rag

AI agents, vector search, and isolated sandboxes

GPT-5.6 costs drop 80%; Claude context fees slash 97.6%

Multi-page OCR + agent orchestration updates

Terminal agents + 975B models: Weekly signals

Vector search lands in DynamoDB natively

Ollama switches to llama.cpp—faster local inference

Elixir LangChain & Agent Loop Patterns

Stacked PRs, LLM efficiency, multilingual RAG