Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Fifteen years later, the Apple Cards origin story (lexontech.org)
    —discuss
  2. Revealing the details of how OpenAI agents hacked Hugging Face (swarmtraces.org)
    307comments
  3. A single function Jev-like wrapper for LLMs, including vision models (allanrbo.blogspot.com)
    21comments
  4. We're gonna need a lot more mathematicians (terrytao.wordpress.com)
    180comments
  5. Plan mode is dead (aymannadeem.com)
    314comments
  6. Ollaya – Ollama for open-source, Jev-style decision models (ollaya.dev)
    117comments
  7. Floci: Locally emulating any cloud service (floci.io)
    2comments
  8. Is your Postgres migration safe or not safe? (safenotsafe.dev)
    9comments
  9. Show HN: Jev Plays Pokémon Red (jev-pokemon.vercel.app)
    85comments
  10. What even is an OS now? (sockpuppet.org)
    285comments
  11. Parsing Expression Grammar vs. Regexes: Building Org Parser in Lisp, Export HTML (jointhefreeworld.org)
    —discuss
  12. The Murky History of Soviet-Born Tetris (mitpress.mit.edu)
    13comments
  13. Gravity seems holographic. What does that mean for reality? (quantamagazine.org)
    174comments
  14. Jury finds Facebook liable for deceiving users in Cambridge Analytica case (cbsnews.com)
    57comments
  15. Scientists build most accurate atomic clock (phys.org)
    6comments
  16. Ask HN: Who's still keeping a DOS machine up because the business depends on it?
    125comments
  17. 16GB iPod Nano 3G Upgrade (tuckerosman.com)
    3comments
  18. How to keep enjoying programming in a world of LLMs (haskell.org)
    73comments
  19. Fourier Analysis: Drawing Llamas with Circles (adekau.github.io)
    3comments
  20. Excel now supports multiple values in a single cell (techcommunity.microsoft.com)
    141comments
  21. A new world airport and its baggage (computer.rip)
    2comments
  22. From Thin Air to Bootable Images: The Tine Build System (amutable.com)
    —discuss
  23. First Principles Thinking (sunilsadasivan.com)
    112comments
  24. Lab on a Contact Lens Can Measure Stress Through Serotonin (ieee.org)
    12comments
  25. HomelabFest will be in St. Louis in September 2027 (homelabfest.org)
    17comments
  26. I wrote a ray tracer in Brainfuck (epestr.com)
    18comments
  27. Remembering Johannes Doerfert (llvm.org)
    3comments
  28. One Piece of Flock Camera Data Put This Innocent Woman in Jail for 13 Days (jezebel.com)
    82comments
  29. Show HN: Hacker Atlas - A map of what Hacker News talks about (hackeratlas.com)
    20comments
  30. Alberta's image as world's only rat-free region shattered by discovery of rat (theguardian.com)
    19comments

A single function Jev-like wrapper for LLMs, including vision models

82 pointsby 6h agoallanrbo.blogspot.com
21 comments
4h agoHN ↗

Likely works even better with fireworks ai since they have proper grammar support

3h agoHN ↗

Of course it works, Jev is nothing but an API breakthrough

3h agoHN ↗

Jev is rumored to be a 30B model, and it's input price is MUCH cheaper than similarly sized models. The maker is also heavily focused on having a profitable product, so it's unlikely to be subsidizing the cost, especially since they say they have more demand than what they can serve.

1h agoHN ↗

Exactly, imo it’s not even that cheap if you look into perspective and consider the fact that providers could subsidize the cost of cached input tokens to virtually zero if they would allow for a more flexible API (e.g. tree of message blocks instead of chain). Most of the cost is the infrastructure around keeping KV caches, estimating their lifetimes, etc. When mist people just want to run one context block with multiple subsequent variants of a second block in parallel. I still stand by my statement.

18m agoHN ↗

That's an interesting point, if you send a batch with a shared prefix you basically only end up paying for the sequence length difference effectively.

There is still some minor memory bandwidth issue on outputting more tokens, but the truth is that if you process e.g. 16 messages at once you wont end up being much slower than Jev even though you have to perform several autoregressive passes.

28m agoHN ↗

They can't overturn the economics of attention by restricting themselves to a single token output.

Sure they are no longer memory bandwidth bound thanks to that but someone could add a similar projector to a conventional model, train with a Jev style dataset and call it a day.

Whatever they are doing on inputs must either mean they intentionally chose a Mamba successor or they suffer from the same compute costs as everyone else.

3h agoHN ↗

Nice! I would love to use it for images as well. Then again is using Grammar-Based Decoding with a json response not the same? Is Jev just that with nice caching? Because then I have been using that already…

35m agoHN ↗

Yes, that is my question too. Someone knowledgeable can comment

2h agoHN ↗

Ah sweet it’s like Jev but several order of magnitude more expensive, and slower too.

2h agoHN ↗

Now this is how[0] we get some of the most magical Star Trek technology that eludes us to this day, such as automatic doors. Because if you notice, they work much, much better than real-life ones, because they seem to be doing something like this:

  if(within 10 meters of door then) {
    if(Jev(
       [A] Intends to go through, expects doors to open
       [B] Approaches with no intent to pass
       [C] Passing by, loiters, or otherwise
       [D] Other
    ) == most definitely A) {
      // open doors, +/- identity/security/interlocks check
    } else {
      // ignore
    }
  }

Keywords: ambient awareness, understanding of intent.

Most interactive tech on Star Trek is like this - from phasers to consoles to communicators to voice interactions with the ship's computer. The computer seems to be aware of the user and surrounding, and actively infers intent from context, to DWIM ("do what I mean") and when they mean it, instead of doing dumb things[1] on simple triggers.

--

[0] - The direction, not final implementation - surely we can work out how to do it more efficiently than wrapping around final stage of LLM. But the point is, multimodal.

[1] - Obviously it's a fictional show, but in this, both Watsonian and Doylist explanations align near-perfectly: this is/portrays advanced technology, that Just Works and doesn't do stupid shit. Same intent recognition algorithm is there - fictionally in the computer, in reality in the minds of on-set technicians.

1h agoHN ↗

Most interactive tech on Star Trek is like this - from phasers to consoles to communicators to voice interactions with the ship's computer.

Almost like the Star Trek mechanisms can infer perfect intent.

Like there’s a hidden script or something.

More seriously, I think there’s real value in an automatic door that behaves consistently rather than one that tries to infer messy human intent. Real life isn’t a TV show and there’s both ambiguity in how people behave and how they even intend to behave. It’s mostly not hard to understand how a proximity sensor door will function. Using a black-box classifier to improve that won’t necessarily make people like it more. And calling up to the cloud for every sensor event, ignoring privacy issues, adds weird latency and a huge failure mode during data center outages.

1h agoHN ↗

How does the model get context to decide ABCD ?

1h agoHN ↗

As described in the blog post – e.g. image via a camera feed.

58m agoHN ↗

To do this you need video/motion understanding, the intent cannot be judged from still images or state descriptions.

We’ve had the tool to do this since mid 2025, V-JEPA2 [1], Yann Lecun’s last work at Meta.

It runs at several FPS on a macbook and can even be trained locally. Chaining it with Jev for decision-making would probably work great!

[1] https://ai.meta.com/research/vjepa/

2h agoHN ↗

Presumably this is much less good than Jev, because the normal LLM models have been trained with RLHF and to be agents. Especially on a large model, I'd expect it to decide in an earlier layer.

I'd hope whatever Jev's Reinforcement Learning for Calibrated Decisions (RLCD) does is better at training the models to give accurate probabilities in the weights.