Why a weekly AI glossary
AI vocabulary changes faster than almost any other field. A term coined in a research paper can be on every product page within weeks, and the gap between hearing a word and understanding it is where confusion — and bad decisions — live. This glossary is a rolling page that explains the newest AI terms entering mainstream usage, in plain English, while they are still unfamiliar. The goal is not to be exhaustive like a full dictionary, but to be current: to catch the words you have started seeing everywhere and tell you what they actually mean, why people are talking about them, and whether they describe something genuinely new or just rebrand an older idea.
Recent terms worth knowing
Agentic AI describes systems that do more than answer a single question — they plan, call tools, observe results, and take multiple steps toward a goal. The shift from “chatbot” to “agent” is one of the biggest framing changes in the field.
Test-time compute (or inference-time reasoning) refers to letting a model spend more computation while answering — generating and checking intermediate reasoning — rather than only making the model bigger during training. It is the idea behind reasoning-focused models that “think” before they respond.
Mixture of Experts (MoE) is an architecture where a large model is split into many specialised sub-networks, and only a few are activated for any given input. This gives the capacity of a huge model at a fraction of the running cost.
Context engineering is the emerging discipline of deciding what information to put in a model’s context window, and how to structure it — a superset of prompt engineering that includes retrieval, memory, and tool results.
How to use these definitions
Treat the glossary as a critical-reading aid, not just a vocabulary list. When a new term appears in a headline or a sales deck, look it up here and ask the key question: does this describe a concrete mechanism, or is it a vibe? Real terms point to something you can verify — a way models are built, a measurable change in behaviour, a specific technique. Marketing terms tend to be vague and superlative. Knowing the difference lets you read AI news the way a practitioner does, separating signal from noise. As new words break through, this page adds them — so check back, and pair it with a fuller A-to-Z glossary when you want the established fundamentals rather than the latest arrivals.