@zauey: Why is everyone obsessed with an AI model that can't write? 🧠 TypeSafe launched Jev, a "System One" model that doesn't write text - it picks from answer options you define and returns a probability for each. ⚡ 20 questions cost about the same as one, because it reads everything in a single pass instead of composing answers word by word. 💸 Input runs $42 per billion tokens (most models price per million) and output is free. 🔮 The real unlock isn't intelligence, it's that decision-making inside software just got cheap enough to use everywhere. #AI #startups #Tech #LLM #devtools Disclosure: Views expressed are my own. This is not investment advice. This is for informational and educational purposes only.
Classifier with general knowledge. Anyone who’s worked with agents - this is a necessary step.
2026-09-22 05:41:45
46
ugofficial :
How can I use jev
2026-09-22 06:40:50
0
gg.deska :
so is it just a classifier?
2026-09-22 09:53:40
5
!truthful :
Most llm providers support structured output, so what’s the point of it besides being extremely fast? Can it classify on tasks that require more domain expertise like security or finance?
2026-09-22 05:54:13
0
Stout :
i love that behavioral psychology is being leveraged. huge fan of Kahneman
2026-09-22 12:29:36
1
andrew k :
I wish it could run locally 😞
2026-09-22 06:56:59
2
Anything :
While looking into it found out that it seems to be just better marketing for essentially something that someone built and open sourced over a year ago. Laya
2026-09-22 08:30:56
9
Neelasha :
Guysss you’re missing the point. Yes it’s a classifier, but what makes it cool is that it’s general purpose. You can use it out of the box for many use cases, whereas you need to train regular classifiers for your use case each time
2026-09-26 13:39:20
3
nickandron :
In simple terms, how is this different than decision trees?
2026-09-22 05:14:36
3
Gareth Clarke :
Can Jev be used to drive decisions within a larger task performed by a different agent? For example could an Astra agent create a Jev sub agent to make a decision within a task?
2026-09-22 16:53:59
0
Derek :
If you set the probability threshold to 85%, it means that there is still an 15% chance it picks the wrong response. That's not a good rate of success. Critical corporate processes need better than 15% success rate.
2026-09-22 14:46:54
6
fiona frills diary :
very interesting!!
2026-09-22 05:16:21
4
alona :
u da real jev
2026-09-22 10:39:34
1
Caiu :
Great explainer 👌
2026-09-22 11:02:36
1
Bramble Rodriguez :
That’s really interesting
2026-09-22 13:14:59
1
ryanbracker3 :
algorithm
2026-09-22 07:07:23
1
AI & Automations :
thanks for sharing this Zauey 😎
2026-09-22 06:02:20
1
Skyler :
Is Jev capable of classifying images/videos or is the capability and speed limited to text?
2026-09-22 12:03:41
1
wwwwaawqqw1 :
undisclosed ad
2026-09-23 22:03:06
0
MrCoconut :
It feels a bit like ML with more steps
2026-09-24 19:41:22
0
selinaintheloop :
a model that just picks from answers you set sounds perfect for apps tbh, way less room for it to go off the rails
2026-09-24 18:24:44
0
Leo Marchetti :
I’m not obsessed with it. I’m not the same as all 😌
2026-09-26 13:03:44
0
I2C_Jason :
Pydantic is all you need
2026-09-22 11:41:57
0
To see more videos from user @zauey, please go to the Tikwm
homepage.