Live data from Hacker News

Viewing profile — frontsideair

frontsideair

HN member
Joined
Fri, Jun 15, 2012, 12:36 AM UTC
HN karma
259
Public activity
89 items

About frontsideair

[ my public key: https://keybase.io/frontsideair; my proof: https://keybase.io/frontsideair/sigs/OkvxxcDuiNS18hJBaXqcsqVwPSSRVzb2q7QR8goRDzs ]

Recent public activity

  1. comment
    Comment #47653132

    I was able to confirm that it actually runs on ANE, I'm impressed.

  2. comment
    Comment #47627262

    > Apple locked it behind Siri. apfel sets it free This doesn't feel truthful, it sounds like this tool is a hack that unlocks something. If I understand it correctly, it's using th…

  3. comment
    Comment #45654488

    Yeah, the initial experience with no colors doesn’t look great. I can implement this when I have some free time, if you feel like doing it please feel free to open a PR. Thanks!

  4. story
  5. comment
    Comment #45178495

    14B Qwen was a good choice, but it became outdated a bit and seems like the new version of 4B surpassed it in benchmarks somehow. It's a balancing game, how slow a token generation…

  6. comment
    Comment #45178257

    Ollama adding a paid cloud version made me postpone this post for a few weeks at least. I don't object them to make money, but it was hard to recommend a tool for local usage and m…

  7. comment
    Comment #45178232

    This is the command probably: sudo sysctl iogpu.wired_limit_mb=184320 Source: https://github.com/ggml-org/llama.cpp/discussions/15396

  8. comment
    Comment #45178212

    I'm interested in this, my impression was that the newer chips have unified memory and high memory bandwidth. Do you do inference on the CPU or the external GPU?

  9. comment
    Comment #45178106

    Good point, let me add a quick note.

  10. comment
    Comment #45178082

    Thank you, it was the integral part of the whole post!

  11. story
  12. comment
    Comment #44814505

    According to the benchmarks, this one is improved in every one of them compared to the previous version, some better than 30B-A3B. Definitely worth a try, it’ll easily fit into mem…

  13. story
  14. comment
    Comment #43277313

    This is the first time I’m hearing about Nebula. How does it compare to Tailscale?

  15. comment
    Comment #43277310

    I saw from the comments that they provide smaller binaries, which may have worked on my Raspberry Pi. Maybe I’ll give it a try one day. https://tailscale.com/kb/1207/small-tailscal…

  16. comment
    Comment #43277295

    Thanks, this could’ve worked for my Raspberry Pi! I would’ve tried it if it was still in commission. Next time?

  17. comment
    Comment #43276980

    That’s exactly why I put the spend limit, and this was the first time to confirm that it actually works.

  18. comment
    Comment #43276973

    Exactly, I had a spend limit since I didn’t want to break the bank. It’s back up now.

  19. story
  20. story
  21. comment
    Comment #39525905

    Synthetic benchmarks should be retired. Build a real app, measure real impact. Unless your users insert 1000 rows, of course. Other than that, I agree with the general premise of R…

  22. comment
    Comment #39522332

    This looks like a treasure trove that I somehow missed! I'll give it a more thorough read later, but gave my star for now, thanks!

  23. comment
    Comment #39522312

    I'm really interested in `home-manager` honestly, and while I agree about the benefits of declarative package management, I find it easier to type commands at a terminal like `nix …

  24. story
  25. story