Chapter 13: Claude From Beginner to Advanced User
A quick note before you begin
Snapshot taken in the third quarter of 2026. Model names, prices, and limits will have moved. Structure and judgment will not. Verify the specifics at the official source, for this product that is claude.ai/pricing and the help center at support.claude.com.
What it is, and how it differs in character
Claude is made by Anthropic, a company founded by former OpenAI researchers, and it competes directly with ChatGPT at the consumer level. Same basic proposition: a conversational assistant that reads, writes, analyzes, and increasingly acts.
The differences are real but narrower than partisans on either side suggest, and they concern emphasis rather than kind. Claude has been positioned around writing quality, long-document work, coding, and a somewhat more reserved conversational manner. Users who prefer it usually cite prose that requires less editing and a lower tendency toward enthusiastic agreement. Users who prefer ChatGPT usually cite the broader feature set and the larger ecosystem.
Both are frontier products. Anyone telling you one is categorically better is describing their own workflow, not a general fact, and the ranking changes with each release cycle anyway.
The plans
The free plan gives you Claude at claude.ai with a Google account, Apple ID, or email address, no payment required, including Projects, file uploads, and Memory. It is a working version of the product rather than a demonstration.
The principal paid consumer plan provides substantially more usage than the free tier, along with priority access during busy periods and access to additional features. Its cost has generally been comparable to a casual lunch each month. Current figures are collected in the dated pricing snapshot in the Bonus Resource Library.
Max comes in two tiers that provide substantially more usage than Pro. It also includes priority access during periods of high demand and earlier access to selected capabilities. Capacity is the main reason to choose Max, so if you rarely reach Pro’s limits and do not need those additional benefits, the upgrade may add little value.
Team and Enterprise plans exist for organizations, adding administrative controls, central billing, and (the part that matters most) different data handling terms.
One structural note worth understanding: limits here are expressed as usage per rolling session rather than as a simple message count, so heavy work in a short window hits them faster than the same volume spread across a day.
These plans change often. Use the description here to understand the choices, then check the dated pricing snapshot and the provider’s official page before paying.
Level two: Configuration
Projects. Grouped conversations with shared context, uploaded documents, and their own instructions. The same project advice applies even more strongly here, because Claude’s long-document handling makes a well-stocked project genuinely powerful. Load in the background material once; stop re-explaining.
Memory. Details carried across conversations. Off by default in some configurations and on in others: check yours, know how to view it, know how to clear it.
Styles. A distinctive feature: you can teach it to write in a particular voice by giving it samples of your own writing, then apply that style to any conversation. If you produce a lot of written work under your own name, this is the highest-value twenty minutes in the product.
Preferences and instructions. The standing note applied everywhere. The same advice applies: keep it short, and include an instruction about how to signal uncertainty.
Search past chats. The tool can retrieve your own previous conversations. Useful, and worth knowing it exists before you re-explain something you covered last month.
Artifacts
The feature most worth understanding, and the one with no exact equivalent elsewhere.
When Claude produces something substantial — a document, a piece of code, a working web page, a chart — it appears in a panel beside the conversation rather than inline. You can then iterate on it directly: revise, extend, correct, and watch it update while the conversation continues alongside.
Two consequences. Written work becomes an object you refine rather than a block of text you keep regenerating. And it can build small working programs — a calculator, an interactive chart, a form — that run immediately in the browser, without you knowing how to code.
For a non-technical reader, being able to build a working tool through conversation can feel like a superpower.
Files, data, and analysis
Upload documents, spreadsheets, images, and PDFs. Ask specific questions. Claude’s document handling is a genuine strength, particularly on long material.
For numerical work, it can write and run code to compute rather than producing figures from its own sense of what they should be. This distinction is the difference between reliable arithmetic and confident guessing, so ask which happened when a number matters.
It can also produce files: Word documents, spreadsheets, slide decks, PDFs. The later file-work guidance explains where the output needs cleanup.
Level three: Delegation and the expanding surface
Claude has spread well beyond the chat window, and the expansion is the main product story of the last year.
Connectors. Links to outside services (file storage, calendars, email, workplace tools) so Claude can work with your actual material rather than what you paste. Convenience and exposure arrive together; Chapter 25 covers the tradeoff.
Claude Code. A coding agent that works in a terminal or editor, reading a codebase, making changes, and running tests. Included with paid consumer plans rather than sold separately. If you write software this is likely the main reason to subscribe. If you do not, skip it.
Cowork. The equivalent for non-developers: handing off multi-step tasks involving many files and tools, in a desktop application, rather than conducting them turn by turn in chat.
Claude in your existing applications. Browser, spreadsheets, presentations, and a Slack integration. The strategic direction is becoming clear across the industry: the assistant is moving to where the work already happens.
Skills and plugins. Reusable packages of instructions for recurring specialized work. They become useful when you repeat a task to a fixed standard and your prompt library outgrows a document.
Where it shines
Held loosely, and true as of writing:
• Long-form writing that needs less editing before it sounds like an adult wrote it
• Long documents: reading them, comparing them, holding a lot of material at once
• Coding, where it has been among the strongest options for some time
• Artifacts, for building small working things without programming
• A somewhat more direct manner, including more willingness to say a plan has problems
Current limitations
A smaller ecosystem. Fewer third-party integrations and shared configurations than ChatGPT.
Tighter free limits. The free tier gives you a capable model but less of it. Heavy users hit the ceiling sooner than on some competitors.
More conservative refusals. It declines some requests other tools will attempt. Sometimes that reflects a real risk; sometimes it is a false positive on an ordinary request. If you hit one that seems clearly wrong, the feedback control in the interface is the mechanism for reporting it.
The familiar failure modes apply here in full. Fabricated citations, silent assumptions, confident errors on thin material. A more measured tone is not a reliability guarantee, and the more reasonable something sounds, the more disciplined your verification needs to be.
One factual difference worth noting because it bears on Chapter 25: Anthropic does not display advertising in its products or allow paid placement in Claude’s responses. Chapter 12 noted that some competing free tiers have begun carrying sponsored content. Whether that matters to you is a judgment about how you want recommendations funded, and it is worth making deliberately rather than by default.
The subscription decision
Ask the same three purchasing questions. Start free, use it for two weeks, then ask whether you are hitting limits, whether you need a specific paid capability, and what an hour of your time is worth.
The Max tiers deserve their own note, because the naming misleads people: they buy volume, not capability. Nobody should buy Max without first exhausting Pro.
Privacy settings to check
Data handling. Understand whether your conversations may be used to improve models on your plan, and what the retention period is. Consumer and business terms differ materially. This is the setting to check before pasting anything from work.
Memory. Review it, know how to clear it.
Connectors. Every connected service is a standing grant of access. Review what you have connected quarterly and disconnect what you no longer use.
Exercise 13.1: Build a style from your own writing
Gather three or four samples of your writing that sound like you at your best. Create a style from them.
Then run the same request with and without it. For most people this is the moment the tool stops producing text that needs rewriting.
Exercise 13.2: Make something with an artifact
Ask for a small working tool you would use: a calculator for a calculation you do often, a checklist that tracks state, a simple chart of numbers you care about.
Then refine it three times. You have just built software.
Exercise 13.3: Run the same task in two tools
Take a real task and run it in both Claude and ChatGPT with an identical prompt.
Compare on output quality, how much editing each needs, and how each handled the parts you know well. This is worth more than any published comparison, including mine, because it is measured against your work.