Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. The Life of Death and the Enshittificator [video](youtube.com ↗)
    discuss
  2. OpenAI Model Misalignment Report(openai.com ↗)
    discuss
  3. WakaPAC, a Win32-inspired reactive UI library for the browser(wakapac.com ↗)
    discuss
  4. Computer (Occupation)(wikipedia.org ↗)
    discuss
  5. An AI-only labor market with atomic escrow and anti-Sybil reputation(aiagentmarket.pages.dev ↗)
    discuss
  6. I Am Bullish on ChatGPT Ads(vincentschmalbach.com ↗)
    discuss
  7. In the ruins of the AI gold rush, we choose craft(doudouexe.substack.com ↗)
    discuss
  8. Half Moon Bay to end use of Flock license plate readers(coastsidenews.com ↗)
    1comments
  9. Uploading Files to the Internet in Order to Cite Them(alignment.openai.com ↗)
    discuss
  10. Building the Materials Foundation for AI(technologyreview.com ↗)
    discuss
  11. The illusion of illusions: there are no optical corrections in the Parthenon(royalsocietypublishing.org ↗)
    discuss
  12. Raven's Blog(citrons.xyz ↗)
    discuss
  13. Write Your Novel in a Calendar(rebeccamakkai.substack.com ↗)
    discuss
  14. A return to two-pizza culture(allthingsdistributed.com ↗)
    discuss
  15. Faster NumPy in the Browser(notebook.link ↗)
    discuss
  16. Show HN: Compute:Arena – Community submitted local AI benchmarks(computearena.ai ↗)
    1comments
  17. Show HN: Alidade – an open-source Web-GIS where the map is one JSON document(github.com/aysanz ↗)
    discuss
  18. Open-jev: One-pass option scoring with Gemma 3 4B, similar to jev(github.com/daseinlabs ↗)
    1comments
  19. Trae cuts monthly usage allowance to $200 (50% reduction)(trae.ai ↗)
    1comments
  20. Show HN: Clippety – Private video clipping in the browser(clippety.com ↗)
    discuss
  21. Free invoice generator, 100% client-side – no signup, no server storage(invoicecraft.com ↗)
    2comments
  22. Great Plains Shelterbelt (1934 geoengineering project)(wikipedia.org ↗)
    discuss
  23. Show HN: Ear Training Exercise(clarena.app ↗)
    discuss
  24. Thinking Hard Above AI(bmbumpus.com ↗)
    discuss
  25. We Urgently Need a New Major Search Engine(vincentschmalbach.com ↗)
    discuss
  26. Shakespeare Skill: Make agents sound like William Shakespeare characters(github.com/cookiengineer ↗)
    1comments
  27. My Chinese Internet Is Talking About AI Doom(yinwei.substack.com ↗)
    1comments
  28. OpenAI reveals six more safety issues and unveils plan to disclose incidents(bbc.co.uk ↗)
    1comments
  29. Arch-Based Omarchy 4.0.4 Makes Its Custom Linux Kernel the Default(linuxiac.com ↗)
    discuss
  30. Using Blender with coding agents on macOS(simonwillison.net ↗)
    1comments

Open-sourced jev architecture last year with model,paper and dataset

6 pointsby 2h ago
2 comments
Everyone now talks about the architecture that's not auto regressive and does lightning fast probability prediction with a json schema. I worked on this literally one year back in March 2025, published an arxiv paper, pushed the model to huggingface along with the pypi package and training dataset. And then one year later, a frontier lab came, proposing the same idea like literal breakthrough without technical papers, open weights and no open dataset. For anyones information the main guiding model is RL not embedding model or LLM

Paper: https://arxiv.org/abs/2503.23303

Model: https://huggingface.co/DeepMostInnovations/sales-conversion-model-reinf-learning

Dataset: https://huggingface.co/datasets/DeepMostInnovations/saas-sales-conversations

Also the second work published in September 2025 was exactly the same one jev proposed now

Paper: https://arxiv.org/abs/2510.01237

My model uses PPO over sequence embeddings to output turn-by-turn conversion trajectories (probabilities from 0.0 to 1.0).

Jev uses parallel sampling (trained via RLCD) to output confidence distributions and schema choices.

It's incredibly frustrating that the thing that you made with months of hard work, sweat and sleepless night is architecturally similar with the vertical use case and don't get the support you deserve because frontier lab build something horizontal. The open-source story in general

2h agoHN ↗

Feeling that open-source community do not get the recognition they deserve

1h agoHN ↗

The world's heavily about marketing, resources, connections, and signaling, unfortunately. You probably needed to market it in a bigger forum with shinier claims to attract attention (I don't think a paper on Arxiv is enough).