What AI is — and what it is not

Start with the job, not the product

People say “AI” when they mean several different jobs: answering a question, drafting text, labeling a photo, routing an email, or running a program that calls other programs. The product name on a homepage is not the job.

A useful first split is this: a **model** is the statistical engine. An **assistant** is a product wrapped around a model (chat, memory, tools, billing). An **agent** is a setup where software can take a sequence of actions, not only return a paragraph.

Pattern engines, not minds

The systems you meet in 2026 chat products are trained to continue text, or to map inputs (pixels, audio, tables) to outputs. They do not know your company. They do not inspect the physical world unless some other system feeds them data.

That is why a fluent answer can still be wrong. Fluency is the training objective. Truth is optional unless you check it.

What this site will not do

LearnLumin will not tell you that a vendor is “the best,” quote a price we did not just read from a labeled source, or treat a press headline as a capability. Tools and models change. The lesson underneath — what kind of system you are using — changes much more slowly.