Hacker News Reader

Comment by sva_ | Hacker News Reader

Does someone have examples of interesting stuff that has been built utilizing Jev/decision models? The way this is hyped up surely there must be some good stuff?

View on HN

On the tiny, personal, end of things, I’m classifying all my incoming email into folders (‘Notification’, ‘Advertising’, ‘Bills’, ‘Personal’, ‘Mailing List’, etc).

Jev makes it easy and cheap to do if you have an email host that provides a way to write arbitrary filter hooks. It’s fast enough that the filter has never timed out despite the network call, and it’s _way_ more accurate/reliable than my previous rules-based system.

I’ve been thinking of using an LLM to do this for a while, but it felt like that would be too complex and perhaps dangerous. With Jev I don’t need to worry about code complexity, prompt injection - or cost, really!.

You're sharing all your email content with jev/typesafe ai?

I’m going through Vercel, which claims to allow Zero Data Retention with Typesafe (https://vercel.com/changelog/typesafe-ai-jev-now-available-o...).

I trust them as much as I trust my email provider, I guess.

I am working on a game https://imperiaquiz.com/en that needs over 20,000 questions. I use jev to classify 'has statement' and 'can be a question' paragraphs/chunks taken from cli script to reduce token usage. Meaning, an LLM only starts work once I've chunked text and marked it as 'to be reviewed' instead of parsing the full content. This reduces tokens usage at least 10x on average

This started as a fun experiment, but I now use Jev on my mac as an advanced auto-correct. Much better than any other option I tried. https://levmiseri.com/nospace (the no space being more of a gimmick, but the autocorrection is good)

I see Jev as a major step towards commoditizing current LLM paradigm. One thing would be to further optimize this particular route to work purely on CPU. This will grant an option to embed this feature into any application, from MS Office to games. The other is integrating this into agent workflow to vastly minimize token consumption.

So, what exactly is the

> interesting stuff that has been built utilizing Jev/decision models

in that case?

It would be interesting to see if a coding LLM trained for tool calling like GLM 5.3 would work well as a Jev model, or maybe even a flow where a model generates options (eg for a plan) and then uses a Jev to refine/optimize the path. Or similarly where else in a harness they’d help.

Not with Jev, but you can use classifiers for a lot of use cases, here’s a few: https://playground.jeffyclassify.com/

Not in a serious manner but I created a testing harness for a Nintendo 3DS game I'm making that uses the OpenAI Decisions API. The main advantage is the speed (~2-300ms per input) which I really need for this purpose.

Also like to see a "layman harness" like LM Studio integrate an open decision model in its workflow.

use it to classify a bunch of research papers (by feeding it section by section, or summaries of sections if too long)

works like a charm

(not a product, so not much to share)

I use Jev in a Claude code hooks to detect dangerous commands.

Reply on news.ycombinator.com