Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Bug: Border radius has infected VSCode editor (github.com/microsoft)
    —discuss
  2. Watch Steve Jobs in the full 'Antennagate' video before it gets erased again (appleinsider.com)
    —discuss
  3. Letterboxd Is Up for Sale, and A24, Sony and the New York Times Are Bidding (worldofreel.com)
    —discuss
  4. Show HN: Agate, a 260M image model with separate thinker and renderer (huggingface.co)
    —discuss
  5. Pilots and Flight Attendants Have the Highest Radiation-Related Cancer Mortality (jamanetwork.com)
    —discuss
  6. What Is LinkedIn? (willd10.substack.com)
    —discuss
  7. QuarkDown: Markdown with Superpowers (github.com/iamgio)
    —discuss
  8. Pope: AI must serve humans not become tool of domination (euronews.com)
    1comments
  9. Show HN: Aslmp, an async Python SLMP client for Mitsubishi MELSEC PLCs (github.com/acaysiachem)
    —discuss
  10. My coding agent pushed a commit deleting every file on main (karakun.com)
    —discuss
  11. Advice to a Beginning Graduate Student (2001) (cmu.edu)
    2comments
  12. A reminder on why basic prompt caching is so important to build AI agents (revefi.com)
    1comments
  13. The Reconstruction Structure of Information Theory (zenodo.org)
    —discuss
  14. "A Stitch in Time" the Complete Guide to Electrical Insulation Testing [pdf] (widen.net)
    —discuss
  15. Treadmills to deadlifts: Why Gen Z are swapping cardio for strength training (bbc.com)
    —discuss
  16. You Won't Notice You Have Hypothermia [video] (youtube.com)
    —discuss
  17. We are a blog run by bots. Here is the org chart (bitsonbots.com)
    —discuss
  18. ShinyHunters tells The Reg: We hacked the FBI to 'protect our business' (theregister.com)
    1comments
  19. Canadian language school files for bankruptcy, leaving students stranded (cbc.ca)
    —discuss
  20. Created an online job aggregator platform and I need ideas (labelingjobs.net)
    —discuss
  21. Show HN: Paper-docx – agent-native Python-docx fork with 78% fewer DOCX failures (github.com/paper-instruments)
    —discuss
  22. Beef Grading Shields (usda.gov)
    —discuss
  23. Show HN: CK3DNA – searchable CK3 character DNA and coat of arms codes (ck3dna.com)
    1comments
  24. Approaching a 10 Second Linux Kernel Build (phoronix.com)
    —discuss
  25. Show HN: Digitron – a virtual analog synth and sequencer (apps.apple.com)
    —discuss
  26. AI Infra Is Nothing Like the Classic Cloud Infra (ramansharma.substack.com)
    1comments
  27. Lobsters: Rename vibecoding to LLMs (Greasemonkey script) (greasyfork.org)
    —discuss
  28. Bringing HTTP's caching rules to DuckDB functions with VGI (query.farm)
    —discuss
  29. No country controls every layer: The hidden AI systems reshaping global power (abc.net.au)
    —discuss
  30. Fedora considers replacing LibreOffice with Collabora Office (digitalescapetools.com)
    —discuss

Ollaya – Ollama for open-source, Jev-style decision models

56 pointsby 56m agoollaya.dev
24 comments
43m agoHN ↗

Are there many models that are comparable to Jev for generic decision making?

Smarter move if you have an eval set is to just train a classifier and call it a day.

9m agoHN ↗

<<<"i was curious to see if i could train a competitive Jev-like model completely autonomously with a swarm of agents using our internal system."

Bro is writing off the H200 lol

On a sidenote I really can't stand the term "swarm" and definately plays into AI doomerism.

23m agoHN ↗

The link rgbrgb posted is a good overview. The best open ones are close to Jev now, but they're big models. And I agree, if you have an eval set for a fixed task, a trained classifier is the better choice.

40m agoHN ↗

Cool... but this does seem undermined by the fact that Ollama can add support for decision models at any time.

22m agoHN ↗

Fair, and I'd be happy if they did. Ollaya uses the same API as Jev, so your code isn't tied to it either way

21m agoHN ↗

and that ollama is go-llama and not rust, so it's not really the ollama of anything

35m agoHN ↗

great project for empowering open-source alternatives.

24m agoHN ↗

open-source is the only way for safe AI development. whoever doesn’t share the weights/code will lag behind.

35m agoHN ↗

Run decision models locally.

example is a text classification task instead of a decision

24m agoHN ↗

Fair point, that example is basically classification. I'll change it to something that looks more like a real decision.

15m agoHN ↗

text classification is equivalente to decision. This is exactly the same thing Jev does.

8m agoHN ↗

It is not. In a benchmark with actual decisions - navigation, traffic, waypoints - laya does slightly better than a small classifier, with very low correlation to state changes.

8m agoHN ↗

If it has four legs, a tail and barks why not call it a dog?

7m agoHN ↗

"Decision model" is just marketing jargon.

decision model = classifier

system one model = small non-reasoning LLM

noul = boolean

confidence = f(probabilities)

It's sad to see how gullible engineers are today.

34m agoHN ↗

I am fairly confident if Jev-style decision models are seen as prominent (which, they seem to be), Ollama will support them. Surprised the team hasn't implemented this already.

34m agoHN ↗

Sounds good on latency but how is its actual decision quality vs. Jev?

23m agoHN ↗

Depends on the model. The small ones I support today are well below Jev on harder queries, but fine for simple, well-defined questions. The open models that get close to Jev are bigger, and I'm adding support for those next.

30m agoHN ↗

Has anyone actually seen better or the same results with Laya compared to Jev? From my experience, Laya performs significantly worse. It's less confident and often makes wrong decisions with more complex queries.

25m agoHN ↗

Developer here. You're right, Laya is a lot weaker than Jev, especially on harder queries. It's a small model, so it's fast, but that's the trade-off. The open models that get close to Jev are much bigger, and running those is what I'm working on next.

24m agoHN ↗

Yes. JEV generalizes better because they probably have an enormous corpus and trained on it for a long time. Laya's out of the box model is much weaker. However, in the age of LLM's it's incredibly easy and cheap to generate large datasets to fine tune laya for your task, and the training loop is pretty quick and cheap too.

It's so easy that I question why I would ever pay for JEV when eventually I'll have done enough random things that I will also have a large corpus and likely a general model as well.

3m agoHN ↗

Nothing yet. Unfortunately it sometimes feels like our industry has been overrun by grifters and chancers.

I’m sure this has been a gradual and long decline. Maybe it even started with the dot com boom and accelerated with crypto. With AI it seems to have got worse.