实测 Jev 模拟交易：只会追趋势、持续亏钱
作者看完 Jev 纸面交易 demo 后认为它无法预判、独立交易必亏，因此 Noesis 只把 Jev 用在信号发现与风控，仅极强信号才直接触发投资。
i’ve gone through nearly every practical jev application i could find lately. most of the investment examples are far too idealized. the paper-trading page below is worth watching. https://jev-poly-crypto-demo-black.vercel.app/ jev keeps losing money. i don’t think jev is really making investment decisions there. it is reflecting the current trend. it can follow momentum, but it cannot anticipate what comes next. let it trade independently and losing money becomes almost inevitable. that’s why jev in noesis is mainly being integrated into signal discovery and risk control. there is only one narrow case where it can trigger an investment directly, when the signal is exceptionally strong. the more interesting possibility is that jev represents the beginning of language models diverging by execution role. this isn’t specialization around a vertical corpus. it is specialization around execution capability.
这条是引用着另一条写的：「NOESIS 用 Jev 做社交信息发现层，将放出 demo」