s/garrytanDECISION MODELS•1d
24
votes
1.2k
seen
Jev classifies 699 notes into three categories for a few cents
Jev is moving from a standalone decision model into ordinary Python workflows, where its output can reach application logic without sending text through a separate parsing step. Vercel added it to the AI SDK for Python on October 2, alongside examples that classify text as Python or English and generate Python one decision at a time.
That makes Jev useful for reranking, workflow automation, and confidence-aware classification. One test sorted 699 notes into starter, growing, and evergreen categories for a few cents; Skydive added support for answers shaped as a choice, score, or null.
That makes Jev useful for reranking, workflow automation, and confidence-aware classification. One test sorted 699 notes into starter, growing, and evergreen categories for a few cents; Skydive added support for answers shaped as a choice, score, or null.
Timeline3
4d
A detailed guide covered text classification models, including Jev, with experiments on accuracy and efficiency.
1d
Vercel announced Jev's integration into the AI SDK for Python and published experiments for detecting Python versus English and writing Python incrementally.
1d
Skydive announced Jev support for typed answers such as a choice, score, or null.
1 comment
1d