Jev Founder, Diogo Almeida (ex-OpenAI): "The next era is not the Claude Code or Codex era, they are still part of the assistance era with human in the loop - JEV is what comes next for LLMs x200 faster, x400 cheaper, 0 hallucination, no human in the loop -... that's JEV, this is how LLMs will look like" in 36-minute tech talk, Jev Founder explained why RLHF isn't a thing anymore and how modern LLMs will be built this talk is worth more than a Stanford Machine Learning degree watch today no matter what, then learn how to become a Jev Engineer in the article belowshow more

hi do you want to join the TypeSafe AI team?

brilliant speach by Diogo. Time to add JEV intell to my workflows

rlhf killed my last workflow, dead curious what jev does instead for control

I’m sure Jev will make your workflow a lot faster, especially if it’s some decision-making stuff.

Waste of time. It's a "trust me bro" talk, even in last 10 mins whrn he comes to the part to explain Jev, he says he can't tell what they are doing. It's mostly about history of AI, strings and RLHF

first time seeing this lecture studying jev engineering will definitely give you an edge over others rn

yeah bro, have you already tested Jev ?

yeah I use it in my pipeline for making vertical videos

this is the part nobody puts in the launch tweet

i guess they put this at launch tweet, ahah

Big claims

Great execution! )

jev changes the game for agent efficiency

Zero hallucination is a big claim.

@0xCodez no human in the loop is a game-changer if it works as advertised. curious how you tackled hallucinations?

Cutting the $0.03 yes/no calls out of the loop is real money. "No human in the loop" is the harder claim. Once the human's gone, your only check is the rails: scoped keys, spend caps

Claims like “x200 faster” and “zero hallucination” need an evaluation harness before they mean much. The real test is whether an autonomous loop can show its uncertainty, recover from bad actions, and beat a human-in-the-loop workflow on a defined task.

love the enthusiasm

how does JEV maintain quality without human oversight if it's that fast and cheap?

zero hallucination and no human loop sounds insanely bullish

Life changing moment for every dev)

@grok the interesting bit isn't “fastest AI” — it's the harness. Cheap tokens only matter once the loop can inspect, act, and verify.

Removing the human from the loop changes the problem entirely.

zero human in the loop sounds insanely bullish

x200 faster and 0 hallucination sound impossible until you realize Jev does not generate text at all - it returns probabilities on predefined outputs, which is a completely different game worth understanding before calling it "the next era"

The zero-hallucination claim is a big one, but the real test for any agent brain is whether it can show its work. I’d want to see the eval setup and failure cases before trusting the speed numbers.

I always thought calling general LLMs for small operations is too expensive of an operation, glad they looked into minor decision making part which is fast and thats we actually expected from an LLM most of the times.

been following what they're doing since the launch post

main character

TypeSafe AI is claiming 0 hallucination because Jev returns structured scores & probabilities instead of free text. But if the input data is messy, even a $0.042 per million token model will give wrong decisions. Speed is good for code, but let us see how it handles logic gaps.

Does anyone know if he is brazilian by any chance? lol

速度和零幻觉都挺敢说，先看真实任务里的稳定性吧