personal Agent产品Reflex Agent，宣称响应速度比Claude Code快3倍，支持真正无限上下文。
Reflex Agent’s release is notable not just for its claimed 3x speed advantage over Claude Code, but for its explicit focus on ‘true infinite context’ — suggesting architectural innovations beyond standard attention windowing or retrieval-augmented summarization. This addresses a core pain point for production agents: maintaining coherent, stateful reasoning across long-running, multi-step user interactions without latency spikes or context drift. For early AI application teams, it signals that performant, open, and embeddable agent runtimes are now viable alternatives to building custom orchestration layers from scratch. Teams building vertical agents — especially in regulated or document-heavy domains like legal ops or insurance underwriting — can leverage Reflex as a battle-tested foundation, accelerating time-to-PoC while retaining full control over data flow and memory handling.