# Claude Chat Settings What are "AI-powered artifacts," and what does enabling them do? # Using Claude If I want to brainstorm about and plan a large coding project, should I switch over to Claude Code for that brainstorming and discussing? Or remain in the normal chat, and then later switch to Claude Code when the programming begins? >**Chat is better for open-ended exploration.** Early brainstorming — "should this be a monolith or microservices," "what's the right data model," "which language/framework fits" — is a conversation, not a coding task. Regular chat gives you a clean back-and-forth without the overhead of a project directory, file tools, or execution context that Claude Code sets up around every response. >**Claude Code shines once there's a real codebase to act on.** Its value is in reading/writing files, running commands, checking things against an actual repo. If there's no code yet, most of that machinery is idle. >**You can carry the plan over.** Once you've landed on an architecture or rough spec in chat, you can paste or summarize it as the seed for a `CLAUDE.md` or an initial prompt in Claude Code, so you're not starting from scratch. --- What exactly are skills, and how could I use them to benefit me as I full-time IT worker at a university that hopes to improve my skills and find a better and higher-paying job, a passionate musician working on original music projects, and an aspiring content creator that wishes to run a channel for long-form political essays with accompanying written pieces posted on my own website, and another channel for posting videos about metal music and of myself playing guitar, drums, and performing vocals, as well as covering other bands' music?