Redditor runs a local agent fleet to mine terabytes of personal books and archives
GunmetalZen · reddit · 2026-10-21
A non-coder Redditor shares a hybrid setup: frontier models build the infrastructure once (with full docs), then local models run the ongoing work.
- Hardware: 2x Strix Halo boxes running GPT-OSS 120B Q4, plus an RTX 5090 running Gemma 31B Q4.
- Infrastructure: a NAS library with a Wikipedia mirror, a personal "Google Earth" with geo-coordinated points of interest (extended to the Moon and Mars), a StackExchange replica, hundreds of thousands of books/reports/magazines, and galleries of historical photos and museum-quality art.
- Ongoing work: 3 local agents run 24/7 mining terabytes of material for images, quotes, factoids and cross-work linkages to decide what's worth reading; one agent stays on standby to answer questions about the library, reachable via Signal with self-hosted SearXNG search — fully local, no external tracking.
His takeaway: local models are solid for tool use, file ops and troubleshooting; only major coding needs frontier help.
More from coding & agent
- Why agents need an 'Agent GitHub' for versioning, permissions, and memory — yangyi · 2026-10-21
- Kapitsa 0.2's KFlash tree speculative decoding beats Splash 1.4.0 with 63 tok/s on Qwen 27B — DotaMate · 2026-10-21
- Dev predicts 2027 will be the most fun year as vibe-coded games take off — TAbrodi · 2026-10-21
- Hamel Husain: public benchmark quality is abysmal — broken graders, sloppy harnesses — HamelHusain · 2026-10-21
- Dev jokes Claude is never blocked from making mistakes, only from fixing them — cjimti · 2026-10-21
- Compile Your LLM Workflow Into Small Model Weights: A LoRA Hypothesis Worth Testing — arpit_bhayani · 2026-10-21