• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    TypeSafe launches Jev, an AI model aimed at decisions rather than text

    Creator Diogo Almeida claims Jev is 20–200x faster and 40–400x cheaper, with free output tokens. Investor DCVC says it led TypeSafe’s $40 million seed round.

    Sasha RushSR
    AKAK
    Sander DielemanSD
    124 Sources, ,

    TLDR

    Jev’s creator, Diogo Almeida, announced its release on September 15, 2026, saying he spent two years in stealth developing the model and a training method called RLCD. He claims Jev is 20–200x faster and 40–400x cheaper, with free output tokens.

    Posts discussing Jev describe it as answering structured questions in parallel and returning choices or scores with probabilities and confidence instead of free-form text. One commenter suggests using that confidence to decide when human review is needed, while urging readers to wait for benchmarks.

    Investor DCVC says TypeSafe emerged from stealth with Jev backed by a $40 million seed round it led.

    Combined views

    25.6M

    124 Sources, first seen 22d ago

    Combined views

    25.6M

    124 Sources, first seen 22d ago

    102.7K likes
    22d ago
    first seen 22d ago
    102.7K likes
    5.3K comments
    67.2K saves
    8.6K reposts
    5.3K comments
    67.2K saves
    8.6K reposts

    Sentiment

    Positive85.1%14.9%Negative

    Summary

    Many accounts welcomed Jev for enabling fast, low-cost decisions that support economical scaling and specific automation use cases, while negative replies questioned its unverified claims and narrow applicability.

    Based on 445 sentiment-bearing replies from 388 accounts across 9 conversations.

    Sentiment

    Positive85.1%14.9%Negative

    Summary

    Many accounts welcomed Jev for enabling fast, low-cost decisions that support economical scaling and specific automation use cases, while negative replies questioned its unverified claims and narrow applicability.

    Based on 445 sentiment-bearing replies from 388 accounts across 9 conversations.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    124 Sources

    Diogo Almeida@CompleteSkepticAfter co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution22d
    AK@_akhaliqRT @CompleteSkeptic: After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last…22d
    Mike Taylor@hammer_mtA few weeks ago someone sent me a screenshot of the InstructGPT paper (that led to ChatGPT) with a name highlighted, and asked if I wanted to test a new language model Diogo was working on that doesn't output text... Couldn't resist, so here's my vibe check: https://every.to/also-true-for-humans/mini-vibe-check-typesafe-s-jev-judged-everything-i-ve-written-in-0-7-seconds?gift=atu6rcw7_-Uyiqx69NVhr2ai0SW65WkA22d
    Santiago@svpinoNew model that looks very different from everything else we've seen before: • It returns decisions with probabilities and confidence • Confidence will let you decide whether you need human review • It can answer multiple structured questions in parallel Let's wait for the benchmarks, but this looks pretty cool!22d
    elvis@omarsar0Recommended read. Jev gives up text generation to make AI dramatically faster. TypeSafe built a new architecture that answers structured questions in parallel, with RLCD training its probabilities to reflect how often it’s right. The team reports 40–200x faster responses on System One queries.22d
    will depue@willdepuediogo is an immensely creative guy and is working on really different types of models with the principle of building composable, programmable AI systems from layers of small inferences. it’s a weird and ambitious idea thats worth tinkering with22d
    Sander Dieleman@sedielem"output tokens are free, because they're finally too 🤬🤬🤬 cheap to meter" 😂 Reinforcement learning for calibrated decisions (RLCD) might finally bring about all the AI-based automation we were promised! Diogo and his @typesafeai crew are making it happen, very exciting! 🫨22d
    Ryan Lowe@ryan_t_lowebeen waiting for this to come out of stealth for a while now. once you sit with it, it seems clear that this set of ideas will enable a new paradigm of AI automation (that will grow alongside the current LLM + reasoning paradigm, which is complementary). lots of economic implications to consider. @CompleteSkeptic and co have been cooking on this for a while, congrats to the team!!!22d
    Marius Vach@rasmus1610The blog post and the use cases sound like this is like the model version of @DSPyOSS I'm excited.22d
    Rohan Paul@rohanpaul_aiAnother brilliant launch for developers: and its 20-200x faster than LLMs because it skips token-by-token generation entirely. TypeSafe AI just launched Jev, > 20-200x faster >40-400x cheaper (w/ output tokens free) > Frontier composable intelligence optimized for decisions So Jev is an AI model built to make software decisions instead of writing text for people. A normal LLM answers by generating tokens sequentially, so software often has to request structured output, parse it, validate it, and decide what happens next. Jev removes that translation layer: give it some data and a predefined question, and it returns a typed choice or score with probabilities and confidence. For example, a support app can ask whether a ticket is urgent, whether it violates policy, and which queue should receive it, then act on those answers directly. In code, Jev behaves like a smart if-statement: ordinary software controls the workflow while the model handles fuzzy judgments that rigid rules struggle with. Jev evaluates multiple structured questions in parallel rather than writing an answer token by token. TypeSafe reports 70-500ms responses and 40-200x faster performance than comparable LLMs22d

    124 Sources

    Diogo Almeida@CompleteSkepticAfter co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution22d
    AK@_akhaliqRT @CompleteSkeptic: After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last…22d
    Mike Taylor@hammer_mtA few weeks ago someone sent me a screenshot of the InstructGPT paper (that led to ChatGPT) with a name highlighted, and asked if I wanted to test a new language model Diogo was working on that doesn't output text... Couldn't resist, so here's my vibe check: https://every.to/also-true-for-humans/mini-vibe-check-typesafe-s-jev-judged-everything-i-ve-written-in-0-7-seconds?gift=atu6rcw7_-Uyiqx69NVhr2ai0SW65WkA22d
    Santiago@svpinoNew model that looks very different from everything else we've seen before: • It returns decisions with probabilities and confidence • Confidence will let you decide whether you need human review • It can answer multiple structured questions in parallel Let's wait for the benchmarks, but this looks pretty cool!22d
    elvis@omarsar0Recommended read. Jev gives up text generation to make AI dramatically faster. TypeSafe built a new architecture that answers structured questions in parallel, with RLCD training its probabilities to reflect how often it’s right. The team reports 40–200x faster responses on System One queries.22d
    will depue@willdepuediogo is an immensely creative guy and is working on really different types of models with the principle of building composable, programmable AI systems from layers of small inferences. it’s a weird and ambitious idea thats worth tinkering with22d
    Sander Dieleman@sedielem"output tokens are free, because they're finally too 🤬🤬🤬 cheap to meter" 😂 Reinforcement learning for calibrated decisions (RLCD) might finally bring about all the AI-based automation we were promised! Diogo and his @typesafeai crew are making it happen, very exciting! 🫨22d
    Ryan Lowe@ryan_t_lowebeen waiting for this to come out of stealth for a while now. once you sit with it, it seems clear that this set of ideas will enable a new paradigm of AI automation (that will grow alongside the current LLM + reasoning paradigm, which is complementary). lots of economic implications to consider. @CompleteSkeptic and co have been cooking on this for a while, congrats to the team!!!22d
    Marius Vach@rasmus1610The blog post and the use cases sound like this is like the model version of @DSPyOSS I'm excited.22d
    Rohan Paul@rohanpaul_aiAnother brilliant launch for developers: and its 20-200x faster than LLMs because it skips token-by-token generation entirely. TypeSafe AI just launched Jev, > 20-200x faster >40-400x cheaper (w/ output tokens free) > Frontier composable intelligence optimized for decisions So Jev is an AI model built to make software decisions instead of writing text for people. A normal LLM answers by generating tokens sequentially, so software often has to request structured output, parse it, validate it, and decide what happens next. Jev removes that translation layer: give it some data and a predefined question, and it returns a typed choice or score with probabilities and confidence. For example, a support app can ask whether a ticket is urgent, whether it violates policy, and which queue should receive it, then act on those answers directly. In code, Jev behaves like a smart if-statement: ordinary software controls the workflow while the model handles fuzzy judgments that rigid rules struggle with. Jev evaluates multiple structured questions in parallel rather than writing an answer token by token. TypeSafe reports 70-500ms responses and 40-200x faster performance than comparable LLMs22d