An AI system that doesn’t just answer a question, but takes a series of actions on its own to complete a task — like booking a flight, writing and running code, or managing a to-do list — usually with the ability to use outside tools along the way.

In practiceYou’ll see this term when companies talk about AI that “does the work,” not just AI that “answers questions.”

The amount of text an AI model can “see” and consider at once when generating a response — including your question, any documents you’ve shared, and the earlier parts of the conversation. Once a conversation grows beyond this limit, the model starts losing track of the earliest parts.

In practiceA “longer context window” means a model can work with bigger documents or longer conversations without forgetting the beginning.

When an AI system states something false with full confidence, as if it were fact — a made-up statistic, a fake source, a wrong date. It happens because the model is predicting plausible-sounding text, not consulting a database of verified truths.

In practiceAlways double-check names, numbers, and sources an AI gives you, especially for anything published or decision-critical.