15min typesafe Jeveloper tutorial 🧠 My team at Treg has been using Jev a lot in prod: - Classify web traffic - Smart onboarding - Jev-reranked search ~10x cheaper, 18x faster, 30% more accurate in our tests compared with GPT6-luna OpenSourced our Jev implementation 👇

@typesafe Jev & Treg are good friends

@typesafe @treg_ai So tired of 30-min onboarding forms can’t wait for every site to use Jev and just know what it needs about me!!

@typesafe @treg_ai yeah, I’m sure everyone will move to this soon but the privacy question here is pretty serious

@typesafe @treg_ai facts!

@treg_ai oppus, tagged wrong @typesafeai above lol

@typesafe @treg_ai Thanks for the guide, chief!

@typesafe @treg_ai I15 mins that can help you understand ai better than 90% of people

@typesafe @treg_ai

@typesafe @treg_ai Fr treg_ai & Jev we needed! Cooking treg_ai team ❤️

@typesafe @treg_ai open to feedback!!

@typesafe @treg_ai Jev + treg_ai most needed combo

@typesafe @treg_ai this is now mindblowing ? i can predict what customer is going to leave ?

@typesafe @treg_ai this is gold

@typesafe @treg_ai 🙌

@typesafe @treg_ai Classification, onboarding, and reranking are exactly where a decision model fits better than a general LLM. Which of the three gave you the biggest win in production?

@typesafe @treg_ai 18x faster and 10x cheaper is a pretty wild combo. Curious to see what people build with Jev + Treg.

@typesafe @treg_ai cool stuff!

@typesafe @treg_ai This is exactly why specialised models matter. Real production use is what makes this stand out.

@typesafe @treg_ai Jev for reranking actually makes a ton of sense. Love the whole concept! Great work :)

@typesafe @treg_ai The Jev + Treg split makes a lot of sense. One handles the decision, the other gives it fresh data. I’d probably start with reranking with enrichment workflows first.

@typesafe @treg_ai I really like Treg and it just keeps getting better 🙌

@typesafe @treg_ai 😎

@typesafe @treg_ai This is prob the best Jev tutorial I've seeen, thanks for sharing!

@typesafe @treg_ai this is impressive! 10x cheaper, 18x faster, and more accurate is a crazy combination

@typesafe @treg_ai sick gide!!

@typesafe @treg_ai jev + treg is an insane combo 🫡

@typesafe @treg_ai super cool🌟

@typesafe @treg_ai this is super cool. Will try it soon :)

@typesafe @treg_ai 18x faster, so the tutorial is really 50 seconds

@typesafe @treg_ai 18x faster is the number that caught my attention.

@typesafe @treg_ai check out the all vid! shared a ton of sick examples there

@typesafe @treg_ai definitely

@typesafe @treg_ai For traffic classification, what happens when the result is uncertain? I'd want an explicit fallback rather than forcing a label.

@typesafe @treg_ai Yep wouldn’t force a label here Jev returns probabilities, so if the top score is too low or two sources are too close, I’d mark it “unknown” and route it to a fallback rule or review queue

@typesafe @treg_ai That’s the fallback I had in mind. I’d also log why it went to review, so I could separate low scores from close calls.

@typesafe @treg_ai embedding Jev in app code lets one primitive serve several workflows

@typesafe @treg_ai damn, so much cheaper

@typesafe @treg_ai i built a custom skill with treg x jev to analyze chatgpt responses for my app and tell me how to rank better🔥

@typesafe @treg_ai 10x cheaper and 18x faster is wild. been building RAG with citations, does jev hold up as a reranker vs cohere?

@typesafe @treg_ai Amazing

@typesafe @treg_ai This is seriously impressive. The cost + speed gains are hard to ignore !!

@typesafe @treg_ai 10x cheaper and 18x faster is insane

@typesafe @treg_ai Three production uses on Jev with the implementation open sourced is something I can actually learn from. At about 10x cheaper and 18x faster than GPT6-luna in your tests, classification jobs I would usually rule out on cost start looking possible. Excited to dig in.

@typesafe @treg_ai 10x cheaper and faster makes the tradeoffs pretty interesting 👀

@typesafe @treg_ai This is an insane combo 🔥