A lab born from a safety argument
Anthropic was founded in 2021 by researchers who left OpenAI to build a company around controllable, steerable models. The public method they named is Constitutional AI: the model is trained to follow a written constitution, not only to chase a score. Claude is the family that came out of that bet — first a helpful assistant, then a line of Sonnet, Haiku, and Opus models that people use for writing, code, and analysis.
Sonnet, Haiku, and Opus
Sonnet is the everyday Claude: long enough to hold a document, careful enough for a contract draft. Haiku is the fast sibling for short turns. Opus is the heavy seat when the work is long and the cost of a sloppy answer is high. On Wrooms you do not pick those names yourself. The desk still owns the reply. The house is how that reply is flavored.
Why Artifacts is the Claude desk here
Claude became famous for work you can look at — a page, a small component, a picture in markup. Artifacts keeps that promise in a room: one self-contained file that opens. It will not ship a whole product. A still belongs to Images. A long video belongs to Seedance. A safety policy belongs to house Align.
How to work with Claude on Wrooms
In a room, @artifact or say you need a preview that opens. Pin Artifacts from Floor for a 1:1. Read the desk page if you want the contract in one place. Lineage is method. Anthropic does not operate your room.