AI, in plain language.

A broad category of systems that perform tasks such as recognizing patterns, generating language, or supporting decisions.

AI that produces text, images, or other content from instructions and context. Its output can be useful and still wrong.

A trained model that works with language. The model is not the same thing as the app or chat interface.

The instruction or question you send. In this course, a useful prompt names the audience, goal, sources, output, and limits.

The messages, instructions, and materials available for the current response. Naming a file does not prove the model can read it.

The original material used to support or check a claim, identified here by a filename and source ID.

What a supplied source directly records. An observation should not be expanded into a cause or frequency it does not establish.

A possible meaning drawn from observations. It should be labeled and checked, not presented as a direct finding.

Something treated as true without direct support from the supplied evidence. Make it visible before relying on it.

A proposed change or next action that still needs evidence from a test. It is not a finding.

A false or unsupported generated claim that may sound convincing. Check it against the original source.

A check against an independent source, observable state, or acceptance criterion. Asking the same model again is not independent verification.

Reusable instructions and supporting material for a recurring workflow. A skill does not grant account or file access.

A package that can include skills and connected apps. Installation does not bypass authorization or workspace permissions.

Authorized access to information or actions in another service, limited by the current account, plan, workspace, and permissions.

A person reviews, corrects, or approves important work instead of treating model output as final.

OpenAI’s AI assistant, available through web and desktop interfaces. The exact tools and controls can differ by account and app.

Anthropic’s AI assistant. The same prompting and checking habits apply, while its tools and installation steps can differ.

OpenAI’s agent for working with code, files, and tools. Its available actions depend on the environment and permissions you provide.

Model Context Protocol, a standard for connecting AI applications to tools and information. A connection still needs the appropriate access permissions.