found the perfect use case for @typesafeai Jev: instant compaction in 2026, why is compaction still a summarization prompt? Jev can make it instant by scoring every tool call and dropping what’s irrelevant
Introducing System One Models and Jev
The Hacker News launch thread for Jev, with the CEO agreeing the shape is a zero-shot classifier and commenters mapping it to encoders, constrained decoding, and typed classification.
- Category
- Sites & Guides
- Published by
- Community
- Author
- —
- Added
- 2026-09-22
Tagscommunitymodelsevaluation
Highlights
- The thread drew 1,955 points and 511 comments, with discussion of the Doom demo, typed answers, and what "cannot hallucinate" actually covers.
- CEO Diogo Almeida agreed the shape is a zero-shot classifier, saying he prefers "zero-shot" over "instruction-tuned".
- He said the architecture is close to the chest for now, with a paper under discussion, and pointed to data rather than architecture as the interesting part.
- Commenters mapped Jev to encoder-only transformers with scalar and ordinal output heads, GLiNER-style label-conditioned encoders, and constrained decoding.
- Skeptics argued type safety is not factual correctness and that a forced boolean answer cannot abstain on out-of-range input.
Watch out
Discussion is dominated by people reacting to the launch rather than API users, and the CEO's architecture answers stay high level.
More like this
Latent Space's AINews roundup of launch day: the Jev thread atop Hacker News, the RLCD decision-model framing, and community readings that compare it to DSPy-style typed signatures for routing and scoring.
Sites & GuidesArticle#community#article#models
Community
638GitHub stars
A curated list of public projects and practices built on Jev: 212 entries across 13 categories, from classification and routing to agent decisions, calibration, infra/SDKs, and finance, with explicit inclusion rules.
Sites & Guides#community#awesome-list#agents
Communityawesome-jev-tools
396GitHub stars
An independent, community-maintained catalog of 640 open-source projects built with Jev across 11 categories, paired with the jevbest.com site and citing public evidence behind each entry.
Sites & Guides#community#typescript#awesome-list
Communityawesome-jev
From the community
Posts from builders shipping with Jev right now.
Instant compaction with Jev
A Claude session from 1M to 86K tokens
This is actually insane. This uses @typesafeai Jev model, as a plugin in Claude to review all the un-nesseasary tool calls, and it takes 1s to run! Like, literally, 1 second to take my Claude session from nearly 1M to ... 86K tokens! 😮 Ask your claude to install it and be Show more
found the perfect use case for @typesafeai Jev: instant compaction in 2026, why is compaction still a summarization prompt? Jev can make it instant by scoring every tool call and dropping what’s irrelevant
Vercel's fx safety reviewer, 18x faster
We're seeing extraordinary results from @typesafeai. Default mode in 𝚏𝚡 is auto, with a safety reviewer analyzing every command. That reviewer runs on GPT Luna today. Jev is up to 18x faster (p95) *and* more accurate. It's coming to @vercel AI Gateway and likely new default.
@fazxes
We benchmarked fx auto mode (safety) classifier with @typesafeai's Jev. tl;dr: ~5-18x faster and more accurate than 𝚐𝚙𝚝-𝟻.𝟼-𝚕𝚞𝚗𝚊, our current top choice
Jev lands on OpenRouter
Jev by @typesafeai is now on OpenRouter, in beta. Jev is a System One model. Instead of generating text, it takes your app's state plus a typed question and returns a typed decision with a probability attached. There is no JSON prompting, parsing layer, and nothing to validate Show more
700 leads scored for $0.09
JEV is INSANE. We gave it 700 high-intent leads and personalised outreach messages. In 40 seconds, it predicted how each message would perform, assigned a confidence score and detected lead-message mismatches. All for just $0.09. JEV can also score leads, analyse buying Show more
Beating Gemini Flash Lite on an eval
Ran @typesafeai's Jev against an existing classifier eval that previously used Gemini 2.5 Flash Lite. It won both on quality (saturated the eval) and speed (6x)
