AI 101 — The AI Glossary: every AI term, plainly explained
The canonical, plain-English reference for every term in modern AI — from tokens to agents, RAG to open-weight models. 27 entries, all sourced, all updated.
AI 101 — The AI Glossary
AI Midday's free, plain-English reference for the most-asked terms in modern artificial intelligence. Every entry is short, sourced, and updated regularly — written for journalists, investors, and engineers who need the right definition fast.
Whether you want a one-line answer to “what is a token?” or a fuller read on fine-tuning, RAG, agents, or open-weight models, this is the page to bookmark. Each term links to a dedicated explainer with a working example, the math when it helps, and the trade-offs you should actually care about.
Browse by topic
Models & training
- What is test-time training?
- What is fine-tuning?
- What is synthetic data?
- What are reasoning tokens?
- What is model quantization?
- What is knowledge distillation?
- What is continual learning?
- What is RLHF?
- What is a mixture of experts?
- What is a token in AI?
- What is an AI benchmark?
- What is speculative decoding?
- What is an embedding?
Agents & infrastructure
- What is chain of thought?
- What is a context window?
- What are open-weight models?
- What is RAG?
- What is MCP?
- What is an AI agent?
- What is a vector database?
Safety & policy
- What is prompt injection?
- What is AI watermarking?
- What is an AI hallucination?
- What is a learned kernel?
- What is a sandbox escape?
- What is AI regulation?
More terms
Why this page exists
AI terminology moves fast. New jargon — context windows, MCP, open-weight models, sandbox escape — shows up in headlines before the previous cycle has explained it. This glossary is where we keep the canonical answer.
If a term is missing, tell us and we'll write it.
