# Build Something (/claude-and-chatgpt-2/build-something/overview) --- type: Document title: "Build Something" description: "The fifth section of Claude and ChatGPT 2.0. How to build with Claude's six artifact types, Docs, Slides, Sheets, Design, Data and Whiteboard, and with the ChatGPT tools that do the same jobs, before you use them in the cases." status: stable order: 0.45 ksor: owner: team:panaversity audience: [ public ] approval: by: process:panaversity at: 2026-11-11T15:26:22Z chapter: "F00F" part: Front expert_status: required objectives: [] word_budget: null prerequisites: [] build_step: null artifact: null field_guides: [] delta_entries: [] concepts: [ "Artifacts", "Docs", "Slides", "Sheets", "Design", "Data", "Whiteboard" ] last_verified: 2026-11-11 review_by: 2026-12-11 sources: - id: anthropic-artifacts title: "What are artifacts and how do I use them? (Anthropic, verified 2026-11-11)" resource: https://support.claude.com/en/articles/17153992-what-are-artifacts-and-how-do-i-use-them - id: anthropic-share title: "Share artifacts (Anthropic, verified 2026-11-11)" resource: https://support.claude.com/en/articles/9547008-share-artifacts - id: claude-launch title: "Introducing Claude Whiteboards, Claude Sheets, and Claude Data (Anthropic, November 10, 2026)" resource: https://claude.com/resources/articles/sheets-whiteboards-data - id: claude-code-artifacts title: "Share session output as artifacts (Anthropic, verified 2026-11-11)" resource: https://code.claude.com/docs/en/artifacts - id: openai-space title: "Getting started with Space in ChatGPT (OpenAI, verified 2026-11-11)" resource: https://help.openai.com/en/articles/20001549-getting-started-with-space-in-chatgpt generated: at: 2026-11-11T15:26:22Z by: process:panaversity trust_tier: unverified build_id: sha256:a2698d592c63f6c6c789f4b01ca90bdbf64ca64800918c9f09b0391aaddbda67 dirty: true ksor_version: 0.0.60 --- *Six things you can build in Claude, and what ChatGPT offers* ## Where this section fits The earlier sections showed that an AI Worker gives you finished work, not just an answer. This section shows you how to make that finished work yourself. It covers six kinds, in Claude and in ChatGPT. After this section come the case studies. In them you write, check and share finished work, such as a brief, a draft and a grading sheet. The habits on these pages are the ones the cases need, so do this section first. > [!NOTE] > **Words used in this section** > > - **Artifact:** a piece of finished work that Claude makes, such as a document or a slide deck. It opens next to your chat. You can change it, come back to it and share it. > - **Export:** to save a copy as an ordinary file, such as a Word file. > - **Beta:** a new feature that works but is not finished yet. It may change. > - **Connected app:** another app you have linked to Claude, such as your calendar or email. > - **Space and Work:** two parts of the ChatGPT app. **Space** keeps your shared pages and decks. **Work** is where ChatGPT does longer jobs for you. > - **Plugin:** an add-on that lets ChatGPT do something new. > - **Formula:** a rule in a spreadsheet cell that calculates something, such as a total. ## Use the desktop apps This book works in the **Claude desktop app**, on a Mac or a Windows computer, and in the **ChatGPT desktop app** for the ChatGPT parts. [The two apps side by side](../new-way-of-working/the-two-apps.md) listed both under what you need to begin. Every step in this section is written for them. - **Claude.** The same steps also work in a web browser, because Anthropic gives the web and the desktop app the same artifact tools.[^anthropic-artifacts] But only the desktop app can read a folder on your computer. On a phone, you can ask for a doc, a deck or a design in a chat and look at the result. - **ChatGPT.** Use the new desktop app, the one with Chat, Work and Codex. Each page says where a tool differs on the desktop. ## Start here: the big idea When Claude makes an artifact, it saves it in the **Artifacts** tab, in the menu on the left side of the Claude desktop app. It stays there, so you can find it again tomorrow.[^anthropic-artifacts] If you do not see the Artifacts tab, open Settings, then Capabilities, and turn on **Artifacts** and **Create and edit files**.[^anthropic-artifacts] To start one, open the Artifacts tab. Under **Make something new**, you see one big button, called a **tile**, for each kind: ![Six things you can build. Six colored cards in two rows. Docs: a document people read and edit; in ChatGPT, Pages, in Space. Slides: a deck to present; in ChatGPT, Slides, in Space. Sheets: a shared table with formulas; in ChatGPT, inside Excel and Google Sheets. Design: something judged by how it looks; in ChatGPT, Sites, which are working web pages. Data: charts from your company's data; in ChatGPT, the @Data plugin in Work. Whiteboard: a board for thinking together; in ChatGPT, no built-in board yet. At the foot: build a draft with the AI, improve it together, check it before anyone sees it.](img/six-things-you-can-build.png) Which tiles you see depends on your plan:[^anthropic-artifacts],[^claude-launch] | Your Claude plan | What you can build | | --- | --- | | Free | Docs, Slides and Design | | Any paid plan | All six. Sheets and Whiteboards are still in beta, and on Enterprise an owner turns them on | In ChatGPT, most of these tools need a paid plan. Which plans differs from tool to tool, so each page says.[^openai-space] One sentence holds the whole idea: > **Pick the tool that fits the job. Make a draft with the AI. Improve it together. Check it before anyone else sees it.** ## Four habits for every kind 1. **Ask for the kind you need.** Say what you want, who it is for, and what to use. If you need an ordinary file, such as a Word file, make the artifact first, check it, and export it at the end. A file is only a copy: if you fix the artifact later, the file does not change.[^claude-code-artifacts] 2. **Improve it together.** Fix small things yourself. Ask the AI for big changes. 3. **Share it on purpose.** Only you can see a new artifact. You choose who can open it, and what they can do. On the Free plan, sharing may be limited, as the [Docs](docs.md) page explains.[^anthropic-share] 4. **Check it before anyone else sees it.** It looks finished, but it can still be wrong. Check the numbers, the names and the facts. Remove anything private. Ask yourself: would a reader understand it without my chat? > [!WARNING] > **Two people can see different data in the same artifact.** Some artifacts read your connected apps. When a colleague opens one, it reads the colleague's own apps, not yours. So each person sees only what they are allowed to see.[^anthropic-share] This protects only what an artifact reads live. Anything already copied into it, such as a number in a doc, everyone who can open it can read. The [Data](data.md) page explains the one exception. Also, open shared artifacts only from people you trust. A page can hide instructions that try to trick an AI, as [Staying safe](../behind-the-screen/staying-safe-a-locked-office-and-a-second-signature.md) explained. ## Artifact, file, connected app or website? Four words in this section sound alike, but they do different jobs. In the examples, Dave is Brightline's controller: the manager in charge of Brightline's money, who approves payments. | You need | Use | Brightline example | | --- | --- | --- | | Work you and Claude keep improving, shared by a link | An **artifact** | The payment-run doc that Dave comments on | | A fixed copy to send, print or keep | An **exported file** | A PDF of the version of the doc that Dave reviewed | | Claude to read or change information in another app | A **connected app** | Claude reads this week's invoices from a Google Drive folder | | A working web page that other people use | A **website**, such as a ChatGPT Site | A page where suppliers check when they will be paid. A real one needs a sign-in, so each supplier sees only its own payments | A short example: on Thursday, Dave wants to comment on the payment run, so you share the **doc**. On Friday he approves it, so you export a **PDF** of the version he approved. The PDF is a fixed copy, and the doc can keep changing. The PDF does not prove the approval itself. That needs its own record, such as Dave's sign-off in the payment system. ## Build on what you already know The earlier sections taught you the other parts of an AI Worker: the borrowed desk in the cloud, your own folders, connected apps, skills, projects and scheduled tasks. You do not leave them behind here. They make your artifacts better. First, one link between them. A cloud job runs on a borrowed desk, and the desk is cleared when the job ends, as [Why the borrowed desk gets cleared](../behind-the-screen/why-the-borrowed-desk-gets-cleared.md) showed. An artifact is kept by Anthropic, outside that desk, in your Artifacts tab.[^anthropic-artifacts],[^claude-code-artifacts] So **an artifact is one way finished work leaves the borrowed desk and stays.** A downloaded file, or a file saved to Google Drive, also stays. ![Build on what you already know. On the left, five earlier parts: your folders, read by the desktop app; connected apps, such as Google Drive; a project, your rules written once; a skill, your method; a scheduled task, which decides when the job runs. An arrow leads to a dashed box in the middle, Claude does the job: a cloud job runs on a borrowed desk, a temporary computer that reads, calculates and writes, and the desk is cleared when the job ends. An arrow leads to a box on the right, the artifact stays: a doc, deck, sheet, design, dashboard or board, kept in your Artifacts tab, outside the desk. You keep it, you check it, you share it. At the foot: the earlier parts decide what Claude can reach and when it works.](img/build-on-what-you-know.png) Each of the next six pages adds one earlier part to its artifact: | You learned | Add it, and | Practice it on | | --- | --- | --- | | **Projects** | You write your rules once, and Claude reads them in every chat in the project | [Docs](docs.md) | | **Connected apps** | Claude reads an app while it builds. What it read is a snapshot of that moment | [Slides](slides.md) | | **Your folders,** through the desktop app | Claude builds from files on your computer. The artifact holds a copy | [Sheets](sheets.md) | | **Skills** | Claude can follow your method when a job matches the skill, and a design system keeps your look | [Design](design.md) | | **Scheduled tasks** | A job updates a doc you already made, at a set time | [Data](data.md) | | **Projects, again** | A plan on a board becomes jobs that run side by side | [Whiteboard](whiteboard.md) | You find each one in a short **Combine it** part, just before the page's **Try it** task. They are optional: the Try it tasks and the final exercise work without them. Some need a paid plan or a connected app. ChatGPT has the same parts under similar names: projects, plugins, skills, scheduled tasks and your own files through its desktop app, as [ChatGPT 2.0](../tour/chatgpt.md) showed. > **Rule to remember:** the earlier parts decide what Claude can reach and when it works. The artifact is what you keep, check and share. Start with [Docs](docs.md). The last page, [One job, three deliverables](one-job-three-deliverables.md), puts everything together in one exercise, with an optional part that uses these earlier parts too. [^anthropic-artifacts]: As verified on November 11, 2026. Anthropic says that on "Claude on the web and Claude Desktop" you can "create, edit, and share artifacts, and start from a template", and that on a phone you can ask for a design, deck or doc in any chat and view the result. Anthropic says artifacts need Create and edit files turned on, and in the app on November 11, 2026, the Artifacts switch said it is required for Create and edit files. On Team and Enterprise plans, an owner turns this on in the organization's settings. On Enterprise plans, Sheets and Whiteboards stay off until an owner turns them on. Some company plans also have Motion, for short animations, which this section does not cover. What are artifacts and how do I use them?, Anthropic. [^anthropic-share]: As verified on November 11, 2026. Share artifacts, Anthropic. [^claude-launch]: Sheets and Whiteboards arrived on November 10, 2026, when the older Dashboards tool was renamed Data. Introducing Claude Whiteboards, Claude Sheets, and Claude Data, Anthropic. [^claude-code-artifacts]: As verified on November 11, 2026, Anthropic's guide to the Code tab also says that to get a file instead of a doc, you can name the format, such as .docx. It says artifact content "is stored on Anthropic-operated infrastructure". Share session output as artifacts, Anthropic. [^openai-space]: Getting started with Space in ChatGPT, OpenAI.