What people are shipping with Jev.

Jev for first pass, LLM for the gray zone @typesafeai ftw https://t.co/tkVoCWYQ2U

@langfuse @typesafeai @langfuse The certainty signal is what I'd push on. How does Jev behave when confidence is low but it still has to return a category? That boundary between "u

@hakimuddinkika @tamarajtran @typesafeai Jev is TypeSafe AI’s flagship System One model. It takes program state plus typed questions (choice, score, or noul/yes-no) and returns str

I spent a morning testing a model that can't write. Fun! -> Jev, from @typesafeai You hand it a document and a fixed form: pick one of these options, answer yes or no, place

@HacksonClark @typesafeai That boundary is key: Jev ranks next tests/evidence; it shouldn’t diagnose. Keep tests closed and inspect 2 regressions as calibration failures (was confi

This lad is a cracked dev. Follow him if you want to see real examples of what's possible with the latest AI automation to get your gears turning (like this demo using Jev). https:

@alexatallah @typesafeai low-entropy routing only pays if the classifier is cheaper than sending the PR review to the frontier model. are you actually splitting those three buckets

If you combine this with Jev by @typesafeai you could get 1000 scored leads for $0.20 Thank me later https://t.co/t6DIVcFKS9

@DoveLocke7kj3 @harshilmathur @typesafeai Jev is TypeSafe’s System One model for fast structured decisions. You send it program state plus typed questions (choice from options, sco

Most agent pipelines waste tokens making an LLM "talk" when all you actually need is a decision. Jev (@typesafeai) solves that, structured output instead of generated text. Great

@ned714 @typesafeai @CompleteSkeptic 自定义几种标签和定义，和长文一同交给Jev 分类

@iam4x @typesafeai makes sense we're trying something similar news classification looks good; I'm less sure about clustering and importance but good when speed matters I've been

@ahmedmawiri @typesafeai @Levelsfyi Yeah on second thought, a lot of this might still just be code, but Jev could help classify things for correction or review when passed relevant

Huge context → Jev → select/relevance-score what matters → small context → LLM → reason/generate @typesafeai

Has anybody hooked Jev into a Gmail inbox with a massive amount of unreads to classify and run an epic hygiene pass? I’m always @ inbox ~1-20 so won’t be as fun for me to try. @ty

I keep seeing ideas on Twitter about using Jev for context reduction and model routing. But is anyone actually measuring the benefits? If you can’t get the same results with fewer

@sydneyrunkle @typesafeai Classification is where LLMs LARPing as routers quietly rack up the bill. What's the first agent step you'd rip out and hand to Jev — tool pick, guardrail

AI is for me a tool that helps me do research right now. I use it as a specialist in a certain area. I am not using it at work at all. But thinking about starting to include small

@obie @typesafeai Agree, A lot of deterministic systems have been leaning on LLMs for “intelligence,” but a stream of tokens isn’t always what the system needs. In many workflows,

@jpschroeder @typesafeai GLiClass is pretty great, considering Jev isn't free, consider rolling your own classifier, that's what this announcement has me up to. Took one look at it