SiteGround Academy

Claude Code can help build software, but it still expects you to understand how software gets written.

Claude Code is built for people who are already comfortable with terminals, codebases, Git, and deployment workflows — in short, developers. If any of those terms are unfamiliar, it’s probably not the right starting point. Beginners who just want a live website or simple web app will get a more direct path with a managed AI website and web app builder like Coderick AI.

What is Claude Code?

Claude Code is Anthropic’s agentic coding tool: an agent that plans work, reads a full codebase, edits files across multiple directories, runs tests, and can create pull requests. It launched as a generally available product in May 2025.

What Claude Code can do inside a codebase

Claude Code operates at the project level. It can read files across an entire codebase, make coordinated edits in multiple locations, run shell commands (instructions typed directly into your computer’s terminal), handle Git workflows (tracking and saving changes to code), and work through multi-step engineering tasks. It understands context that spans dozens or hundreds of files at once, which is what makes it useful for developers managing complex projects.

Where Claude Code runs: terminal, code editor, desktop app, and browser

Claude Code works in terminals, code editors, desktop apps, and developer platforms like GitHub. The terminal version has the most features and gets updates first. The underlying engine is the same across all of them, so settings you configure in one carry over to the others.

If none of those are part of your day-to-day workflow, that’s a good sign Claude Code isn’t the right starting point — and this article will help you find one that is.

If terms like terminal, CLI, or CI/CD are new, the vibe coding glossary covers the vocabulary beginners run into most.

Why Claude Code is different from a no-code AI builder

A no-code AI builder asks what the user wants to create. Claude Code asks what the developer wants to do inside a code project that already exists. The distinction matters: Claude Code doesn’t generate a finished website from a description. It helps technical users work faster inside a codebase they’re already managing or are prepared to manage.

How does Claude Code work for beginners?

Claude Code works for beginners who are willing to learn a development environment first. That means a terminal, a code project, and a connected account, none of which the tool creates for you.

What you need before using Claude Code

Anthropic’s quickstart lists three requirements before you start: 

1. A terminal or command prompt open 
2. A code project to work with 
3. A Claude subscription (Pro, Max, Team, or Enterprise), Claude Console account, or access through a supported cloud provider. 

That’s before a single prompt is typed.

Why a terminal or command prompt can be a barrier

The terminal is a text-only screen on your computer where you type instructions instead of clicking buttons. There’s no visual interface, no drag-and-drop, and no menus. For users who’ve never opened one, figuring out how to find it, open it, and type the right commands is the first obstacle — and it comes up before any building begins.

Why having a code project matters

Claude Code is designed to work with an existing project. You open your terminal, go to the folder where your project lives, and run the tool from there. For beginners who don’t have a project folder, repository, or a codebase yet, Claude Code may feel like the wrong starting point.

Who is Claude Code actually for?

Claude Code is built for users who want to work inside a development environment. 

Developers working with existing codebases

Claude Code’s primary use case is helping developers move faster inside projects they already own. It can read a large codebase, understand how the pieces connect, and make changes across multiple files simultaneously. That’s not useful until the codebase exists.

Technical founders and product teams

Technical founders who write code and product teams with engineering resources can use Claude Code to move faster on features, debug issues, and automate repetitive development tasks. The time savings are real, but they only show up for users who understand the technical context already; they know enough to check the AI’s work and point it in the right direction.

Teams that use GitHub, pull requests, or CI/CD

Claude Code integrates with GitHub Actions, supports pull request workflows, and can run inside CI/CD pipelines. For teams that already use these tools daily, Claude Code fits into the workflow they have. For beginners, those are new concepts to learn on top of the tool itself.

Users who want control over code and deployment

Some users specifically want to own their source code, manage their deployment pipeline, and have access to every file the project creates. Claude Code is built for that use case. It’s a code-control path, not a managed website or app publishing environment. If that level of control is the goal — and the technical fluency to exercise it is already there — Claude Code is well suited to the work.

Why can Claude Code feel difficult for beginners?

Claude Code might feel technical if you’re a beginner because it works inside development workflows. Each of its strengths assumes a layer of technical knowledge the user has already built.

It still uses developer tools and project files

Claude Code’s terminal CLI is its most capable interface. Working with it means being comfortable in a command-line environment, which requires knowing how to navigate directories, install packages, and run commands. Users who have never done those things won’t get far before hitting something unfamiliar.

It can require Git, repositories, and approvals

Git is a version control system that tracks changes in a codebase. Claude Code works with Git naturally: it can make commits, create branches, and generate pull requests. For a developer, those are routine tasks. For a beginner, Git is a separate system to learn, with its own terminology, setup steps, and ways to go wrong.

It does not remove hosting, DNS, SSL, or deployment decisions

Generating code and launching a website are two different things. Claude Code helps produce or modify code. Getting that code onto the internet still requires a hosting account, a deployment pipeline, DNS configuration (connecting your domain name to the server where your site lives), and SSL setup (the security certificate that makes your site load as https rather than http). None of that is simplified by the coding agent. A beginner who uses Claude Code to build something still has to solve the publishing problem afterward.

It may create a working project before the user knows how to launch it

This is a common experience for non-technical users who try Claude Code: they get working code faster than they expected, then find themselves stuck because they don’t know what to do with it. The code exists; the infrastructure to run it publicly still has to be chosen, configured, and maintained. 

Cost is another reason beginners look beyond coding agents. Claude Code pricing and deployment costs after the project is built often come as a surprise, since hosting, domain registration, and any external services are separate expenses on top of the subscription.

What should beginners look for in a Claude Code alternative?

If you want to build a site without Claude Code, the right alternative depends on the goal. For most beginners, the goal isn’t “generate code” — it’s “get your website live.” That changes what the tool needs to do.

A plain-language building experience

A useful beginner-friendly Claude Code alternative lets users describe what they want in plain language without needing to translate it into technical instructions. “I want a booking form on the homepage” should be enough. The tool should handle the interpretation, not the user.

No terminal or local setup requirement

Any tool that requires installing software, running commands, or configuring a local development environment before the first page renders is adding friction the beginner doesn’t need. The building experience should start in the browser, not in a local terminal setup.

Built-in publishing and hosting

Building and publishing are usually treated as separate steps, but for most beginners they should be part of the same experience. An alternative that generates a project but leaves the user to figure out where to put it hasn’t solved the actual problem.

Custom domain, SSL, backend, and database support

A beginner-friendly tool should handle the infrastructure decisions that would otherwise require research, separate service accounts, and manual configuration. Custom domains, SSL certificates, form backends, and basic database functionality should be part of the managed environment, not add-ons.

Easy updates after the site or app goes live

Launching is only the beginning. A tool that makes it easy to build something but hard to change it later is a problem waiting to happen. Making changes after launch should be as easy as building in the first place and ideally through the same interface.

Why is Coderick AI a better fit for many non-technical users?

Coderick AI is a better fit for many beginners when the goal is not to manage a codebase, but to get a working website or web app live. Instead of starting with a terminal, repository, deployment pipeline, hosting account, DNS records, and database setup, users describe what they want and build it inside a managed AI website/app environment.

It starts with describing the website or app you want

Coderick AI is a web-based AI website and app builder that runs inside a SiteGround account. There’s no installation and no terminal. Users open it in a browser, describe the site or app they want to build, and Coderick AI generates the project from there. It’s designed for users with zero coding or design experience.

It creates structure, layout, pages, forms, and basic logic

Coderick AI doesn’t just produce code. It generates a working project: navigation structure, page layouts, form functionality, and basic application logic (things like a contact form that sends you an email, a booking system, or a price calculator). For more advanced starting points, it also accepts file uploads including .html, .css, .js, .jsx, .ts, .tsx, .scss, and .md files, so users who prototyped something in Claude Design or another tool can hand those files over and continue building from there. That’s the shorter path described in how to turn a Claude Design project into a real website. If the starting point is a prototype rather than a codebase, the broader AI prototype to published website workflow explains how to move from design output to live site.

It runs projects on managed hosting instead of a manual deployment pipeline

When a Coderick AI project is published, it goes live on SiteGround’s hosting immediately. SSL installs automatically once a custom domain points to SiteGround’s nameservers. There’s no separate hosting account to set up, no deployment pipeline to configure, and no server to manage. The infrastructure is part of the product, not an afterthought.

It lets users edit and expand the project through chat

After launch, the project is still editable through the same chat interface. Users can request layout changes, add new pages or forms, adjust content, and extend functionality without leaving the builder or touching any code. For small businesses and solopreneurs who need to manage a site on their own over time, that’s the part that matters most.

When is Claude Code still the better option?

There are cases where Claude Code is clearly the right tool. These are not edge cases — they’re the situations Claude Code was built for.

You need source-code control

If you need to own every file and track every change made to the project, Claude Code is the right path. It works directly with Git, keeps a full history of changes, and gives you visibility into what was modified and when. 

You already have a codebase or development team

Claude Code can read and work across an entire existing codebase. If a project already has thousands of lines of code, an established directory structure, and a team of engineers collaborating on it, Claude Code fits into that workflow in a way a managed builder doesn’t.

You need custom integrations or complex engineering

Some projects need to connect to multiple external services, require users to log in with their own accounts, or run complex logic behind the scenes that a managed builder can’t handle. Claude Code gives  technical users the control to direct that work precisely.

You want GitHub, pull requests, tests, and manual deployment

For teams that run their delivery process through GitHub, manage code reviews through pull requests, and deploy on their own schedule, Claude Code integrates directly with those tools. The full developer workflow remains available. 

This isn’t unique to Claude Code. The same pattern shows up across coding agents and it’s why some users end up looking for OpenAI Codex alternatives for beginners for exactly the same reasons.

Which path should beginners choose?

The right tool depends on the goal, specifically, whether the goal is to work with code or to get something live.

Ready to build without the developer setup?

If the real goal is not learning a coding-agent workflow but launching a site, the more direct path skips the terminal entirely. Coderick AI is a managed AI website builder for non-technical users that handles the build, the hosting, the domain connection, and the ongoing edits inside one managed environment. Describe what you want, publish when it’s ready, and update it through chat when it needs to change.

[96]

Frequently asked questions about Claude Code alternatives

To start vibe coding, pick one small project, describe what you want in plain language, let an AI builder, like Coderick AI, create a first version, then test and refine it until the main result works. Starting small, with a landing page, a portfolio, or a simple web app as a first vibe coding project, will get you further faster than trying to build everything at once.

What should beginners know before they start vibe coding?

Below are the most important things to know about vibe coding for beginners before your first session.

Vibe coding is fastest when the project is small

Vibe coding is an approach to building software where you describe what you want in plain language and an AI tool generates the first version. The clearer your description, the better the result. And at least when you’re first starting with vibe coding, small projects are simply easier to describe clearly.

One beginner’s mistake is starting too big. A landing page has one clear job. A portfolio has a defined structure. A small calculator has a fixed number of inputs. Those boundaries are actually what make the first session manageable, as the AI has less to guess at, and you have less to review.

The first AI output is a draft, not the final version

The AI interprets your project brief, makes assumptions about gaps, and fills in details you didn’t specify. That’s exactly what it should do. Your job is to review those assumptions and correct the ones that don’t match what you actually wanted.

Expect the first version to be roughly right, not exactly right. The goal of that first generation is a working draft you can test and improve.

What should you review?

Content accuracy, layout across screen sizes, links, and the main action (a button click, a form submission, a calculation) all need a check before you move on.

Reviewing isn’t optional, even for simple beginner projects. Check that the content is accurate, the layout works on different screen sizes, the links go to the right places, and the main action (a button click, a form submission, a calculation) produces the correct result.

If you’re using Coderick AI, version control makes that process less pressured: you can restore any previous version if something goes wrong, so experimenting doesn’t carry risk. The visual editor also lets you click directly on elements to adjust them without writing a new prompt every time — useful when the fix is small and obvious.”

Not every project is a good beginner project

Some projects look simple on the surface but carry hidden complexity. Payments, private user data, login accounts, and third-party integrations like CRMs or booking platforms all come with technical and legal responsibilities that go beyond a first session. Tools like Coderick AI can handle many of these, but they work best as a second project, once you’re comfortable with the basic loop and confident in what you’re asking the AI to build.

Start with a project you can describe clearly and test on your own once it’s built.

How does vibe coding work step by step?

Vibe coding follows a short repeating loop: describe, generate, test, refine. Each pass tightens the result.

Describe what you want

The first step is writing your prompt: a clear description of the project. This isn’t a technical specification. It’s a plain-language brief that covers what you’re building, who it’s for, what it should do, and how simple the first version should be.

Generate the first version

The AI takes your description and builds a first draft. Depending on the tool and project, this might be a full web page, a working form, a functional calculator, or a basic web app layout. Review the output before you do anything else. Read what was built, click through it, and note what works and what doesn’t. This is the ground for your next steps.

Test what works

Testing means trying the project the way a real user would. Click the buttons, fill in the forms, resize the window. The gaps that appear during testing are the most valuable feedback you’ll get, because they show you where the AI interpreted your brief differently from what you intended. Note everything that needs improving; you’ll work through it one item at a time.

Refine one thing at a time

This is another step where most beginners go wrong. Go back to your notes and pick the single most important issue. Ask for that one change. Depending on the tool you use, you can either prompt it or make other changes. 

For example, with Coderick AI, you can describe the change you need in the chat, or, for something small like a text edit, click directly on the element and change it. One focused prompt produces a cleaner result than five problems lumped together — work through the list in order.

Your 30-minute vibe coding action plan

This is a practical first vibe coding session. It works regardless of what tool you choose (see how to choose the best one for your case) and the type of project: landing page, a portfolio, a lead form, or a small tracker. Do the steps in order.

1. Choose one small project. Decide what you’re building before you open any tool. Write it in one sentence.
2. Write one clear goal. What does success look like at the end of this session? “A live landing page with a working contact form” is a goal. “A good website” is not.
3. Add the audience and context. Who is the page for? What do they need to understand or do when they see it? Add that to your brief.
4. Generate the first version. Paste your brief into the tool and let it build. Don’t prompt again until you’ve read and clicked through the full result.
5. Test the main action. Find the single most important thing a user would do on this page (clicking a button, submitting a form, reading a price) and test it. Note what works and what doesn’t.
6. Ask for one focused improvement. Pick the most important gap from your test and write a focused prompt to fix it. One thing.
7. Save the prompt or version that worked. You’ll want to reuse the ones that produced the best result.
8. Decide the next step before ending the session. What would you improve in the next session? Write it down before you close the tab. You’ll thank yourself later.

What should your first vibe coding project be?

The best first vibe coding project has a clear goal, a simple structure, and a result you can judge on your own. The project should be simple enough that you know immediately whether it worked.

Simple business landing page

A landing page has one goal, one audience, and a handful of sections — headline, description, social proof, call to action. That structure is well-defined enough that AI tools produce a solid first draft quickly, and the quality is easy to judge: does it communicate what the business does, and does the call to action work?

Personal portfolio or service page

A portfolio or service page is a good beginner project because you already know the content. You don’t have to make up product details or invent testimonials. You can describe your work accurately, and the AI can focus on structure and layout rather than guessing at copy.

Lead capture form

A lead form has a clear outcome: someone fills it in or they don’t. That makes it straightforward to test and easy to measure. Keep the first version simple: a name, an email, maybe one other field. Avoid CRM or email integrations in the first session.

Small internal tracker

A tracker for personal use (tasks, expenses, a reading list, a simple habit log) is low-stakes and genuinely useful. Because you’re the only user, you set the rules and there’s no pressure to make it production-ready before you’re happy with it.

Simple calculator or generator

A calculator or generator has defined inputs, a defined output, and a result you can verify immediately. If the math is right, the project works. If it’s wrong, you know exactly what to fix.

First project options for beginners:

How do you choose your first vibe coding tool?

Vibe coding tool choice is a real decision, but it’s not the most important one. The project comes first. Choose your AI builder based on what you’re building, not based on which one has the most impressive demo or the longest feature list.

Choose by project type, not hype

Different tools are built for different outputs. Some let you preview the result in the browser and publish it live with one click. Others are built for backend logic or full-stack apps that expect you to run code in a terminal. 

Pick a tool that matches the output you need: a website, landing page, small web app, tracker, calculator, or form. If you don’t have a technical background, look for a tool where you write your prompt, it builds, and you’re looking at a live website. If a tool asks you to install something, write code, or configure a database and hosting separately, it’s probably not the right starting point.

Look for editing, preview, and publishing support

A useful vibe coding tool for beginners does three things: 

1. It shows you the result as it builds 
2. It lets you make edits without writing raw code 
3. It handles the publishing step without asking you to set up hosting separately

These are the baseline for a tool that actually fits a beginner workflow.

Why Coderick AI is a good first tool for websites and web apps

Coderick AI is built for non-technical users who want to go from description to live result without a developer setup in the middle. You describe your website or web app in plain language, the AI website builder generates a working version, and you can preview, edit, and publish directly from the interface.

For beginners building a landing page, portfolio, service site, or small web app, that workflow covers the full loop: describe, generate, test, refine, publish. Coderick AI also supports visual editing alongside the AI prompting interface, so you can click directly on elements to adjust text, colors, or layout without writing a new prompt every time. There’s no separate hosting account to configure and no deployment pipeline to manage. The published URL is ready as soon as you decide the project is done.

What are your first 3 vibe coding prompts?

The three beginner prompts below follow the describe-generate-refine loop. Keep each one focused. A shorter, clearer prompt almost always produces a better first result than a long one that covers multiple goals. If terms like prompt, constraints, first version, or refinement still feel new, this vibe coding glossary can help you check the basics before writing your first prompt.

Prompt 1: describe the project

The first prompt starts the session and sets the scope for every output that follows. Include the project type, the audience, the goal, and the main sections or features. Keep constraints in the prompt too: if you want the first version to be a single page, say so.

Here’s your first prompt formula:

Example: “Build a one-page landing page for a freelance copywriter. The audience is small business owners looking to hire someone for website copy and email campaigns. Include a short bio section, a list of services, two or three testimonials, and a contact form. Keep the design clean and professional.”

That’s enough for a solid first draft and here’s what Coderick AI generates from this first prompt:

Prompt 2: improve the first version

After reviewing the first output, pick one specific thing to change. Don’t list everything at once.

In our case, there are two competing CTAs in the hero section. “Start a Project” and “View Services” are both asking for attention, which splits the visitor’s focus at exactly the point where you want them to do one thing. If the goal is conversion, there should be one primary action.

Prompt 2 example: “Remove the ‘View Services’ button from the hero. The only action in that section should be ‘Start a Project’.”

One clear instruction, one specific area, one result to review. The result will be more accurate than asking for five changes in one go. And the result is clear:

Prompt 3: test and fix issues

After testing the main action (in this case, the contact form), report the exact issue.

Going through our example website, the placeholder text in the message field reads “Tell me about your project, timeline, and budget.” This asks for three things before the visitor has even decided to reach out.

For someone who just wants to ask a quick question, that level of detail can feel like too much commitment too soon and may stop them from sending the message at all. The fix is simple: describe the problem and the expected behavior in the same prompt.

Prompt 3 example: “Change the message field placeholder to ‘How can I help you?’ The current placeholder asks for a project, timeline, and budget upfront, which may put off visitors who aren’t ready to share that detail yet.”

That gives the AI both the fault and the fix target, and you’ll get a more accurate result than a vague “fix the form.”

And voila:

What beginner mistakes should you avoid?

Now that you have the core prompts down, here are the mistakes worth watching out for before they slow you down.

Starting with a project that’s too large

A project is too big for a first session when the scope has too many moving parts to evaluate clearly. Multiple user roles, conditional logic, or integrations before you’ve built the core feature are all signs to break it down. The project itself is something you can judge on your own once it’s built.

• Too broad: “Build a website for my freelance copywriting business with a homepage, services page, about page, blog, client portal, and a contact form with automated replies.”
• Better: “Build a one-page site for a freelance copywriter. The main goal is to get potential clients to send an inquiry. Include a short bio, three service cards, and a contact form that shows a confirmation message after submitting.”

Giving vague instructions

First, think through what you want, then describe the goal, audience, pages or features, and constraints in as much detail as you can to the AI. “Make it look nice” is not a prompt. “Change the background to off-white and make the headline font larger” is. The AI generates what you describe, which means vague input produces vague output. 

Changing too many things at once

After the first version, it’s tempting to list every issue and ask for them all at once. That produces messy results. One change at a time is slower per prompt but faster overall. You spend less time untangling results that fixed one thing and broke three others.

Publishing without testing

A working draft and a publishable project are not the same thing. Before sharing anything publicly, click every link, test every form, check the page on a phone, and read the content out loud. Things that looked fine in the build view sometimes reveal problems the moment a real person tries to use them.

Ignoring errors you don’t understand

If the AI generates an error message you can’t explain, don’t skip past it. Either ask the tool to explain what the error means, or (for anything beyond a basic website) ask a developer before continuing. Errors in the early build are easier to fix than errors discovered after the project is live.

When should you ask a developer for help?

Vibe coding covers a lot of ground, but there’s a clear line where the project scope exceeds what a beginner should manage alone. Knowing that line before you hit it saves time and avoids building something that creates real problems later. Let’s examine where to draw that line.

When money or private data is involved

Payments, billing, and financial data require security standards that go beyond a first vibe coding session. That said, tools like Coderick AI support Stripe integration through a restricted key, which means the payment processing stays on Stripe’s side: a simpler and safer setup than building payment logic from scratch. 

The line to watch is when you start storing, handling, or taking responsibility for financial data yourself. The same applies to any project that stores personal information about other people: customer records, medical details, private communications. These need proper security review, compliance checks, and infrastructure that a beginner project won’t have by default.

When login or permissions matter

The moment your project needs to restrict who can see what, separate admin access from regular user access, or let people log in at all, login and permissions move from a feature request to a security decision. 

You become responsible for what users can access, how their data is stored, and what happens if an account is compromised. And those aren’t questions a prompt can answer. They require proper planning, security review, and in most cases, a developer.

When the project runs a core business operation

Some projects need more than a great website — they need custom backend logic, specific infrastructure, or compliance requirements that go beyond the scope of any website builder. If that’s the case, bring a developer in early.

When you see repeated errors

One error is a prompt problem. Two related errors might be a logic gap. Three or more errors in the same area that keep returning after you ask for fixes are a sign the underlying structure has a problem that prompting can’t resolve cleanly. That’s a developer conversation.

When you need custom integrations

Connecting to a payment gateway, syncing with a CRM, pulling data from an external API: these integrations have authentication flows, error handling requirements, and rate limit considerations that add up quickly. Getting it right the first time is much easier with a developer who’s built the integration before.

Ready to build your first vibe coded project?

The first session teaches you more than any guide can: what context the AI actually needs, how much scope is too much for a single prompt, and how quickly a clear goal produces a usable first version.

Coderick AI is built for this beginner workflow: a clear brief, a working first version, and a direct path from draft to live without a developer setup in the middle. If you’ve got a project in mind and 30 minutes, that’s enough to have something real to show.

[96]

Frequently asked questions about getting started with vibe coding

Update, July 1, 2026: Following an official statement from Anthropic on June 30th, access to Claude Fable 5 has been restored. The US government had ordered Anthropic to suspend the model on June 12th while it investigated a potential security concern; that directive has now been lifted, and Fable 5 is available again across Claude.ai, Claude Code, Claude Cowork, and the Claude Platform. Mythos 5 has also been partially restored for a set of approved US organizations. All other Claude models were unaffected throughout. Read below to understand more about each model and how Fable 5 fits.

Anthropic just handed the public a model that was, until June 9, 2026, too powerful to release openly. Claude Fable 5 sits at the top of the Claude model family, built on the same foundation as the restricted Mythos 5 with safety guardrails added for general use. Whether it belongs in your workflow, though, is a different question.

What is Claude Fable 5?

Claude Fable 5 is Anthropic’s most capable publicly available AI model, launched June 9, 2026. It’s built for complex, long-running tasks that previous models struggled to sustain. It comes from a model tier Anthropic had, until now, considered too powerful to release openly.

What is Claude Mythos 5 and what does it have to do with Fable?

For the past year, Anthropic has been quietly developing a tier of models called Mythos: models powerful enough that the company considered them too capable for public release. The concern was specific. At Mythos-level reasoning, a model can identify and exploit security vulnerabilities and navigate sensitive areas of biology and chemistry with a degree of precision that raises real risks. Enticing, sure, but problematic.

Anthropic’s answer was to build two versions of the same model. Claude Mythos 5 stays restricted, available only to cyberdefenders, critical infrastructure providers, and select research partners through Project Glasswing. Claude Fable 5, on the other hand, is the same model with safety classifiers added on top, launched for general use on June 9, 2026. 

What are the safety filters and what happens when they trigger?

Because Fable 5 inherits Mythos-level reasoning, it comes with classifiers that monitor every request. So if a prompt touches restricted territory (such as detailed cyberattack tooling, dual-use biology, certain chemistry applications, or frontier AI model development), the model declines and for most of these categories automatically reroutes to Claude Opus 4.8, though behavior may vary depending on the specific topic flagged.

Anthropic reports this happens in fewer than 5% of sessions, so for normal business work you’re unlikely to ever see a refusal. One caveat: shortly after launch, some users reported false positives on completely benign prompts. Anthropic has acknowledged the issue and is working on fixes. If it happens, Opus 4.8 handles the task in the background and your workflow continues.

What’s actually different about Fable 5 — and what isn’t

Where Fable 5 pulls ahead of previous Claude models

The meaningful difference between Fable 5 and Opus 4.8 isn’t any single capability but rather how much further the model can run before it loses the thread. Previous models handle complex tasks well but tend to need check-ins. A long research project, a detailed document analysis, or a multi-step workflow will often require you to re-prompt, correct course, or fill in gaps. Fable 5 is built to sustain that work longer before that happens.

Specific areas where the gap is noticeable:

• Long-horizon autonomous work: Tasks that previously needed to be broken into chunks and supervised can run end-to-end. Anthropic’s benchmarks on agentic performance show Fable 5 consistently ahead of Opus 4.8 on tasks requiring sequential decisions across extended runs.
• Maintaining focus across long documents: Opus 4.8 could handle a 1-million-token context window, and so can Fable 5. But remembering early constraints, tracking long document threads, and avoiding contradictions across that full window is where Fable 5 shows a practical difference.
• Vision and document analysis: Fable 5 posts the strongest vision benchmark scores Anthropic has published. Reading dense technical graphs, extracting data from multi-page PDFs, and interpreting complex screenshots all produce more reliable results.
• Multi-step workflows: Tool use and agentic workflows, where Claude calls external services, processes the results, and continues working, are smoother. The model is less likely to stall when it hits an unexpected result.
• Coding: This is where Fable 5 really shines. It’s Anthropic’s strongest coding model to date. It can write, review, and debug complex code with fewer errors and less back-and-forth than previous versions, which makes a real difference if you’re building or maintaining a website, app, or any custom tool.

What Fable 5 still won’t do well

Fable 5 still needs a clear, well-defined task to produce good results. If your brief is vague, the output will be too. It won’t push back and tell you the question you asked is the wrong one, and it can’t replace judgment on decisions that require real-world context.

The safety classifiers can also produce unexpected refusals on benign prompts. And for anything routine, the cost premium is hard to justify when Sonnet or Opus will get you most of the way there for less.

How much does Claude Fable 5 cost?

If you’re on a Claude subscription plan

If you’re using Claude through claude.ai, Fable 5 comes with a temporary usage allowance rather than a flat inclusion. Pro, Max, Team, and select Enterprise plans include it for a limited portion of weekly usage before it shifts to usage credits (check Anthropic’s usage credits page for the current terms). Standard Enterprise seats don’t get an included allowance at all and need usage credits enabled from day one. Opus 4.8, Sonnet 5, and Haiku remain on standard plans without this kind of cap

If you’re paying per token via the API

For developers and businesses accessing Claude through the API, pricing is per million tokens. Fable 5 sits at the top of the scale at exactly twice the price of Opus 4.8, following the same 5x output-to-input ratio consistent across all models.

Source: Anthropic API pricing

*Sonnet 5 launched with introductory pricing of $2 per million input tokens and $10 per million output tokens, in effect through August 31, 2026. Standard pricing shown above applies after that date.

The Claude AI model family, and which one is right for your business

Claude is not one model. It’s a family, with each tier designed for a different balance of speed, cost, and capability. 

Not sure which one fits your needs? Here’s a quick guide, with more detail in the sections below.

If you’re also weighing Claude against other AI tools, our ChatGPT vs Claude vs Gemini guide covers the key differences across use cases.

When to use Haiku or Sonnet

Haiku is the right choice when you’re running high volumes of simple, structured tasks: classifying support tickets, tagging content, generating short product descriptions at scale. It’s fast and costs a fraction of the upper-tier models.

Sonnet 5 covers everyday business work and quite a bit more: drafting emails, writing blog posts, summarizing meeting notes, answering customer questions, building basic content. It’s also Anthropic’s most agentic Sonnet yet, closing much of the gap with Opus 4.8 on reasoning, tool use, and coding. 

For most small business owners using an AI assistant day-to-day, Sonnet is where to start, and it now handles a wider range of tasks than “start here” used to mean. Speed is good, quality is close to top-tier, and the cost is still the most manageable of the three paid tiers.

When Opus 4.8 is still the smart default

Opus 4.8 is the workhorse for work where accuracy and reliability matter most, short of a multi-day autonomous run. Detailed research reports, nuanced client documents, code reviews, and long-form analysis all fall comfortably within what Opus handles well. 

Sonnet 5 can take on a lot of this too now, so the honest dividing line is less about task type and more about how much margin for error you’re comfortable with. When the work is high-stakes enough that you’d rather not double-check it, Opus is worth the extra cost over Fable 5’s price.

If you’re weighing Claude Code pricing for a development project, Sonnet 5 will cover a large share of coding tasks at a lower cost, with Opus 4.8 as the step up when you need the extra reliability.

When Fable 5 is worth it

Fable 5 earns its cost when the task is genuinely large and benefits from running autonomously start to finish. Good candidates include:

• Researching and writing a comprehensive market analysis or business report from scratch, pulling from multiple sources without you stepping in between each step
• Running a full content audit across a large website and producing a prioritized action plan in one go
• Analyzing a large set of customer reviews, support tickets, or survey responses to surface patterns and recommendations
• Building out a multi-page proposal or pitch deck that requires synthesizing information from several documents

If the task could be completed with a few rounds of Opus 4.8 prompts, Fable 5 is probably not the right spend.

How to access Claude Fable 5

Claude Fable 5 is generally available across all of Anthropic’s main surfaces: Claude.ai (web and mobile), Claude Code, and Claude Cowork. It is also available through the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.

The bottom line: matching the right Claude model to your needs

Claude Fable 5 is a genuinely significant release, but alas “significant” doesn’t mean “right for everyone.” For most small businesses, the model to reach for daily is Sonnet 5, with Opus 4.8 as the upgrade for anything where reliability matters more than cost. Fable 5 earns its place on tasks that are genuinely large, long-running, and autonomous: the kind of work where a skilled contractor completing a project unsupervised over several days is the right mental model.

For the day-to-day work that makes up most of a small business’s AI needs, SiteGround AI Studio gives you access to ChatGPT, Claude, Gemini, and more than 15 specialized AI agents for SEO, social media, email marketing, and WordPress — all in one workspace, under one subscription.

[97]

Frequently asked questions

Random prompts get random results. A vibe coding workflow gives you a repeatable way to work with Coderick AI and other AI website and web app builders so each session builds on the last instead of starting from scratch.

A vibe coding workflow is a structured process for working with AI tools: you define one outcome, give the AI the right context, review the first output, refine with focused follow-up prompts, and document the decisions that shaped the project. The goal is a consistent, reviewable build process instead of a prompt session that spirals into something you can’t explain, repeat or undo.

What is a vibe coding workflow?

A vibe coding workflow consists of the following points: 

1. А defined start (the goal)
2. А defined end (a result that meets your success criteria)
3. А series of steps in between that keep the session on track

Workflow vs. random prompting

Random prompting is an unstructured approach: you open an AI tool, type whatever comes to mind, and evaluate the result by feel. There’s no defined goal, no way to judge whether the output is right, and no record of what you asked for. When the result is off, you rewrite the prompt from scratch and hope the next version lands closer.

Here’s a quick example of random prompting: 

“Build me a website for my HVAC business.” Result is off. “Make it better.” Still off. “Add more stuff.” 

A vibe coding workflow moves through defined phases, as covered earlier in this guide. Here’s what that looks like in practice:

Step 1. Describe: “Build a one-page site for a residential HVAC company. Main action: service request form. No ecommerce, no blog. Clean layout, phone number at the top.”

Step 2. Review: Result is close, but the form is below the fold and the heading is too generic.

Step 3. Refine: “Move the form above the fold.” Done. “Rewrite the heading to focus on same-day repairs.” Done.

Three prompts, one clear outcome. That’s the workflow.

Start with one outcome before you prompt

The clearest prompt you’ll ever write comes from a goal you’ve already defined on paper. Spend five minutes here and the rest of the session gets easier.

1. Name the thing you want built

Be specific about the deliverable. “A website” isn’t a deliverable. “A three-page website for a yoga studio with a class schedule, a contact form, and an About page” is. The name of the thing sets the scope for everything that follows.

2. Define who it’s for

The audience shapes every decision the AI makes: the language it uses, the features it includes, the layout it defaults to. “Homeowners needing HVAC repairs” and “facilities managers for commercial properties” are both potential audiences for the same type of service, but they expect very different experiences.

3. Write the success criteria

A success criteriа is a check you can run after the first output to decide whether it worked. It doesn’t need to be long. “The main call-to-action button links to a contact form” is a success criterion. “The page looks good” isn’t.

Two or three criteria is enough for a first session. The goal isn’t to define everything the project will ever need; just enough to judge whether the first output is on track.

4. Set the first-version boundary

The first version doesn’t need to include everything. A smaller scope produces a cleaner output and gives you a clearer baseline to build from. Decide what the first output must include and what it doesn’t need yet. This stops the AI from adding features you didn’t ask for and gives you a clean baseline to review.

What context should you give before prompting?

Context is what separates a useful first version from one that needs to be thrown away. The AI builds from the context you provide. If the description is incomplete, the output will fill the gaps with guesses.

Business context

Tell the AI what kind of business or project this is for. “Residential HVAC company serving suburban homeowners” gives the AI enough to make reasonable decisions about tone, layout, and functionality.

Audience context

Describe who will use the result. Age, technical confidence, and the specific thing they need to do all matter. “Homeowners who aren’t especially tech-savvy and want to book a repair quickly” is useful. “Regular people” is not.

Style and brand context

If there are visual or tonal constraints, say them upfront. Color preferences, an existing logo, a formal or casual voice. Without this context, the AI is more likely to produce generic output.

Feature context

List what the first version needs to include. The goal here isn’t to define the whole project — it’s to get a reviewable first draft. Three to five features gives the AI enough to build something concrete without producing an output that’s too complex to evaluate in one sitting. More detail can come in follow-up prompts once the structure is right.

Constraints and must-not-do items

What the AI should avoid is just as important as what it should include. No ecommerce. No user accounts. No multi-page navigation for now. These constraints prevent the AI from solving problems you didn’t ask it to solve.

How should your prompting workflow work?

A prompting workflow has four moves: create, review, fix, and test. Each one has a job. Don’t combine them in a single prompt.

Prompt 1: create the first version

The first prompt is the broadest prompt in the workflow. Include the goal, the audience, the key context, and the first-version boundary. Ask for the first complete version. Don’t hold back important context here because you’re trying to keep the prompt short.

A first prompt for Coderick AI might look like: 

“Build a one-page website for a residential HVAC company. The audience is homeowners in suburban areas. The main action is filling out a service request form. No ecommerce, no blog, no user accounts. Clean layout with a phone number visible at the top.”

And here’s the first result that Coderick AI generates:

Prompt 2: Review and improve structure

Before refining details, check whether the structure is right. Is the layout doing what you wanted? Is the information in a sensible order? Is the main action easy to find? A structural problem is harder to fix later, so catch it now.

Prompt 3: Fix one issue at a time

Once the structure is right, move to specifics. One change per prompt. “Move the contact form above the fold” is a good prompt. “Move the contact form, change the heading text, add a testimonial section, and make the buttons blue” is four prompts crammed into one, and the AI will handle some of them well and others poorly.

In our case, if we want the service request form to be the only action a visitor can take in the hero section, we’d use this prompt:

“Remove the phone number button from the hero. The only primary action should be the service request form.”

Prompt 4: Test the main user action

Before you call the project done, test the one thing the project is supposed to do. For a booking site, that means going through the booking flow. For a landing page, that means reading it as a first-time visitor. What’s missing or broken?

Document what changed

AI sessions have no memory between prompts. When a session runs long, it’s easy to lose track of what you changed, why you changed it, and which version actually worked. A one-sentence note after each significant prompt — what you asked for and why — is enough to keep the session recoverable.

For example: “Removed the pricing section because it confused the target audience in testing.”

A session that runs four prompts deep already has three decisions worth keeping: what you built, what you changed, and why. Without a record, the next session starts from scratch. With one, it starts from where you left off.

In Coderick AI, every saved state is stored in version history, which means you can roll back to any earlier version if a change takes the project in the wrong direction. Pair that with a short decision log and you have two layers of recovery.

How do you review the first AI output?

Reviewing the first AI output is a specific skill, not just a gut check. Work through a short checklist before you ask for changes.

✅ Check if it matches the original goal

Does the output match what you asked for in the first prompt? Not whether it looks good, but whether it does the job you defined. A beautiful website that doesn’t have a contact form failed the brief.

✅ Check what’s missing

What did you ask for that isn’t there? Go back to your first-version boundary and compare it to the output line by line.

✅ Check what’s wrong

What’s present but incorrect? Wrong copy, wrong layout, wrong tone. Note each issue separately so you can fix them one at a time.

✅ Check what’s too complicated

What did the AI add that you didn’t ask for? AI tools tend to add features. Anything you didn’t request is a candidate for removal before the project gets bigger.

✅ Check what needs human judgment

Some things the AI can’t decide for you: 

• Whether the copy sounds like your brand
• Whether the color palette fits your audience
• Whether the flow makes sense to someone who doesn’t know the product 

Mark those for a manual review pass.

How does prompt refinement work?

Prompt refinement is the discipline of improving the AI output with targeted follow-up requests. It’s not rewriting the whole brief from scratch. It’s asking for one specific change, checking whether it worked, and deciding what to do next.

For example: instead of going back to the AI with “redo the whole page, it’s not quite right,” you’d say “move the service request form above the fold.” Check the result. Then: “rewrite the heading to focus on same-day repairs.” Check again. 

Ask for one change at a time

This is the most important rule in the refinement loop. One change per prompt keeps the output predictable and the session recoverable. When you ask for three things at once, the AI changes all three but you can only evaluate one at a time, so you don’t know which change caused the result you’re looking at.

Reference the previous output

Tell the AI what it just built. “In the version you just created…” or “Keep everything from the last version except…” grounds the follow-up prompt in a specific state. Without this, the AI may drift back to defaults.

Give the reason for the change

“Simplify the navigation because users are arriving from mobile and need to reach the contact form in one tap” is a better prompt than “simplify the navigation.” The reason helps the AI make a proportionate change instead of guessing what level of simplification you want.

Avoid rewriting the whole request

Rewriting the whole prompt from scratch restarts the context window and loses decisions you already made. The AI forgets the decisions you’ve already made. Work from the existing output, not from a blank slate.

Stop when the output meets the success criteria

Go back to the success criteria you wrote before the first prompt. If the output clears them, stop. You don’t need perfection. You need something that works well enough to test with real users or review with a real audience.

Why prompt documentation matters

Prompt documentation is the habit of recording what you asked for, what changed, why it changed, and what you kept. It’s the difference between a session you can hand off, restart, or learn from and one that only makes sense in the moment.

Save the prompt that worked

When a prompt produces a result you want to keep, save it. Save it somewhere outside the AI session. The session context gets messy over time, and what worked in turn 3 is easy to lose by turn 15.

Record what changed

Note every significant change you made and why. Not every tiny adjustment, but every decision that affected the direction of the project.

Track decisions you don’t want to lose

Some decisions aren’t about prompts, they’re about the project. “Decided to remove the pricing page because it confused the target audience” is a decision worth recording even if it happened in a conversation, not a prompt.

Keep a short changelog

A changelog doesn’t need to be formal. A date, a description, a reason, and a result is enough. This is the pattern that makes sessions restartable.

Context rot is what happens when a long AI session accumulates so many follow-up prompts and partial rewrites that the AI’s responses stop being consistent with the original goal. A changelog is the best defense against it because it lets you restart a clean session with a summary of the goal, decisions, and current version.

Restart with clean context when needed

When a session gets long enough that the AI starts contradicting earlier outputs or forgetting constraints, start a new session. Paste in a short summary of the goal, the decisions made so far, and the current state of the project. You’ll get cleaner results from a fresh context than from a conversation that’s been running for fifty turns.

When should you stop prompting and start reviewing?

There’s a point in every session where more prompting makes things worse, not better. Knowing when you’ve hit that point is part of the vibe coding workflow.

Stop prompting when:

• The same issue keeps coming back after you’ve addressed it twice
• The output from the last prompt is worse than the output from the previous one
• You can’t clearly explain what change you’re about to ask for
• The next change involves a judgment call that only you can make
• The first version is good enough to test with a real audience

At any of these points, the right move is to stop prompting and review manually, test with a real user, or start a new session with a cleaner brief. For higher-risk builds, run the security checks before publishing.

A reusable vibe coding workflow template

Use this template at the start of any new project. Fill it in before you open Coderick AI or any other AI website and web app building tool.

Put the workflow into practice with Coderick AI

Everything in this vibe coding workflow maps directly to how Coderick AI is built. You describe the goal, Coderick AI generates the first version, and you refine the project without switching tools or managing infrastructure.

The Visual Editor handles credit-free refinements for design changes you want to make without reprompting. Version history works as your decision log: every saved state is a checkpoint you can roll back to or reference when you start a new session. Publishing happens inside Coderick AI, with hosting and SSL included.

The result is that the full workflow: plan, prompt, review, refine, document, test, publish — happens in one place. No separate tools, no technical decisions before you can start.

Test the full workflow free for 14 days.

[96]

Frequently asked questions about vibe coding workflow