AI terms, explained the way I actually use them
No textbook, no consultant pitch. Every entry is a short, citable definition plus a paragraph from practice — what the term actually means in a solo marketing operator's daily work, and when it becomes relevant for your business.
The base model applications build on — and why models are becoming interchangeable.
Read → SovereigntyModel weights you can download and self-host. Not the same as open source.
Read → AgencyAI operates browsers and applications like a person — no interface required.
Read → GovernanceThe stopping point where a named person must approve before the flow continues.
Read → CraftBuilding instructions so the output is reproducibly usable — instead of correcting it five times.
Read → ArchitecturePursues a goal, uses tools, takes several steps. What separates it from a chatbot is the ability to act.
Read → RiskWhen a model invents something and delivers it in the same confident tone as an established fact.
Read → FoundationLarge language model: predicts the most likely continuation — everything else follows from that.
Read → Law & transparencyAI-generated media that deceptively recreates existing people or events — with a disclosure duty in force since 2 August 2026.
Read → ConceptAI systems that independently plan and execute multi-step tasks — instead of just responding to a single prompt.
Read → ArchitectureRetrieval-Augmented Generation: a language model fetches facts from an external source before it answers.
Read → ProtocolAn open standard that lets AI models access tools, data and systems in a uniform way — without a custom adapter for every app.
Read →From "prompt engineering" to "fine-tuning" to "hallucination" — the rest of the glossary follows week by week. Suggestions? Get in touch.