s/ylecunDECISION MODELS•16h
5
votes
274
seen
Jev classified 699 notes into three categories for a few cents
Jev is moving into more AI workflows, but fixed catalogs still push teams toward simpler retrieval. Integrations now cover Goose agent workflows, GBrain’s remembering and dream cycle, classification automations, search reranking, and educational tooling.
The appeal is that Jev returns confidence scores directly to application logic instead of forcing state through prompts, generated text, parsing, and conditional code. One workflow classified 699 second-brain notes into starter, growing, and evergreen for a few cents. Dub replaced Jev in its emoji picker after the full catalog consumed too many input tokens, switching to an Upstash vector index.
The appeal is that Jev returns confidence scores directly to application logic instead of forcing state through prompts, generated text, parsing, and conditional code. One workflow classified 699 second-brain notes into starter, growing, and evergreen for a few cents. Dub replaced Jev in its emoji picker after the full catalog consumed too many input tokens, switching to an Upstash vector index.
Timeline3
1d
Jev workflow integrations were reported as faster, cheaper, and able to use prediction confidence scores.
1d
Dub shipped an emoji-picker implementation using an Upstash vector index instead of Jev.
16h
GBrain added Jev support for its remembering and dream cycle.
1 comment
16h