LLM | Haobin Tan

Claude 101

Tue, 02 Jun 2026 00:00:00 +0000

Meet Claude

Tue, 02 Jun 2026 00:00:00 +0000

TL;DR

What is Claude: An AI assistant designed to be a steerable, collaborative thinking partner.
Claude’s capabilities: Writing and content creation, research and analysis, coding assistance, problem-solving and reasoning, Learning new things
Access: Available via Claude.ai, Claude Code, Slack, and Excel.
Prompting: Communicate naturally like a coworker.
- Provide context / set stage (+ adding context), define tasks, specify rules, and iterate on responses.
- Personalize claude with memory and styles to help Claude work better for you.
AI Fluency: Guided by the 4D Framework (Delegation, Description, Discernment, and Diligence) to evaluate and integrate Claude effectively.

What is Claude

Claude is an AI assistant designed to be your thinking partner.

built to be helpful, harmless, and honest
more than a chatbot
designed to be steerable and collaborative

Claude’s capabilities

Writing and content creation

Drafting emails, essays, reports, and creative content with tailored tone and structure.

Research and analysis

Summarizing documents, extracting key insights, and performing analysis on text or data.
Helps you explore research angles, compile findings, and analyze data to surface meaningful insights.

Coding assistance

Writing, debugging, explaining, and translating code across multiple programming languages.

Problem-solving and reasoning

Brainstorming ideas, handling complex cognitive tasks, mathematical problems, strategic thinking and analysis, and research

Learning new things

Explaining complex topics, summarizing concepts, and answering educational queries.

Ways to access Claude

Claude.ai: the primary way most people interact with Claude.
Claude Code: an agentic coding tool that is designed for developers
Claude and Slack
Claude for Excel

Converation with Claude

The best approach when speaking to Claude is like you would a coworker—naturally, concisely, and conversationally.

Writing effective prompts

Setting the stage
- What is your role and what are your objectives?
- Is there context about your work that Claude should know about?
Defining the task
- What action do you want Claude to take?
- Do you want Claude to write, analyze, build, or something else?
Specifying rules
- What’s the style or tone you want Claude to use?
- Are there examples that you can attach to show Claude what you’re looking for?

Example

I’m the marketing lead at an indie streaming startup, and we’re preparing an investor pitch deck for Series A investors. Can you research the current state of the independent film streaming market and identify key trends, competitor positioning, and growth opportunities? Use current web research with citations and structure it as a professional report of up to 5 pages, with an executive summary, market analysis, competitive landscape, and growth opportunities.

Setting the stage: We tell Claude this is for an investor pitch deck for a new indie streaming app—that’s the context and objective.
Defining the task: We provide the specific action (research the market) with relevant details (trends, competitors, opportunities).
Specifying rules: We ask for current web research with citations, structured as a professional report—telling Claude exactly what style and format we need.

Adding context

Uploads, connectors, and custom preferences offer ways to give Claude even more context about your work.

Some practical ways to use file uploads:

Upload a document and ask Claude to summarize the key points
Share an image and ask Claude to describe or analyze what it sees
Attach a spreadsheet and ask Claude to identify trends in the data
Upload code and ask Claude to explain how it works or find bugs

Iterating on Claude’s responses

Conversations with Claude are meant to be iterative.

If Claude’s first response isn’t quite what you wanted, you have several options:

Ask follow-up questions: Build on Claude’s response by asking for more detail, a different angle, or clarification.
- E.g., “Can you expand on the second point?” or “That’s helpful, but can you make it more concise?”
Provide feedback: Tell Claude what you liked and didn’t like about its response.
Redirect or restart: If Claude went in a different direction than you intended, simply steer it back.
- E.g., “Actually, I was asking about X, not Y. Let me clarify…”
- Worst case, restart your conversation in a new chat to fully refresh the context.

Personalizing Claude

Two features that help Claude work better for you over time to increase the power of your prompts

Memory: Automatically saves key context from your conversations — your role, preferences, past decisions, and working style — so you don’t have to repeat yourself every time you start a new chat.
Styles: let you customize how Claude communicates. Choose from preset options — like concise, formal, or explanatory — or create your own custom style by describing exactly how you want Claude to write.

Getting Better Results

Common challenges and how to fix them

As you start working with Claude, you’ll likely encounter moments where the response isn’t quite what you expected. Here are some of the most common challenges and how to address them:

Challenge	What’s happening	Try this
Claude’s response is too generic	Your prompt didn’t include enough context about your specific situation	Add details about your audience, role, or constraints. Instead of “Write an email about the project delay,” try “Write an email to our enterprise client explaining that the software integration will be delayed by two weeks. They’ve been patient so far but this is the second delay. Keep it professional but apologetic.”
The response is too long (or too short)	Claude is guessing at appropriate length	Be explicit: “Give me a two-paragraph summary” or “Keep this under 100 words” or “I need a comprehensive analysis—length isn’t a concern.”
Claude didn’t follow my format	Claude understood what you want but not how you want it presented	Show, don’t just tell. Provide an example of the format, or describe the structure explicitly: “Use bullet points with bold headers for each section.”
I got confident-sounding information that turned out to be wrong	Claude occasionally generates plausible but incorrect information, especially with specific facts or niche topics	For high-stakes work, verify key facts independently. Ask Claude to cite sources or indicate confidence level. Enable web search to ground responses in current information.
The tone isn’t right	Claude defaults to helpful and professional, which may not match your needs	Describe the tone in plain language: “Make this more conversational” or “This should sound authoritative and formal.” Provide an example of writing in the style you want.

The iteration mindset

Your first prompt rarely produces a perfect result—and that’s okay. Think of your initial prompt as the start of a conversation, not a one-shot request.

Effective Claude users:

Treat first drafts as starting points. Review what Claude produces, identify what’s working and what isn’t, then refine.
Give specific feedback. “Make it shorter” is fine, but “Cut the first two paragraphs and make the conclusion more action-oriented” is better.
Know when to start fresh. If a conversation has gone off track, sometimes it’s faster to open a new chat with a clearer prompt than to try to redirect.

AI Fluency

AI Fluency is the ability to collaborate effectively with AI tools—not just knowing which buttons to click, but developing the judgment to use AI well across different situations.

The 4D Framework for AI Fluency

Delegation

Deciding on what work should be done by humans, what work should be done by AI, and how to distribute tasks between them.
Includes understanding your goals, AI capabilities, and making strategic choices about collaboration.

Description

Effectively communicating with AI systems.
Includes clearly defining outputs, guiding AI processes, and specifying desired AI behaviors and interactions.

Discernment

Thoughtfully and critically evaluating AI outputs, processes, behaviors and interactions.
Includes assessing quality, accuracy, appropriateness, and determining areas for improvement.

Diligence

Using AI responsibly and ethically.
Includes making thoughtful choices about AI systems and interactions, maintaining transparency, and taking accountability for AI-assisted work.

Evaluating Claude for your workflows

As you start integrating Claude into more of your work, you might wonder: how do I know if Claude is actually good at this particular task?

Evals (short for evaluations) are a way to develop intuition for assessing Claude’s outputs on the tasks that matter to you. They’re systematic ways to test how well Claude performs on specific types of tasks that matter to you.

Running simple evals helps you:

Understand where Claude adds the most value in your workflow
Identify tasks where you’ll need to provide more context or examples
Build confidence in Claude’s outputs for recurring tasks

A simple and practical eval approach:

Gather examples. Collect 5-10 examples of a task you do regularly—emails you’ve written, reports you’ve created, analyses you’ve done.
Create test prompts. Write prompts that would generate similar outputs. Include the context you’d naturally have when doing this work.
Compare outputs. Run your prompts and compare Claude’s responses to your examples. Ask yourself:
- Does Claude capture the key information?
- Is the tone and style appropriate?
- What’s missing or could be improved?
Refine your approach. Based on what you learn, adjust your prompts, add examples to show Claude what good looks like, or identify where human review is essential.

Claude Desktop App

The Claude desktop app gives you three ways to work with Claude: Chat, Cowork, and Code

Chat is the same Claude you know from claude.ai, plus quick entry, screenshots, dictation, and connectors that come from running natively on your computer.
Cowork is an agentic tool — you give it a goal, connect it to your tools and resources, and let it do the work.
Code is for building software, from writing and testing code to deploying it.

	Chat	Cowork	Code
Optimized for	Quicker exchanges: exploring ideas, iterative drafting, quick answers, learning through dialogue	Complex or sustained work: research, analysis, file organization, producing finished documents and deliverables	Building software: writing, testing, running and deploying code
Key features	Quick entry, dictation	Work from local folders, plugins, subagents, scheduled tasks	Ask/Code/Plan modes, visual diffs, git integration, local and remote environments
Tools and extensions	Connectors, Skills, Claude in Chrome	Connectors (local and remote), Skills, Claude in Chrome, Plugins, Computer Use	Connectors, Skills, Claude in Chrome, Plugins, Hooks

Organizing Work and Knowledge

Tue, 02 Jun 2026 00:00:00 +0000

Introduction to Projects

What are projects?

Projects are self-contained workspaces with their own memory, chat histories, knowledge bases, and customized instructions. Think of them as dedicated environments for specific work streams.

Projects are ideal for s

toring knowledge Claude should reference
organizing related chats around a specific topic or work area,
collaborating with team members who need access to the same shared context

When to use projects

Projects are particularly valuable when you’re working on something ongoing—not just a one-off question. Consider creating a project when you have a workflow with:

Reference materials you’ll use repeatedly (meeting notes, survey results, reports, historical data, etc.)
Consistent requirements for how Claude should respond (always use formal language, always cite sources, always follow our template)
Team collaboration needs where multiple people should work from the same foundation

Creating projects

Set up project

Add project instructions to tell Claude how to behave across all conversations in this project

Good project instructions typically include:

Instruction	Example
Context about what you’re working on	“This project is for creating marketing content for our B2B software product.”
Process instructions	“First consider a blog structure that will entice this audience, then write the draft.”
Tone and style preferences	“Use a professional but conversational tone. Avoid jargon when possible.”
Specific requirements	“Always include a call-to-action at the end of marketing copy.”

You can also use project instructions to automate workflows (Think of instructions as programming Claude’s behavior for this project.)
- E.g., “When I upload a meeting transcript, create a structured summary using this template.”

Build knowledge base
- Knowledge base is where you upload documents that Claude should reference
- What to upload:
  - Reference documents
  - Background materials
  - Technical documentation or specifications
  - Examples of work you want Claude to emulate

Pro tip: Name your files descriptively. Claude uses file names to understand and retrieve the right information.

Working within project

Once your project is set up, you can start chatting with Claude. Each conversation within the project automatically has access to your knowledge base and follows your project instructions.

Examples

Q4 product launch

Upload your product specs, competitive analysis, and messaging brainstorming notes. Claude will have this context top of mind for any inquiry or document draft.

Research support

Centralize your competitive review, user research data, and customer feedback. Claude can help you synthesize sources, draft reports, and maintain consistency across recommendations.

Client account hub

Keep your client’s brand guidelines, past deliverables, and communication history in one place. Set instructions so Claude matches their tone and references their specific context when creating proposals or reports.

Event planning workspace

Upload venue contracts, speaker bios, and attendee data. Claude can help generate run-of-show documents, attendee communications, and post-event reports that stay consistent with your event’s theme.

Job description generator

Gather past job descriptions, team charters, and internal headcount request docs. Work with Claude to draft job descriptions that reflect your team’s actual work and culture.

Best practices for projects

Start focused, then expand

Begin with a specific use case rather than trying to create one project for everything. You can always add more content as you go.

Keep your knowledge base current

Outdated documents can lead to outdated responses. Review and update your project knowledge periodically.

Write clear instructions

Be specific about what you want. Vague instructions lead to inconsistent results.

Name your documents descriptively

(e.g., ‘Q4-2025-Sales-Report.pdf’ not ‘report.pdf’) and group related files together. Claude uses filenames and proximity to understand relationships between documents.

Reference documents by name

When asking questions, you can mention specific documents to help Claude focus its search: “Based on our Q3 report, what were the top customer concerns?”

Creating with Artifacts

What are artifacts?

Artifacts are standalone, interactive outputs that Claude creates in a dedicated window alongside your conversation.

Common artifact types

Documents: including markdown, plain text, Word docs, PDFs, PowerPoint, and Excel.
Code snippets: Working code in any programming language.
HTML pages: Complete web pages with HTML, CSS, and JavaScript in a single file.
SVG images
Mermaid diagrams: Flowcharts, sequence diagrams, Gantt charts, org charts, and more.
React components: Interactive UI elements with real functionality—calculators, dashboards, games, data visualizations.

Creating artifact

Creating an artifact is as simple as having a conversation. Just describe what you want, and Claude will determine whether to present it as an artifact.

If Claude doesn’t automatically create an artifact when you expect one, you can explicitly ask: “Create this as an artifact” or “Show me this in an artifact.”

Tips for getting the most from artifacts

Be specific about what you want

Specify layout preferences, design requirements, style options, and specific functionalities you expect the artifact to have.

Example: “Build a monthly budget tracker where I can input expenses by category, see a pie chart breakdown, and get a warning when I’m over budget” is better than just “Build a budget tracker”.

Describe the end user

Telling Claude who will use the artifact helps it make appropriate design choices.

Iterate incrementally

Ask Claude to add one feature or make one change at a time. This makes it easier to identify what’s working and catch issues early.

Request artifacts when needed

If you ask for something substantial and Claude responds in the chat instead of creating an artifact, just say “Please create that as an artifact.”

Working with Skills

What are Skills?

Skills are folders of instructions, scripts, and resources that Claude loads dynamically to improve performance on specialized tasks. Think of them as expertise packages—they teach Claude how to complete specific tasks in a repeatable way.

Types of Skills

Anthropic Skills
- Created and maintained by Anthropic.
- Include enhanced document creation capabilities for Excel, Word, PowerPoint, and PDF files.
- Anthropic Skills are available to all paid users and Claude invokes them automatically when relevant—you don’t need to do anything special to use them.
Custom Skills are ones you or your organization create for specialized workflows and domain-specific tasks.
- For example, you might create a skill that applies your company’s brand guidelines to presentations, structures meeting notes in a specific format, or executes your organization’s data analysis workflows.

Using Skills in practice

You typically don’t need to think about them—Claude handles skill selection automatically based on your request.

Security considerations

Because Skills can include executable code, it’s important to use them thoughtfully:

Only install custom Skills from trusted sources
Anthropic’s built-in Skills are tested and maintained by Anthropic
Custom Skills you upload are private to your individual account
If you’re installing a custom Skill from an external source, review its contents before use to understand what it does.

Creating custom skills

Custom Skills let you teach Claude your specific workflows, brand guidelines, and ways of working—so Claude can apply that knowledge automatically whenever it’s relevant.

The easiest way to create a custom Skill is through conversation with Claude itself. You don’t need to write code or manually create files—Claude handles the technical structure for you.

Start a new chat and tell Claude what you want to create.
- E.g., “I want to create a skill for writing quarterly business reviews” or “I need a skill that applies our brand guidelines to presentations.”
Answer Claude’s questions.

Claude will interview you about your workflow, asking things like: What should this skill do? What makes good output for this type of work? Can you give examples of when you’d use this skill?
Upload reference materials if you have them (templates, style guides, brand assets, or examples of work, etc.).
Save your skill. When finished, Claude generates a file containing your properly structured skill. All you have to do is save it and the skill will be ready for Claude to use.

Skills vs. Projects

Think of it this way: projects store knowledge, skills perform tasks.

Projects are knowledge hubs.
- They hold the reference materials Claude needs to understand your work—project specs, meeting notes, research documents.
- When you upload files to a project, Claude draws on that information across every conversation within that project.
Skills are procedural machines.
- They encode how Claude should execute a task—the specific steps, order of operations, and methodology you want followed every time.
- Skills shine when you have repeatable workflows you want Claude to run consistently.

The two features complement each other: A skill can reference knowledge stored in a project. The project provides the what (information), the skill provides the how (process).

	Projects	Skills
Purpose	Store knowledge Claude references	Define processes Claude executes
Best for	Long-term context, reference materials, team collaboration	Repeatable workflows, multi-step tasks, consistent methodology
Example	Customer hub, research buddy, feedback generator	Process guidelines (like brand or legal), Blog drafting, PDF creation
Persistence	Knowledge available across all chats in the project	Instructions applied when the skill is invoked

Example

If we think of Claude as a digital employee, a Project is his “dedicated office” that stores all the background knowledge (the What)—such as medical data and design specs for a “Cat Health App”—ensuring he always has access to these reference materials whenever he is in this office. On the other hand, a Skill is a “standard operating procedure (the How)” that he has mastered—such as a structured workflow for crafting social media posts by extracting selling points, adding emojis, and generating catchy headlines. When you ask him to write a post within the “Cat App Office (Project),” he instantly runs his “content creation assembly line (Skill),” flawlessly processing the professional cat data from the project into a highly engaging, structured, and polished output.

Expand Claude Research

Tue, 02 Jun 2026 00:00:00 +0000

Connecting Tools

What are connectors?

Connectors transform Claude from an assistant into an informed collaborator by giving Claude access to the same tools, data, and context that you use every day. -> Instead of starting every conversation from scratch, Claude can work directly with your actual information.
Connectors allow Claude to read information and perform actions on your behalf.
The Model Context Protocol (MCP) powers connectors.
- MCP is like USB-C for AI—a universal standard that allows Claude to connect to many different applications through a single, consistent interface.
- Following this open standard, developers can build connectors for any tool, and those connectors work seamlessly with Claude.
Two types of connectors: web connectors and desktop extensions.
- Web connectors link Claude to cloud services like Google Drive, Notion, Slack, and Asana.
- Desktop extensions run locally on your computer through the Claude Desktop app, giving Claude access to local files and native applications.

Finding and connecting tools

Anthropic maintains a directory of recommended connectors at claude.ai/directory. The directory is organized into two tabs:

Web: Cloud services and applications (Gmail, Notion, Slack, Asana, Linear, Stripe, and many more)
Desktop extensions: Local tools that run on your computer through the Claude Desktop app

Using connectors in work

Once you’ve connected your tools, Claude considers them when responding to your requests.

Some practical ways to use connected tools:

Project management (Asana, Linear, Jira)

“What are my highest priority tasks due this week?”
“Create a new task for reviewing the Q4 budget proposal”
“Summarize the status of our product launch project”

Communication (Slack, Gmail)

“Find the email thread where we discussed the vendor contract”
“Draft a reply to the latest message in the #marketing channel”
“What did the team decide about the timeline in yesterday’s discussion?”

Documentation (Notion, Google Drive, Confluence)

“Search our documentation for our brand voice guidelines”
“Summarize the meeting notes from last week’s product review”
“What does our style guide say about using contractions?”

Business tools (Stripe, PayPal, Salesforce)

“Show me revenue trends for the past quarter”
“What’s the status of the Acme Corp opportunity?”
“List recent transactions over $1,000”

Security and permissions

Important considerations:

Scoped access: Permissions are specific to what the connector needs and you can toggle individual permissions on and off within each application’s menu.
Claude sees what you see: Claude can only access data you have access to. Connecting your work email doesn’t give Claude access to your CEO’s inbox—only your own.
Revocable at any time: You can disconnect a service through Claude’s settings or through the third-party service’s security settings. Just as with Skills, you can also find or build custom connectors. Exercise the same caution — only install connectors from trusted sources.

Enterprise Search

Research Mode for Deep Dives

What is Research?

Research is an advanced feature that transforms Claude from a conversational assistant into a systematic investigator.

When you enable Research, Claude doesn’t just answer your question—it explores it from multiple angles, synthesizing information from across the web and your connected integrations. -> Think of it as having a skilled research assistant who can help gathering information, cross-referencing sources, and compiling a comprehensive report.
Research is particularly valuable when you need more than a quick answer. It’s designed for situations where a thorough understanding requires pulling together information from multiple sources, comparing different perspectives, and synthesizing findings into actionable insights.

When to use Research

Use Research when you need:

Comprehensive reports that synthesize information from multiple sources
In-depth analysis across the web and your connected integrations (like Google Workspace)
Thorough investigations that would typically require hours of manual work
Comparative analysis, such as evaluating competitors or vendor options
Reports with citations you can verify

Research is ideal for tasks like:

Market analysis and competitive research
Planning complex projects, like team offsites or product launches
Synthesizing information from your email, calendar, and documents
Creating technical documentation that draws from multiple sources
Preparing briefings that require current, verified information

Consider web search instead when:

You need a quick, specific fact (like today’s stock price or a company’s address)
The answer requires only one or two sources
Speed matters more than comprehensiveness

Consider extended thinking instead when:

You need deep reasoning on a complex problem that doesn’t require external information
You’re working on mathematical problems, code debugging, or logical analysis
The answer comes from reasoning through a problem rather than gathering information

Consider enterprise search instead when:

You need answers that draw from your organization’s internal knowledge — documents, Slack threads, emails, meeting notes
You’re onboarding and want to quickly find how your company handles something (like policies, processes, or past decisions)
You’re asking a question that’s specific to your company, not the public web

How Research works

Enable Research = Activate an agentic, multi-step process that goes far beyond a simple web search

Claude autonomously decides what to search next based on what it has already found, pursuing leads and filling gaps without you needing to direct each step.

Claude plans its approach. When Research is enabled, extended thinking automatically activates. This lets Claude break down your request, identify what information it needs, and plan how to investigate different angles of your question.
Claude conducts multiple searches. Rather than running a single search, Claude conducts many searches that build on each other. It determines what to investigate next based on what it finds, pursuing promising leads and filling in gaps.
Claude synthesizes findings. After gathering information from multiple sources—including the web and any connected integrations like Gmail, Google Calendar, or Google Drive—Claude compiles everything into a comprehensive, well-organized report.
Claude provides citations. Every claim in Research reports links back to its source, making it easy to verify information and dig deeper when needed.

Tips for effective Research prompts

Be specific about your goals. Instead of “Tell me about the EV market,” try “Analyze the electric vehicle battery market—identify key players, technology trends, and supply chain challenges that might affect investment decisions.”
Specify the sections or structure you want. Claude will organize its findings around the structure you provide. For example: “Compare venue options for a team offsite including: location and accessibility, meeting space and amenities, catering options, and pricing considerations.”
Include relevant constraints. Budget ranges, timelines, geographic requirements, and other parameters help Claude focus its research on relevant options.
Ask Claude to help refine your prompt. If you’re not sure how to frame your research question, you can even ask Claude to help you write a better Research prompt before enabling the feature.

Working with connected integrations

When you have Google Workspace or other integrations connected, Research becomes even more powerful. Claude can pull context from your emails, calendar, and documents alongside web research.

Claude

Tue, 02 Jun 2026 00:00:00 +0000