Jev is just a classifier. It enables a software engineer to create and use a bert-style classifier within minutes, replacing what use to take a specialized ML team often months to do (create a dataset, train a model, serve it themselves).

The thread explains Jev as a BERT-style classifier that engineers can build and deploy in minutes, with more generality, less data drift, prompt-based updates, and automatic future improvements.

Jev is just a classifier. It enables a software engineer to create and use a bert-style classifier within minutes, replacing what use to take a specialized ML team often months to do (create a dataset, train a model, serve it themselves).

- The Jev classifier is also a lot more general, so it doesn't experience the same issues with data drift that narrowly trained models do. - Updating a Jev classifier requires a bit of prompt engineering instead of improving the dataset and retraining a model. - The next release of Jev will probably just make the classifier smarter without any work on your port - Holy shit what's gonna happen to what software can accomplish when you you can embed general classifiers all over the place?
Engagement counts are snapshots taken when collected.
Only posts collected on this site, sorted by likes.
Jev Score · Probabilities as 100 votes
| Rating | Virtual votes | Subtotal |
|---|---|---|
| 100Save-worthy | 0 | 0 |
| 75Specifics | 14 | 1,050 |
| 50Opinion | 77 | 3,850 |
| 25Reaction | 9 | 225 |
| 0Noise | 0 | 0 |
20% filter threshold · Separate from quality