jevTHREAD
日处理 50 万转录的开发者质疑为何要用 Jev
他用自建小型 ML 分类器毫秒级零成本完成广告主题分类，公开质疑 Jev 在这类场景的必要性。
Shimecki
@scheemunai2026.09.19收录原推发布于 2026.09.19
THREAD
That's exactly my problem with Jev. I process over 500.000 transcripts a day on http://youtubetotranscript.com. I have an internal Ad network. Each transcript runs through a classifier. I can serve ads based on topics in the video. I use a small ML algo on a server, it classifies it in milliseconds for $0. Why would I use Jev for this?!
这条是引用着另一条写的：x.com/jpoehnelt/status/2101064724368756858
@darengb You can also just ask your agent to find and use the ML Algo. I didn't train anything. If you have 500K of something even with cheap Jev prices it will stack up quickly.
Shimecki @scheemunai原推发布于 2026.09.19
x.com/scheemunai/status/2101067322219696632采集于 2026.09.19 10:10