用 Jev 给 8 万封邮件分类再交给 LLM 整理
作者主张 Jev 是分类器不是 LLM，输出概率、score、legend 和置信度；他用它跑完 8 万封邮件，再让 LLM 按结果排序清理，效率提升 100 倍。
Unpopular take: it's a skill issue. Most people posting are vibe coders who can't tell the difference from an autoregressive large language model and a cat. JEV is neither, is a classifier, and should be treated as such. It's built to give you probabilities, score, legend and confidence. I had JEV classify all my emails (+80k) and now I'm using that data together with my LLM to make the sorting, and cleaning 100x more efficient. I do agree with Ole to some degree, that there are a lot of grifters, who seemingly just jump on the next hype train, shill whatever they think they have solved. But loads of the examples that have been floating around have extremely interesting, and when other try to replicate or use, they have no clue HOW to use JEV. The all show up thinking its an LLM, the expect it to work as an LLM because thats what they have gotten used to. So my take is that JEV helps sifting out the grifters from the actual builders. Below I'll add some actual real good examples of JEV in action. Most of it, above my paygrade.
这条是引用着另一条写的：x.com/itsolelehmann/status/2102318027232743779