Course
BASICModule 2: The AI Landscape· 2/2

Claude in Detail

Where Claude lives

One Anthropic account opens several entry points:

  • claude.ai — the web chat, the main interface. This is also home to Projects, Artifacts, file uploads, web search.
  • Desktop app (Windows/Mac) — the same functionality plus integration with local files and MCP servers on your machine.
  • Mobile app — a chat in your pocket, voice input, photos of documents straight into the conversation.
  • Claude Code — a terminal agent for working with files, code, and automation. The pro modules of the course are built around it.
  • API — access for scripts and your own programs, pay per token.

Chat history syncs across the web, desktop, and mobile versions.

The model family

Inside a subscription you switch between models of different classes. The logic for choosing:

  • Larger models (Opus and up) — complex reasoning, long documents, serious code, "think it through properly" tasks. Slower, limits burn faster.
  • Mid-range models (Sonnet) — the workhorse: quality close to the larger ones on most tasks, noticeably higher speed. A reasonable default.
  • Smaller models (Haiku) — instant answers to simple questions, high-volume operations via the API.

Rule: start with the mid-range model; if the answer falls short, repeat the request on the larger one. Paying with time and limits for the larger model on a "how is this word spelled" question makes no sense.

Key interface features

  • Files. PDFs, spreadsheets, images, code — drag into the chat, ask your question. Claude reads documents in full, including screenshots and photos.
  • Web search. For time-sensitive questions Claude searches the internet and answers with links to sources.
  • Artifacts. A separate panel where Claude creates documents, code, interactive pages, and diagrams — edited iteratively, not lost in the chat feed. We'll cover this in detail in a pro module.
  • Projects. Persistent workspaces with their own knowledge base and instructions. Also in a pro module.
  • Data analysis. Claude writes and runs code to process uploaded data — it counts honestly rather than "estimating" (remember the lesson on hallucinations and arithmetic).
💡

Voice mode and photo input in the mobile app are badly underrated: dictating a task on the go and photographing a document instead of retyping it are the fastest ways to "feed in" context.

Plans

  • Free — mid-range models, a tight message limit. Enough to get acquainted and handle occasional tasks.
  • Pro (~$20/mo) — larger models, Projects, limits several times higher. The entry point for regular work; enough for moderate use of Claude Code too.
  • Max (~$100–200/mo) — limits for intensive daily work, priority during peak hours. Justified when Claude Code becomes your main working tool.
  • API — separate pay per token, independent of the subscription. You'll need it in the automation module.

Subscription limits are rolling: use up your quota and you wait for the next window to start (a few hours). This matters for planning your work: it's better not to leave heavy tasks for the end of a quota.

Key takeaways

  • Entry points: web, desktop, mobile, Claude Code, API — history syncs.
  • Model choice: mid-range by default, larger for complex tasks, smaller for high-volume.
  • Interface strengths: files, search, Artifacts, Projects, honest data analysis via code.
  • Pro is the entry point for regular work, Max is for living in Claude Code, the API is for scripts.
CHECK YOURSELF
1. Which model is sensible to choose by default for everyday tasks?
2. You need to compute statistics over a table with 5,000 rows. What is more reliable?
A Map of ModelsAnatomy of a Good Prompt