Index by capability — RAG, tools, memory & more

The system looks things up in a library of documents and feeds the relevant pages to the model before it answers.

The model can take actions in the real world — send an email, run code, update a record — by calling tools.

The assistant remembers things about you between conversations, so future chats start smarter.

A behind-the-scenes manager that loops — think, act, look at the result, think again — until a goal is done.

Several AIs work as a team — a manager hands subtasks to specialist workers and combines their results.

The AI is given a screen, mouse and keyboard — it looks at the display and clicks and types like a person to get things done.

Screens that check messages going in and answers coming out, blocking obvious abuse, leaks, or banned content.

The rules about what the AI is allowed to do — which keys it gets, and when a human must approve.

The model itself and the machinery that runs it — turning text into numbers, predicting words, and serving answers at scale.

Where the model's weights, packages and tools come from — public hubs and registries you download and trust before anything runs.

The flight recorder and the dashboards — a record of everything that happened, plus alarms that watch for trouble.

Where you talk to the AI, and where the app quietly bundles its instructions together with your message and any documents.