MIT 量化教授用 Jev 过滤 120 亿金融事件
Jev 实时筛出可疑事件，只把需要深挖的案子交给 Opus 5.5 调查，绝大多数事件不进入昂贵推理层。
an MIT quant professor built a Jev + Opus 5.5 system that can monitor 12 billion financial events every 13 min the system watches incoming data in real time, while Jev decides which events are suspicious and Opus 5.5 investigates the cases that actually need deeper reasoning. Jev filters billions of events and decides which ones deserve attention. Opus then examines the selected cases, connects the relevant data, and explains what might be happening. most events never reach the expensive reasoning layer. I found a breakdown of the pipeline they used, and the way Jev reduces the workload for Opus is surprisingly clever. going to post how the system works next, so stay tuned and send it to someone building AI infrastructure. would you trust an AI to decide which transactions deserve a human investigation?
这条是引用着另一条写的：x.com/me_barnyx/status/2101976764779999381