The AI Landscape, Briefly
Narrow, generative, agentic: three different things get called AI. Knowing which one you're using sets what to expect from it.
Time to read: about 1 minute
This is the light version. A deeper look at specific vendors and models, worth doing before you commit budget to one, comes later in Module 7. This reading is just the map, three kinds of AI, and one honest way to compare models yourself before you take anyone's word for which is best, including this program's.
Three words, so we mean the same thing
Built for one task
Bounded by design. A spam filter, a recommendation engine. It does not generalize past its job.
Produces new content
From a prompt. What most of this program means when it just says "AI."
Takes multi-step action
Toward a goal, with limited supervision. Module 4's whole subject.
One prompt, two models, no names
Arena (arena.ai, renamed from Chatbot Arena and LMArena in January 2026; the old lmarena.ai address now redirects there) runs one prompt through two anonymous models at once, side by side. You vote for the better answer without knowing which model wrote it. Millions of those blind votes build a public leaderboard, but the leaderboard is not really the point for you. The habit of checking is.
This is the cheapest way to learn whether you prefer the model you already use or only the one you are used to.
Take one question from your own work this week. Run it through Arena. Notice which answer you pick, and whether it is the model you already use. Either result teaches you something.
Module 7 covers the deeper evaluation: specific vendors, specific tradeoffs, and what to adopt.