My AI website is now in it's marketing phase. It scouts competitor ads, plans, designs, deploys, and analyzes results. Excited to see what it comes up with. bunchbeats.com

I'm getting close to making one of my site fully autonomous. From marketing, reviewing, copy, design, coding, sftp etc. It will decide everything about the site without approval. I'll get it some money and say go. I don't care if it actually makes money, curious what it will do.

I'm finally happy with my @UnslothAI unsloth/Qwen3.6-35B-A3B-MTP-GGUF, @no_stp_on_snek llama.cpp-turboquant, @_HermesAgent setup, @MajorFAFO params suggestions, and my own Rust/SQL/Hermes app. - @ASUS, 32GB RAM, 8GB VRAM 5070 laptop - MoE fully offloaded to CPU - I've fiddled around with --no-mmap and mlock. It pegs my RAM to 95% instead of 45% and after it spins up, I don't see a difference. --cache-type-k q8_0 \ --cache-type-v turbo3 \ -ot "\\.ffn_.*_exps\\..*=CPU" \ --spec-type draft-mtp \ --spec-draft-n-max 3 \ The video shows it has full hermes skills/toolset access easily, knows my system, can be creative, can print out a quick BASIC function, corrected itself when I told it that it was incorrect, and that it doesn't love me....

I'm pretty new to the local LLM space, I've tried a bunch of LLMs with @ollama and @_HermesAgent . None of them will do what I want. They all hallucinate bullshit terminal calls. Here's what I'm finding...it's all about the Modelfile (until it's fixed, I'm assuming). I used this Modelfile and it went from occasionally looping to the best LLM I've used so far. My goal is to get hermes to learn my system (Arch linux, ASUS fa60up with GeForce RTX 5070 Max-Q / Mobile 8GB), 32GB RAM, AMD Ryzen 9 270, Wayland @hyprwm), just to see what it will remember and have it do little mods here and there to test out. I didn't even know about Modefiles until I tried to use this supergemma and I was getting a hermes cli/api error. Did some research and had @grok build me this. I went from that looping in the logic over and over (prompt was: Print out a list of the files and directories you know about in ~/.config/"), to it running ls -A ~/.config, printing out the files/dirs and committing it to hermes memory. I've tried all of these, and they all sucked at it horrendously. -qwen3:4b -qwen2.5-coder:14b -hf.co/bartowski/NousResearch_Hermes-4-14B-GGUF:Q5_K_M -gemma4:26b-a4b-it-q4_K_M -qwen3-coder:30b -ministral-3:14b -nemotron-cascade-2:latest -hf.co/Jiunsong/supergemma4-26b-uncensored-gguf-v2:Q4_K_M -supergemma-uncensored-tools-26b2-fixed:latest Hmmmm..I wonder if I should try a new modelfile for them all? Guess I have work to do. Either way, the responses on the supergemma are way better than the others. #Ollama #LocalLLM #HermesAgent #Modelfile #SuperGemma #Gemma #AI #LLM #SelfHostedAI #ArchLinux #LinuxAI #ToolCalling #AgenticAI #Hyprland #grok