A large language model (LLM) is a model trained on large amounts of text. It can answer questions, summarize, translate, classify and write code. On its own it knows only what it has learned; with retrieval it answers from a given body of knowledge, usually via embeddings: vectors that let you compare text by meaning. With tools it can act. This topic collects how language models are used, from a script that calls an API to agents and knowledge bases they help maintain.
Subtopics
- AI Agents AI apps whose language model calls tools to get work done, such as Claude Code and Claude Desktop, and how to run them safely.
- Model Context Protocol Open protocol through which AI apps such as Claude Code and Claude Desktop connect to servers that offer them tools and data.
- On-Device AI AI models that run on the user's own device, in an app or in the browser, so the input stays there and no server does the work.
- Retrieval-Augmented Generation Answering questions with an LLM by first retrieving relevant passages from a knowledge store and giving them to the model as context.
- Covered in
-
- Measuring the Search of My Knowledge Base Five local embedding models against 26 test questions: why the search now runs on BGE-M3 instead of multilingual-e5-small, with a new minimum score.
- Photo Sidecars with a Vision Model The script vision.ts of my old site: EXIF data and a vision model's alt text, titles and tags in a JSON file next to each photo, generated locally before the deploy.
- RSS News with n8n and OpenRouter Guide: n8n reads security notices, a language model rates and rewrites them, and a restricted SSH script commits them to the Git repository of a static site.
- Safari Bookmark Sorter Python script that sorts Safari bookmarks into folders with Claude Haiku and hands the result to Safari as an import file, so iCloud syncs it.
- See also
-
- LLM Wiki Andrej Karpathy's pattern for personal knowledge: an LLM agent reads each source once and works it into a growing set of interlinked notes.
- Tags