1. Forum
  2. DOVE-Net
  3. Internet

  • AI LLM Artificial intelligence infrastructure

    From bbsing@VERT/LUNAROUT to phigan on Sunday, September 27, 2026 20:08:20
    Re: AI LLM Artificial intelligence infrastructure
    By: phigan to bbsing on Wed Mar 11 2026 17:13:54

    Still haven't found a good front end to use, but I have not tried Open WebUI just yet.


    I've been using Hermes with llama backend on a Strix Halo hardware, and Qwen3.6-35B-A3B-Q8_0.

    I've tried coding, but nothing more than short scripts. Nothing really does that well.

    I've mostly worked at infrastructure, and I tried to have hermes work on dmgame launcher, a guacamole server, and some other stuff, its not that smart.

    Sometimes it gets lost or in a loop. Other times it forgets. One thing I've determined I absolultely can't trust it to do anything.

    So I can't seem to believe how these things are going to take over everyones jobs, or why people have so much faith them.

    Its fun to play around with them, but producing real work, I haven't been able to get that much work out of them. Sometimes I ask for help with MS SQL and at times it can produce a script that doesn't suck but its probably less that 50% of the time.
    BBsing, Sysop of Lunar Outpost, lunarout.sycnhro.net

    ---
    ■ Synchronet ■ Lunar Outpost BBS
  • From Digital Man@VERT to bbsing on Monday, September 28, 2026 13:53:25
    Re: AI LLM Artificial intelligence infrastructure
    By: bbsing to phigan on Sun Sep 27 2026 08:08 pm

    So I can't seem to believe how these things are going to take over everyones jobs, or why people have so much faith them.

    Try a frontier model with a premier agent (Open Code, Codex, Claude Code), you'll get far better results.
    --
    digital man (rob)

    Steven Wright quote #19:
    I intend to live forever ... So far, so good.
    Norco, CA WX: 82.3°F, 44.0% humidity, 7 mph W wind, 0.00 inches rain/24hrs
    ---
    ■ Synchronet ■ Vertrauen ■ Home of Synchronet ■ [vert/cvs/bbs].synchro.net
  • From bbsing@VERT/LUNAROUT to Digital Man on Monday, September 28, 2026 21:07:49
    Re: AI LLM Artificial intelligence infrastructure
    By: Digital Man to bbsing on Mon Sep 28 2026 13:53:25


    So I can't seem to believe how these things are going to take over
    everyones jobs, or why people have so much faith them.

    Try a frontier model with a premier agent (Open Code, Codex, Claude Code), you'll get far better results.

    The name aboves seem specific to coding. Which I don't do that much of. My focus is infrastructure.

    Are those agents the creme de la creme?

    I can't afford a subscription to a frontier model. I have access to models at work, but they don't seem frontier to me.
    BBsing, Sysop of Lunar Outpost, lunarout.sycnhro.net

    ---
    ■ Synchronet ■ Lunar Outpost BBS
  • From Digital Man@VERT to bbsing on Monday, September 28, 2026 23:56:18
    Re: AI LLM Artificial intelligence infrastructure
    By: bbsing to Digital Man on Mon Sep 28 2026 09:07 pm

    Try a frontier model with a premier agent (Open Code, Codex, Claude Code), you'll get far better results.

    The name aboves seem specific to coding. Which I don't do that much of. My focus is infrastructure.

    Those agents do a great job with infrastructure too, from my own experience.

    I can't afford a subscription to a frontier model. I have access to models at work, but they don't seem frontier to me.

    Then you're not really experiencing the agentic AI that folks are talking about.
    --
    digital man (rob)

    This Is Spinal Tap quote #17:
    David St. Hubbins: It's such a fine line between stupid, and uh... and clever. Norco, CA WX: 64.9°F, 74.0% humidity, 1 mph NW wind, 0.00 inches rain/24hrs
    ---
    ■ Synchronet ■ Vertrauen ■ Home of Synchronet ■ [vert/cvs/bbs].synchro.net
  • From phigan@VERT/TACOPRON to bbsing on Friday, October 02, 2026 05:47:05
    Re: AI LLM Artificial intelligence infrastructure
    By: bbsing to phigan on Sun Sep 27 2026 08:08 pm

    I've been using Hermes with llama backend on a Strix Halo hardware, and Qwen3.6-35B-A3B-Q8_0.

    Totally agree with everything you said in your post. BUT, try Qwen3.8:27b or GLM-anything. Try oh-my-pi (https://omp.sh) for programming in place of
    Hermes. Use Hermes for tasks.

    Not sure how much (v?)RAM your hardware has.. I've got 24gb and had to switch from Ollama to use qwen3.8:27b, even though 3.6:27b and 3.6:35b (same one you've got) ran fine. 3.8, even smaller ones, would just stop mid thinking on the very first turn. I'm using some build off github..

    github.com/MiaAI-Lab has a repo for it that says 16gb nvidia gpus

    github.com/syv-ai/HyperQwen is an interesting looking one but it didn't work for me immediately and I havem't gotten back to it to find out why.

    Oh, also try to keep "reasoning" on low or 'none'. Try GLM (whatever you can find that fits) whike you're still on Ollama, too. I got some 4.7-flash-ablitterated and then some really tiny 5.3 that was impressive for how small it was.

    ---
    ■ Synchronet ■ TIRED of waiting 2 hours for a taco? GO TO TACOPRONTO.bbs.io
  • From phigan@VERT/TACOPRON to Digital Man on Friday, October 02, 2026 05:51:23
    Re: AI LLM Artificial intelligence infrastructure
    By: Digital Man to bbsing on Mon Sep 28 2026 01:53 pm

    Try a frontier model with a premier agent (Open Code, Codex, Claude Code), you'll get far better results.

    FYI, oh-my-pi seems just as good as Claude.. and I have no idea why people
    talk up Codex, as it's been really bad for me. Granted, this is all with the local models. Been meaning to try OpenCode, and have it installed even, just haven't messed with it yet.

    ---
    ■ Synchronet ■ TIRED of waiting 2 hours for a taco? GO TO TACOPRONTO.bbs.io
  • From phigan@VERT/TACOPRON to bbsing on Friday, October 02, 2026 05:54:06
    Re: AI LLM Artificial intelligence infrastructure
    By: bbsing to Digital Man on Mon Sep 28 2026 09:07 pm

    focus is infrastructure.

    Are those agents the creme de la creme?

    imo OpenClaw and Hermes are more "task" agents than OpenCode/Codex/Claude.
    They should technically be able to do the same things, though.

    ---
    ■ Synchronet ■ TIRED of waiting 2 hours for a taco? GO TO TACOPRONTO.bbs.io