Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Jeff – Jev-compatible 0.8B decision models, trained at home, ~30 ms (github.com/firelex)
    94comments
  2. Pirating the Pirates (mubi.com)
    213comments
  3. 12,000-year-old Göbeklitepe burials explain scattered bones (archaeologymag.com)
    18comments
  4. MicroLLM Lab – Try 7 tiny LLM's in the browser (stateofutopia.com)
    59comments
  5. 1996 chat room simulator connected to Win95 and System 7 web desktops (lolchat.rip)
    5comments
  6. Sonnet 5.5 (anthropic.com)
    389comments
  7. Scientists solve 1840s space weather mystery (arstechnica.com)
    36comments
  8. World Labs Is Joining AMD (worldlabs.ai)
    69comments
  9. What is the best shape of a city? Modelling effect of urban form on distance (sagepub.com)
    6comments
  10. Hijacking the PS5's RTMP stream (yashgarg.dev)
    65comments
  11. How to win a beer with high-dimensional statistics (jamiesimon.io)
    —discuss
  12. Kids turned low-traffic NPR Spotify comments into a secret group chat (thisamericanlife.org)
    171comments
  13. Parley: Federated, decentralised chat that speaks plain IRC (mills.io)
    167comments
  14. Does Reddit have an astroturfing problem? What the data suggests (petervijeh.com)
    133comments
  15. California farmers are struggling to sell grapes as demand for wine drops (kqed.org)
    75comments
  16. ESP32S3 cluster running 1.58-bit (BitNet) Language model (github.com/low-zi-hong)
    1comments
  17. It's Time to Investigate the AI Labs (calnewport.com)
    102comments
  18. Nvidia wants to put a watchdog chip next to every AI agent (cnbc.com)
    140comments
  19. Show HN: HN.watch – Videos of all Hacker News posts (hn.watch)
    79comments
  20. Cf: The Agentic CLI for the Cloudflare API (cloudflare.com)
    44comments
  21. OpenAI Says It Will Not Release Newest A.I. Model Over Safety Concerns (nytimes.com)
    15comments
  22. First Steps of the PLC Organization – Independent Public Ledger of Credentials (plcred.org)
    19comments
  23. What reversing, modernising old games tells us about the economic impact of AI (isfine.org)
    17comments
  24. Updated Google Maps shows destruction of the city of Rafah (twitter.com/aliabunimah)
    106comments
  25. Behold the pawpaw (cbc.ca)
    9comments
  26. Show HN: Destroy Any Website with Stickman (spritefusion.com)
    27comments
  27. What heraldry and Japanese mon can teach about visual-identity generators (benovermyer.com)
    21comments
  28. Launch HN: Vespper (YC F24) – SOTA Docx MCP (vespper.com)
    8comments
  29. Coding is not solved (alexewerlof.com)
    425comments
  30. Who wrote Elizabeth I's most scathing letters? (smithsonianmag.com)
    21comments

Show HN: Jauvex, one app for Claude+Codex+Grok+Jev with two-way voice chat

1 pointsby 1d agogithub.com
1 comments
1d agoHN ↗

Hi author here, I've been working day & night over the last week building Jauvex, haven't been so obsessed about building something in a while.

I originally started this project as a two-way voice chat for Claude, because I've been missing that from Codex, which has a great voice chat. I really enjoy talking to my computer and orchestrating agents that way instead of typing.

I tried to improve some things that Codex hasn't yet: Jauvex doesn't talk over you, it wont interrupt you, instead it will wait until you finished talking and only then will reply to you.

I tried to reverse engineer the way that Codex built their voice chat, and I came up with something like this:

1) Immediate response ("One moment, I'm checking that") 2) Context Reinforcement ("So you want to do...") 3) Response Summary ("Here's what I've found about what you've asked...")

Jev is connected optionally as a way to decide whether the first message is a statement, a question, a request, or something else, which helps the voice chat to feel more natural.

I've also built a Pro version which is invite only, and it allows you to add multiple organizations, use the same agents with your team (multiplayer AI agents), routines & workflows, todo lists.

Testing now as well different transcribers and text-to-speech models (testing NVIDIA's over Whisper, but Whisper.cpp has given better results).

And my plan next is to be able to download an agent and move to different workspaces.

The version released currently allows you to experimentally change the provider of the main agent (Jauvex agent) without losing the context, so you can start a conversation with Claude and then continue it with Codex normally.

Next up adding support for Linux & Windows.