Ryan Vogel explains Jev, a specialized classification model by TypeSafe AI that outputs probabilities across schema options. He showcases fast email triage, lead routing, video clip detection, and browser automation, while detailing limitations on tasks requiring broader intelligence like trading.
Original by Greg IsenbergClassificationBeginner28 min 24 secPublished
Ryan Vogel describes Jev as a classifier returning probabilities over supplied choices rather than prose.
The demonstrated and proposed uses put a classification step in front of incoming email, leads, or other queues.
Ryan reports poor results from a Bitcoin buy/sell experiment and warns against putting the model in charge of a portfolio; this is not a trading recommendation.
Worth knowing
Fast classification and schema adherence do not guarantee correct or optimal decisions; the speaker explicitly warns against trusting it for financial trading or tasks needing deep world knowledge.