He Uses Local LLMs Like a 3D Printer: 12 Games, 29 Mods and Countless Tools Built Solo
Quebber · reddit · 2026-09-05
Running an uncensored Qwen3.8 27B (Q8, full 256k context) on a Minisforum MS-S1 395+ Max with 128GB unified memory (96GB allotted to VRAM), the poster treats local LLMs like a 3D printer: when he lacks a piece of software, he spends a few hours building it with his custom agent framework.
Actual output so far: 12 adult games, a house AI, a game tracker that auto-fetches FAQs/wikis, 17 Skyrim mods plus 12 for Fallout New Vegas, an RSS-fed house temperature tracking system with heating suggestions, and mapping software for his mobility scooter that checks routes for construction blocking pavements — plus hundreds of small tools. While writing the post, his framework finished a hook that OCRs Japanese visual novels and translates them via a local LLM.
His core point: he used to lazily buy or download software; now local agents make "just build it" the first instinct, like a 3D printer changed how people acquire physical objects. He asks if others use local models this way.
More from coding & agent
- Skip the Chain-of-Thought: Monitor Agent Inputs and Outputs Instead — nicolascraske · 2026-09-05
- Voice agent testing tools compared: Cekura, Cyara and TestMu solve different problems — Fishful_Revenge · 2026-09-05
- LinkedIn VP Breaks Down Hiring Assistant: The Problem Is Fragmented Recruiting Workflows — CodeByPoonam · 2026-09-05
- Agent frameworks are 10% of the problem—state hygiene is what breaks production systems — Deepfeet-09 · 2026-09-05
- Dev workflow: iterate plans with big models, then hand off execution to small reasoning models — intellectronica · 2026-09-05
- Burn Bar for Omarchy visualizes Claude/Codex token burn, quotas and GPU load locally — DanWahlin · 2026-09-05