EXTENDED COURSE

Claude

Claude in your browser, from first sign-in to a real research workflow — no install, just a conversation. This hands-on course starts on the free tier: ask better questions with context and structure, upload PDFs, CSVs and images to work with your own files, get cited web answers you can verify, and build live Artifacts you can edit and share. Then go deeper — organize recurring work into Projects with their own documents, instructions and memory, and finish by deep-reading a real paper end-to-end, asking Claude to reason carefully and check itself.

1Lessons7hands-on, ~15–25 min each

Get started

Work like a pro

Go deep

06 Capstone: deep-read a real paper

2Examples9what people built

Internal tools & ops1

Claude Scientist AI-native SME

Course starter

Keep a running AI-contribution log per project

Ask Claude to help you create a structured markdown log — one entry per session — recording which task was delegated, what the model produced, and what human verification was done. Commit the log alongside your project files.

If a co-author, editor, or auditor asks "where exactly did AI help?", you have a traceable record rather than a vague recollection.

Try it yourself

Prompt: "Create a template for an AI-use log with columns: Date, Task, Model, Output summary, Verification step taken."

Content & marketing1

Claude Founder Creator

Course starter

Draft a one-pager you can edit live

Ask Claude to write a one-pager or landing copy as an Artifact — a live document beside the chat you keep refining by asking for changes, not copy-pasting.

A finished, editable document that evolves with the conversation.

Try it yourself

"Draft a one-page product overview as an artifact: problem, solution, who it's for, 3 benefits."

Research & data tools1

Claude Scientist AI-native SME

Course starter

Turn off model-training before pasting unpublished data

Before pasting any unpublished results, patient measurements, or proprietary formulas into Claude, navigate to Claude.ai → Settings → Privacy → and confirm that "Improve Anthropic's products" is off (free tier) or use the API with no retention by default.

Your unpublished data is not used to train future model versions, protecting both research priority and participant confidentiality.

Try it yourself

Check Claude's current model card or privacy settings page for the exact toggle name — it changes between product versions. The API has no human-review data retention by default.

Knowledge & docs6

Claude Scientist

Course starter

Deep read of a long paper

Upload a 30-page PDF to Claude and ask it to extract the methods, list every assumption, and flag where the conclusions outrun the data.

A structured critique of a long document in minutes, holding the whole paper in context.

Try it yourself

Drag in a PDF → "Summarize the method, list assumptions, and note where claims exceed the evidence."

Claude Scientist

Course starter

Rough notes into a methods section

Give Claude your bullet-point protocol and have it draft a clean, journal-style methods section — then tighten the wording in follow-ups until it matches your voice.

A solid first draft of the hardest section to write, grounded in exactly what you did.

Try it yourself

"Here are my steps as bullets. Draft a methods section in past tense, passive voice, ~250 words."

Claude Scientist Founder

In the gallery

A tough first reviewer for your draft

Paste an abstract, grant paragraph, or pitch and ask Claude to critique it like a skeptical reviewer — unclear claims, missing evidence, weak logic — before a human sees it.

The obvious objections caught early, in private.

Try it yourself

"Review this abstract as a critical reviewer. List the three weakest claims and how to fix each."

Claude AI-native SME Small biz

Course starter

A Project that knows your standard context

Put your style guide, key documents, and background into a Claude Project once, so every chat in it answers with your context already loaded — no re-pasting.

Consistent, on-context answers across many sessions from one shared workspace.

Try it yourself

Create a Project, add 2–3 reference docs, then ask a question that depends on them.

Claude Scientist

Course starter

Write an AI-use disclosure for your methods section

Ask Claude to draft a one-paragraph methods disclosure stating which tasks used AI assistance, which model was used, and that all outputs were verified by the authors — ready to paste into a manuscript.

Readers and reviewers know exactly where AI was involved, satisfying most journal integrity policies without guessing at the right wording.

Try it yourself

Prompt: "Draft a transparent AI-use statement for a methods section. I used Claude to help structure the literature review and draft the discussion. I verified all claims against primary sources. Include the model name and that no data were shared."

Claude Scientist Creator

In the gallery

Have Claude critique its own draft before you accept it

After Claude produces a draft section, send a follow-up: "List every factual claim in the above draft that you are less than fully confident about, and flag where a reader should verify against a primary source."

You catch hallucinated statistics and unsupported assertions before they enter your manuscript, instead of relying on Claude's confident tone as a proxy for accuracy.

Try it yourself

Treat the critique list as a mandatory checklist — do not finalise the section until each flagged claim is verified in a primary source you can cite.

How is the Claude course so far?