Key takeaway:
1. Claude is Anthropic's AI assistant. You can use it free at claude.ai, no card and no API key.
2. The model names (Haiku, Sonnet, Opus, Fable) are speed-versus-depth choices, and beginners should leave the default alone.
3. Anthropic put its whole training curriculum online free at Claude Academy on 20 August 2026, certificates included.
4. The skill that separates people who get value from people who quit is the 4D framework: Delegation, Description, Discernment, Diligence.
5. Skills are the feature worth learning early. They turn a one-off prompt into something Claude reuses every time.
Anthropic just made the whole thing free to learn. On 20 August 2026 the company opened Claude Academy, the same curriculum it uses to train its own staff, with real certificates, no subscription and no API key.
So this is a good week to start. Below is what Claude actually is, what it costs, which model to pick, and the framework that turns it from a novelty into something you use every day.
I have been building with Claude since the first Sonnet release. What trips up beginners is rarely the tool. It is knowing what to hand over and how to ask.
What is Claude?
Claude is an AI assistant built by Anthropic. You type, it answers. It reads documents, writes and runs code, searches the web, remembers your projects, and produces files you can download.
Three things make it feel different from other assistants once you use it for real work:
- It handles long documents well. Paste a contract, a codebase, or a 60-page report and ask questions about it.
- It writes in a steadier voice. Give it a sample of your writing and it will match your rhythm more closely than most models.
- It says when it is unsure. Ask it to flag uncertainty and it usually will, which matters more than any benchmark when you are relying on the answer.
You reach it at claude.ai in a browser, on iOS and Android, or as a desktop app. Anthropic also builds Claude into other surfaces: Claude Code for developers, Claude Cowork for everyday work, plus Chrome and Microsoft 365 integrations.
Claude's models, explained without the jargon
The model picker is where beginners stall. Here is the whole lineup in plain terms.
- Haiku is the fast, cheap one. Good for quick questions and high-volume work.
- Sonnet is the everyday default. Sonnet 5 landed on 30 June 2026 and handles the large majority of real tasks.
- Opus is the deep thinker. Reach for it on hard reasoning, long analysis, and gnarly code.
- Fable 5 is the frontier-class model Anthropic released publicly in June 2026, its most capable tier available to subscribers.
- Mythos sits above all of it and is not publicly released. Anthropic restricted it over cybersecurity capability concerns, so you will not meet it in the app.
The beginner move: leave the default alone. Switch to a deeper model only when a task has several dependent steps or the cost of a confident wrong answer is real. That single habit beats any amount of model trivia.
What Claude costs in 2026
There is a genuine free tier, and it is better than the free tier reputation suggests.
- Free, $0. Chat on web, iOS, Android and desktop. Web search, memory, file creation with code execution, and credits toward Fable at half the weekly limits.
- Pro, $20 a month ($17 billed annually). More usage, plus Claude Code, Claude Cowork, Claude Design, Claude Science, Projects, and Microsoft 365.
- Max, from $100 a month. Choose 5x or 20x more usage than Pro, higher output limits, and early access to new features.
- Team, $20 to $25 per seat, with central billing, SSO, and no training on your content by default.
- Enterprise, priced per seat plus usage, with audit logs and a HIPAA-ready option.
Prices vary by region, so check the pricing page for yours. Start free for a week. Move to Pro when you hit limits or when you want Claude Code and Cowork, which is where the real gains are.
Your first ten minutes with Claude
Skip the tutorial hunting. Do these four things in order.
- Give it a real task, not a test. Paste something from your actual work: a messy email thread, a spreadsheet, a job description.
- Tell it who it is and what you want back. "Act as a hiring manager. Turn this into a job post under 300 words with three must-have requirements."
- Push back once. Say what is wrong with the first answer. The second answer is usually the one you keep.
- Turn on Projects. Drop in your files and context so you stop re-explaining yourself every morning.
If you want ready-made starting points, our Claude prompt library is built for exactly this: copy, paste, swap in your details, run.
The 4D framework: the actual skill worth learning
The most valuable course in Claude Academy is AI Fluency: Framework and Foundations. It is 14 lessons, about four hours, with a quiz and a completion badge. It was built with two academics, Rick Dakan of Ringling College of Art and Design and Joseph Feller of University College Cork, so it is a teaching framework rather than a product tour.
It teaches four competencies. They are the difference between people who get value from AI and people who bounce off it.
Delegation
Deciding what to hand over. It means knowing your own problem, knowing what the tool is actually good at, and judging whether this particular task suits it. Bad AI experiences usually start here, with someone delegating a job the model was never going to do well.
Description
Communicating clearly. Specify the product you want, the process to follow, and how you will judge performance. This is prompt engineering, described honestly: less about magic words, more about saying what you actually mean.
Discernment
Judging what comes back. Read the output critically, spot where it drifted, and steer it with a second and third pass. The people who get the most from Claude treat the first answer as a draft.
Diligence
Owning the result. Be transparent about what was AI-assisted, verify anything factual, and think before you ship. You are accountable for the output regardless of what produced it.
Learn those four and every model in every app gets easier, including the ones that do not exist yet.
Two free curriculums worth your time
If you want structure rather than trial and error, these two are the best free options available right now.
Claude Academy
Over 20 courses across three tracks: AI fluency for non-technical people, product training like Claude 101, and developer deep-dives on Claude Code, the API, MCP servers, Agent Skills and Subagents. There are versions aimed at educators, students, nonprofits and small businesses. It routes you by where you already are rather than dumping a course list on you.
One honest note: these courses have been building all year under the Anthropic Academy name. What is new is the home and the guided paths, not 20 brand-new courses. It is still the most complete free AI curriculum any lab has published.
Microsoft AI For Beginners
If you want to understand what is happening under the hood, Microsoft's AI For Beginners curriculum is 12 weeks and 24 lessons, free under MIT, with 66,000+ stars on GitHub. It covers symbolic AI, neural networks, computer vision, natural language processing and AI ethics, with Python notebooks in both PyTorch and TensorFlow.
The two pair well. Claude Academy teaches you to work with AI. Microsoft's curriculum teaches you how AI works. Start with the first, add the second if the mechanics interest you.
Claude Skills, the feature to learn early
A Skill is a set of instructions and files Claude loads when a task calls for it. Instead of pasting your brand guidelines into every conversation, you write them once as a Skill and Claude applies them whenever it is relevant.
This is the jump from using Claude to configuring it. A marketer builds a Skill for their tone of voice. A developer builds one for their repo conventions. An accountant builds one for their reporting format.
It is also the fastest payback for a beginner, because the work compounds. Every Skill you write keeps earning. We package the ones we use most in our own Claude Skills products, including a mini-course that walks through building your first one.
What Claude is good at, and where I still use something else
Being honest about this saves you frustration.
- Long documents and codebases. Claude is my first choice, and it is the reason I keep Claude Code open all day.
- Writing that has to sound like a person. Give it a voice sample and it holds the voice.
- Everyday work with files and apps. That is what Claude Cowork is for.
- Image generation. Claude does not make images, and pretending otherwise wastes your afternoon. We covered the workarounds in can Claude generate images.
- Coming from ChatGPT? The habits transfer. Our ChatGPT for beginners guide covers the same ground on the other side.
Five beginner mistakes
- Asking tiny questions. Claude handles context well. Give it the whole document, not a summary of it.
- Accepting the first answer. One round of feedback is where the quality lives.
- Chasing the biggest model. Default first. Escalate when the task earns it.
- Never using Projects. Re-explaining your business every session is a tax you can stop paying today.
- Skipping verification. It sounds certain when it is wrong. Click the citations.
Frequently asked questions
Is Claude free to use?
Yes. The free plan covers chat on web, mobile and desktop, with web search, memory and file creation, plus limited credits toward the top-tier model. Paid plans start at $20 a month.
Is Claude better than ChatGPT?
They are close, and it depends on the job. Claude tends to win on long documents, code and writing that has to hold a voice. Try both on your own real task for a week and keep the one that needs less editing.
Which Claude model should a beginner use?
The default. Switch to a deeper model when a task has several dependent steps or when a confident wrong answer would cost you something real.
Do I need to know how to code?
No. For non-technical people, Claude's value sits in writing, analysis and everyday work. The AI Fluency course assumes no prerequisites at all.
Are the Claude Academy certificates worth anything?
They are completion badges rather than accredited qualifications. The value is the curriculum and the framework. The badge is a nice extra for a LinkedIn profile.
What are Claude Skills?
Reusable instructions and files Claude loads when relevant, so it applies your standards automatically instead of you pasting them into every chat.
Where to go from here
Pick one real task this week and give it to Claude properly: a role, the context, and the output you want. Push back once on the answer. That single loop teaches you more than any course.
Then take the four hours for AI Fluency, because Delegation, Description, Discernment and Diligence outlast every model release.
Start with prompts that already work: browse the Claude prompt library, or grab a free guide to keep beside you.
