One curriculum. Every role runs through it.
From "what's a token?" to a working, evaluated agent.
Live now
Foundations
Zero prior knowledge assumed. Tokens, LLMs, prompting, embeddings, vector databases, RAG, and your first agent.
Live nowIntermediate
Chunking, retrieval quality, tool use, memory, and your first real agent, ending in a multi-tool capstone.
Live nowAdvanced
Multi-agent patterns, guardrails, observability, and shipping a traced, evaluated agentic RAG system.
What's new
Latest from Few-Shot Academy
2026-09-13Agent security starts before the first promptTool allowlists help when a model makes a bad decision. They cannot stop startup hooks or an exposed gateway from acting outside the model, so those paths need their own boundaries.2026-09-07AGENTS.md: one guide for your coding agentsWhat AGENTS.md does, how to write a useful one, and which coding tools load it automatically. Includes an example and setup for Claude Code and Gemini CLI.2026-09-04GPT-6 Astra: the harness is the productGPT-6 Astra scored 54.8% or 99.9% on the same benchmark depending on the harness. That gap matters more than the AGI headline.