Introducing System One Models and Jev

I think I missed why is this faster? What I’m reading here is it’s similar to constrained decoding but I’m not seeing the explanation of why it’s able to get those results.

Will need hands on to truly tell, but the doom demo seems very promising. If it can play that with text descriptions of where stuff is by distance and degrees in a 3D context then many GUI automation tasks should be easily doable

Would it be fair to say that this is tailored for tool-selection subagents?

I don't know if it's just me but comments under Twitter post felt like paid partnerships to me.

The idea sounds cool though

This could be killer for ingame AI for grand strategy games like Victoria 3 or EU5

So it's kind of like BERT but you don't have to train it for each request/response shape ?

Huh this looks fantastic. The Doom demo really sold for me that this could be a great tool for accelerating QA at my gamedev studio. Signed up for early access.

Woof, that page is hard to read. I don't understand what they've done to the way text is rendering but it's not great for my eyes.

insane doom demo i wonder what the limits of its intelligence are? i'm guessing it's not great at reasoning tasks, it seems breaking down the problem helps significantly, but how much does a problem need to be broken down for reliable performance? also this would be huge if it could run locally but it seems like there's no intention to do that at the moment

So... a classifier model?

This puts the human even more out of the loop I'll guess?

Is the tradeoff of the parallel output that we don't get arbitrary string generation? like output # of tokens is fixed ahead of time?

Either way, really cool and impressive.

> AI Map Reduce over Big Data

> Search for relevant information over giant corpuses

Do you mean as an alternative to embeddings?

This is best thing to use for decision making, evaluation, classification. If I'm not wrong.

Is this a Markov/Diffusion model with some sort of external Engram memory? If so, this could be extremely interesting.

Funny how it can do everything but not chat. Sort of how when I was a kid I thought of a medicine that could cure any disease except the common cold.

If it is so cheap, why such a limited release?

Very cool. LLMs have been borderline unusable as functions for the longest time, very excited for this direction.

"can't hallucinate" feels like some word game Olympics

What's the difference compared to just taking an embedding and feed forward a simple net trained for the task?

If I understand correctly, it can play chess and rubic cube better than LLM ? ( may be go too ? )

If it work as good as they say it does, confidence score + really fast response when you want very fast response, basically.. To me it is a crime against humanity to not open source it. Just get the money from cloud inference and cloud agentic sessions or whatever but open source it. This tech, a good harness, a good model provider, and you have basically a AGI building machine.

This is quite the paradigm shift. Can't wait to get my hands on it.

> Extraordinary claims require extraordinary evidence so see below for the receipts.

Yes, that’s the kind of attitude I want to see in these model releases

Can this be used in conjunction with a text-generating LLM for better quality code generation?

this reminds me of laravel boost, which does something similar.

it can generate classes/models/routers etc via tool calls, doesnt write the actual code.

Can HN have a tag for open-weight vs closed-source models please? The progress is nice, but if it is not released at least in papers or open-weight? These are just ads?

The whole page reads like it was vibe-written by an AI. If I'd built something as disruptive as this claims to be, I'd have spent at least fifteen minutes writing the announcement myself. Every time I see 'we' in an announcement like this, I picture one guy alone in his basement.

The technology and the results are very handwavy. What is RLCD exactly ? What are scores on benchmarks compared to LLMs ?

This website does not inspire confidence at all, it all sounds like a marketing piece. I wish it was true, some kind of text-prompted classifier with LLM performance would be cool, but I can't trust it with what we are given.

Seems like "some" of LLMs tasks are now Jev tasks.

This looks and feels a lot like productionized conformal prediction

If I’m understanding correctly, this will work well for self driving cars?

Can’t wait to use this. Amazing work.

Okay, so it doesn't output text, that much is understood. What are the inputs like? I'm assuming maybe a text input? maybe an AST definition? Really hard to tell how this works at all from the demos, especially since we can't really try it out.

Why did they pick the name System One? It's not really explained what "System One tasks" and "System One shaped queries" are. Things that need a fast response?

Does this imply it's a very small model? I couldn't find anything about the model itself.

The eval is baffling me

> we assume there is a correct compute graph (a “workflow” represented in code) and use the predictions of the largest, smartest, and most expensive external models as reference probabilities. ... Rephrased: every model gets the same workflow. We test how they compare to the average of the smartest models (in this case, Astra and Fable).

They assume there is a correct graph, but they don't compare to that, they compare to the average of the smarts models? So the smartest models are getting it wrong but you compare that anyway as a benchmark? So the outcome is "how much of a Fable am I getting" etc. Why not compare the actually correct thing?

But then even on this hand constructed eval, the first plot is showing Jev at less than Sonnet 5 accuracy. It is barely better than Luna. There are two Opus 5's and two Sonnet 5's without explanation. What is the plot showing?

I gave up.

Super intrigued by this - large scale automation using LLMs is quite annoying due to deprecation cycles of models from frontier labs and cost of running your own being prohibitive when you have a blend of them.

A little off-topic, I have to admit, I did play Doom back in the 90s and I know it is just a game.

But is it just me or other folks feel uncanny seeing "a machine" playing Doom with low-latency as the first demo.

While I'm optimistic that humanity is good (but sometimes makes bad decisions), My first thought was seeing such models used by armies.

Is there a downloadable technical report somewhere?

Any relation / inspiration to GLiClass?

Can this be used in practice to write code?

um what is going on with the outfit changes in the launch video...

https://x.com/CompleteSkeptic/status/2099925682726002904

hi

I like it. What is it?

This is potentially huge and can crash the Big Two's stock prices or block their IPOs completely.

I could put this to use today.

I think we'll see a bunch of different architectures over the next five years.