A skeptic's practical experiment using a local LLM (Llama3.2 via Ollama) to solve a real problem: sentiment analysis for news data. The author, who runs a computational news analysis system on a Raspberry Pi, had long struggled to build effective sentiment analysis with traditional code. By crafting a simple prompt and querying a locally-hosted model, they achieved a working sentiment analyser with minimal effort. The takeaway is pragmatic: LLMs are just another tool, best used for tasks they're genuinely suited for — like contextual text analysis — rather than as a universal replacement for everything.
Table of contents
What is an LLM good at doing, and What Can it Do For Me?First, Find Your LLMIn Which I Become A Prompt EngineerWow! I Did Something Useful With It!Sort: