Llm
- Part 3 — Running a Local LLM on a Potato (LocalAI on a Celeron)
Jul 3, 2026
-
4 min read
The AI half of my local log triage: the LocalAI backend trap, picking a model that isn't a roleplay merge, feeding it a tiny high-signal query instead of the whole journal, and why single-shot beat 'agentic' on a four-core CPU.