Claude

Most of the features below are on the Pro plan or above. If you only have free Claude, treat this as a map of where the road goes.

Claude Chat, Cowork, and Code

Claude comes in three modes, all powered by the same model. If you get confused about which to reach for, this tends to be a good way to think about it: if it only takes a minute or two, go with chat. If you want to write code or build an app, use Code. For everything else, Cowork is usually the place to go.

Chat is the default mode and the one most people spend most of their Claude time in. You get it on claude.ai and inside the Claude desktop app for macOS and Windows.

Cowork is the agentic mode inside the Claude desktop app. You point it at a folder, give it a goal, and it plans, executes, and lands a real deliverable: a working spreadsheet, a slide deck, a sorted folder of files. The personal productivity track in Level 3 leans on Cowork heavily; this course covers the moves so you can already pick it up today.

Claude Code runs in your terminal, and now also inside the Claude desktop app, pointed at a project folder. It can read the repo, run commands, edit files, and check its work. The right mode when the task is changing code, investigating a log, or building a small tool. We go deeper on Claude Code in the Level 3 tracks (vibe coding and building your own tools); for now, just know it exists.

Projects

A Project is a persistent workspace with its own knowledge base, custom instructions, and chat history. When you have a topic you keep coming back to with the same background docs, the same writing voice, the same set of references, that is a Project. You upload the material once, write the instructions once, and every chat inside the Project starts with all of it loaded.

Free plans give you up to five Projects; paid plans are unlimited and add more knowledge-base capacity. Drop in PDFs, Docs, code files, transcripts, anything Claude should always have on hand for that topic. Set a one-line instruction for tone or format. From then on, Project chats are pre-loaded, and you skip the “let me re-explain everything” step.

Artifacts: from text in a chat to a file you can use

Artifacts is the side panel where Claude builds the actual deliverable. A doc, a deck outline, a small web app, an SVG diagram, a spreadsheet. The chat stays on the left, the artifact renders on the right, and you iterate by talking to it (“add a date filter”, “make the headline shorter”, “swap to a dark theme”). The mental shift is to stop thinking of Claude as a chat that writes code and start thinking of Artifacts as the place small tools and deliverables get built.

Building is cheap, maintaining is not. Build a lot, keep the two or three you reach for a second time, throw the rest away without regret.

Imagine with Claude: “show me” visuals

Sometimes a diagram lands faster than a paragraph. Claude’s visuals mode (officially “Imagine with Claude”, currently in beta on web and desktop) lets you ask for an inline chart, diagram, or interactive visual right inside the conversation. Trigger it with phrases like “show me”, “draw this”, “can I see that”, or “visualize this”.

This is different from Artifacts. Artifacts are persistent, downloadable, shareable deliverables that open in the side panel. Imagine visuals are temporary and live inline with the rest of the chat: a quick illustration to make a point, not a file to ship. Reach for it when you want to understand something faster, not produce a deliverable.

Memory

Claude remembers context across chats by default now. What you are working on, how you like things, who your team is. You can also tell Claude directly to remember something (“remember that I always want responses in plain language”, “remember that we are migrating to PostgreSQL”). The instruction lands in the memory summary and applies to your next conversation. You can ask Claude to read back what it remembers, edit it, clear it, or turn the whole thing off.

Treat it like a colleague who takes notes on what you said: useful, but check what they wrote down before they act on it. The “you are responsible” rule applies here too, because the model will quietly act on a stale fact unless you correct it.

Research mode

Research mode is Claude’s long-running web investigation. Standard Research takes minutes and returns a cited brief. Advanced Research runs autonomously for up to forty-five minutes across hundreds of sources, the kind of question that would normally cost you most of an afternoon. The discipline is the same as for any agentic task: scope it tightly, edit the plan before approving, spot-check the citations when the report lands.

Cowork

Cowork’s range is wider than most people expect. A few of the moves that make it worth opening:

• Real deliverables, not just text. Cowork produces working .xlsx with formulas, .pptx decks, formatted Word docs and PDFs. The output lands as a file you can send, not as a chat message you reformat.
• File-system work. Point it at a messy folder and ask it to rename, sort, deduplicate, or pull together the three documents that actually matter for the meeting tomorrow.
• Plan, then execute. For anything multi-step, Cowork shows you the plan first and waits for approval. You can redirect at any step. The same posture as the principles: start with a plan, keep the human in the loop on consequential moves.
• Live Artifacts. A dashboard or tracker that connects to your apps and stays current, so reopening it tomorrow pulls fresh data instead of a snapshot.
• Scheduled tasks. Save a task (“compile the weekly hiring report from these sources”) and let Cowork run it on a cadence: every Monday morning, the report is waiting.
• Computer use, as a fallback. If a job has no API or connector path, Cowork can drive your browser or screen directly. Slower and less precise than a real integration, so it reaches for that last.

Skills

A Claude Skill is a packaged workflow your agent picks up by name when the task matches. Same idea as Gemini Gems and ChatGPT Custom GPTs, with the most explicit format. Write the rules once (your team’s writing voice, your status-update format, your release checklist), name the skill, save it. Skills work across all three modes (Chat, Cowork, Code), and the Anthropic skill marketplace ships official ones for common tasks (slides, spreadsheets, PDFs, brand kits).

The next course in this track goes deep on writing your first one.

Hands-on

Build me an interactive weekly hiring tracker from the attached CSV: filters by team, a chart of offers vs starts, and a card with the conversion rate. Dark theme.

Reflect

• Which of the three modes are you under-using, given the work you actually do? If you live in chat but most of your week is wrangling files and decks, the friction is the mode, not the model.
• Pick a topic you keep coming back to in chat (a customer, a deal, a research area). Could that whole topic be a Project, with the docs uploaded once and the chats anchored to them every time?