Claude

2.1 Get a useful answer from Claude

Claude is an AI chatbot you run in your browser, a hosted end‑user app you simply open and prompt.

Receive a useful answer to an actual query on your first try

Type your question in the box at claude.ai and press Enter. Ask in any language you like — Claude answers in the language you write in. There's nothing to install.

1. Visit claude.ai and sign in using a Google account or email address
2. Enter your real question in the composer and press Enter
3. Add follow‑up prompts such as “go deeper on the example” or “I didn’t follow step 2” to continue the conversation

• You'll see A clear plain‑language response appears instantly and the chat window stays ready for further replies
• Takeaway Claude works as a simple browser conversation – sign in free, pose a real question, and keep refining the reply
• Check Why should you treat Claude's first reply as the start of a dialogue rather than a final answer, and what does that mean for how you respond to it?
• Cost Free plan: no credit card, with daily message limits that reset every few hours. Pro raises those limits and adds heavier features; Max goes higher still. Start free.

2.2 Create precise prompts for Claude

Claude is an AI chatbot accessed through a web interface, ready‑to‑use without building custom agents.

Do this first Get a useful answer from Claude

Get sharper answers by giving context, assigning a role, specifying the output format and iterating instead of restarting

Anthropic's own guidance: wrap distinct parts of a prompt in simple tags like <context> and <task> so Claude can tell instructions from data. Setting a role ("You are a careful biostatistics reviewer") focuses its tone and rigor.

1. Provide context and assign a role, e.g., “You are a careful biostatistics reviewer” followed by relevant background
2. Request a specific output format such as a numbered checklist, bullet points or a short paragraph
3. Wrap each part of the prompt in tags like , and to keep sections distinct
4. If the response misses something, reply with a targeted correction like “tighten point 3” instead of re‑asking

• You'll see A focused, correctly‑formatted answer that updates when you request small corrections
• Takeaway Define the persona, supply background, specify output shape and use tags, then refine by replying rather than starting over
• Check Why does Anthropic recommend wrapping distinct parts of a prompt in tags like and , and what problem does it prevent?
• Cost Free on every plan — prompt-craft costs nothing and is the single biggest lever on answer quality, on the free tier as much as on Pro.

2.3 Get answers from PDFs, CSVs and images

Claude is an AI chatbot you interact with through its hosted web interface, not a development platform you code against.

Do this first Create precise prompts for Claude

Upload a document or picture and receive answers that reference its actual contents

Click the attachment (paperclip / "Add files or photos") button below the message box, attach the file, then ask your question. Asking Claude to QUOTE the exact sentence keeps it honest and lets you verify against the source.

1. Attach a file using the attachment button (PDF, Word, TXT, CSV or PNG/JPG/GIF/WebP)
2. Ask a question about the content, such as “Which statistical test did they use?” or “What does this gel image show?”
3. Request an exact quote by adding “quote the exact sentence you’re basing that on”

• You'll see Claude returns answers that include exact quotes and data extracted from the attached file
• Takeaway Attach files instead of copy‑pasting and always ask for source quotations to verify claims
• Check When Claude summarizes a document you've uploaded, why should you explicitly ask it to quote the exact sentence it's basing each claim on?
• Cost File upload works on the free plan. Limits to know: up to ~500MB and 20 files per chat; Claude reads text and visuals in PDFs under ~100 pages, and text-only for very long PDFs. Heavy use hits the free daily cap faster.

2.4 Get up‑to‑date answers with source links

Claude is an AI chatbot you access through a web interface, delivering conversational answers directly to users.

Do this first Get answers from PDFs, CSVs and images

Receive current responses that include citations and verify at least one source yourself

When a question needs current facts, Claude can search the web and cite what it found. The trick: explicitly ask for the source links and ask it to flag disagreement — then click through and read at least one source yourself.

1. Type your question into the chat input field and press Send
2. Add “Give me the source links” to request citations
3. Click a citation link to open the source
4. Include “flag where sources disagree” to highlight contested points

• You'll see A reply showing clickable citation pills and, if requested, a note marking where sources disagree
• Takeaway Checking a cited source turns a confident guess into a verified fact
• Check Why does asking Claude to flag where sources disagree give you a more trustworthy research answer than asking for a clean summary?
• Cost Web search is available on the free plan. It's most valuable exactly when your own knowledge (or Claude's) might be out of date — current literature, evolving methods, recent tools.

2.5 Build an editable Artifact beside the chat

Claude is an AI assistant accessed as a hosted chatbot in your browser, generating editable Artifacts beside the chat.

Do this first Get up‑to‑date answers with source links

Create a document, diagram or mini‑tool that appears in its own side panel and can be refined through conversation

When Claude builds something substantial — a diagram, a document, a chart, a snippet of working code — it opens an Artifact in a side panel. You then edit it by replying ("add a 'batch correction' step before normalization") instead of regenerating from scratch.

123 Enlarge

1. 1 Current visibility public link, chat stays private
2. 2 Remove public access artifact becomes private and cannot be republished
3. 3 Embed artifact snippet allows insertion into external pages Why this exists →

Best viewed on desktop — tap Enlarge to read the numbered controls.

1. Enter a prompt asking Claude to generate a flowchart, protocol, table, SVG figure or interactive calculator
2. Refine the output by typing instructions such as “make the boxes blue” or “add a QC failure branch”
3. Click Copy to copy the Artifact to the clipboard
4. Click Download to save the Artifact locally
5. Click Share to obtain a public link for the Artifact

• You'll see A live flowchart (or document, chart, or calculator) displayed in a side panel, updating as you edit it and ready to copy, download or share
• Takeaway Ask for an Artifact to turn text requests into usable, editable outputs that you can shape by talking to the model
• Check What is the key difference between Claude's Artifacts and ordinary text replies, and why does that difference matter when you need to iterate on a deliverable?
• Cost Artifacts are available on all plans, including free; to run code or generate downloadable files you may need to turn on "Code execution and file creation" in Settings. Heavy generating eats the free daily limit faster. Note: Claude does not generate photographic images — it makes diagrams, SVGs, and code, not AI photos.

2.6 Create a Project that provides reusable context for Claude chats

Claude is an AI chatbot accessed through its web interface, designed for direct conversation and task assistance.

Do this first Build an editable Artifact beside the chat

Start each new Claude chat with your uploaded files, style instructions and ongoing work already loaded

Open Projects in the left sidebar, create one, and add a knowledge base (upload your key papers/protocols) plus project instructions. Every chat inside that project inherits the documents and the instructions automatically.

12 Enlarge

1. 1 Project visibility private to you, not the team Why this exists →
2. 2 Project settings button modify name, description, privacy Why this exists →

Best viewed on desktop — tap Enlarge to read the numbered controls.

1. Click Create project in the sidebar, give it a name and upload the relevant documents to its knowledge base
2. Open Project settings, add instruction text such as “prefer British spelling” or “always cite sources”
3. In Settings → Profile Instructions, add any account‑wide preferences you want applied to all projects
4. Go to Settings → Capabilities and toggle Memory on to let Claude build a running summary that persists per Project

• You'll see A Project where every new conversation automatically knows your uploaded files, preferred style and ongoing work
• Takeaway Define context once in a Project and let Claude’s memory retain it for future chats
• Check What is the difference between project instructions and the memory feature in Claude, and when would you want to use each one?
• Cost Projects, project/profile instructions, and memory are all on the free plan; free accounts get up to 5 Projects, while Pro/Max give unlimited Projects and a much larger knowledge-base capacity. (Searching across all your past chats — Chat Search — is a paid feature.)

2.7 Analyse a research paper with Claude

Claude is a conversational chatbot accessed through a web interface, not a development platform you code against.

Do this first Create a Project that provides reusable context for Claude chats

Run Claude over an uploaded paper, have it reason step‑by‑step, verify its conclusions and reveal the strongest reviewer objection

The trick that lifts quality on hard analysis: ask Claude to reason step by step AND to verify its own answer against the source before finishing. Combined with file upload (Lesson 2) and quoting, this turns a summary into a genuine critical read.

1. Upload the PDF by clicking Upload the paper and provide Claude with the numbered question set for a structured review
2. Prompt Claude to reason by selecting Ask it to reason, using phrases such as “reason step by step” and “re‑check your reasoning against the actual text before you conclude”
3. Challenge Claude with a reviewer stance by clicking Pressure-test and asking “what’s the single strongest objection?”, then verify any quoted passages in the paper yourself

• You'll see A structured, quote‑backed critique showing claims, method gaps, statistical red flags and a hard reviewer objection
• Takeaway Combine uploading, stepwise reasoning and self‑verification to produce a defensible critical read
• Check Why does asking Claude to re‑check its reasoning against the actual text and correct anything it got wrong before it finishes improve the quality of a critical paper analysis?
• Cost Everything here runs on the free plan; a long, multi-question deep-read will use up the free daily message limit quickly, where Pro/Max simply let you keep going. Careful step-by-step reasoning works best on Claude's strongest models, available across the plans.