export declare const MARKDOWN_DEMO_CONTENT = "# What is Spark Design\n\n**Spark Design** is a React design system for AI-native interfaces. It provides production-ready primitives, chat patterns, and theme tokens that fit into the same workspace as your app.\n\n## Core capabilities\n\n- **Context-aware**: Understands your needs from open files and the current project\n- **Multimodal**: Handles text, code, diagrams, and structured output\n- **Executable**: Runs code and tools in a controlled environment, from $0 \\rightarrow 1$\n\n---\n\n## Why \u201CCowork\u201D\n\n> Traditional conversational AI is more like \u201Cadvice over the shoulder\u201D\u2014you describe the problem, it suggests steps.  \n> Cowork aims to **sit at the same table**: edit docs together, write code together, move tasks forward together.\n\nUsing a blockquote like this separates \u201Cthinking / context\u201D from \u201Cfinal answer\u201D and keeps long replies easy to scan.\n\n### How it differs from regular Claude\n\n| Dimension  | Regular Claude       | Spark Design              |\n|-----------|----------------------|--------------------------------|\n| Mode      | Turn-by-turn Q&A     | Ongoing edit in one workspace |\n| Context   | Mostly the thread   | Docs, code, and project        |\n| Output    | Text and code blocks | Direct edits in your files    |\n\nThe table above is **responsive**: on narrow screens it scrolls horizontally via `overflow-x: auto` (inline code here) so the layout doesn\u2019t break.\n\n### Typical use cases (ordered list)\n\n1. **Clarify requirements**: Turn PRDs or notes into concrete, executable steps\n2. **Code together**: Complete, refactor, or fix bugs in real projects instead of \u201Cpaste an example\u201D\n3. **Docs and knowledge**: Draft specs, API docs, meeting notes\n4. **Task flow**: From idea to deliverable, $\\frac{idea}{time} \\rightarrow deliverable$\n\n---\n\n## What happens in a session (H4 subheadings)\n\n#### Execution-log style (H4 / H5)\n\nUse smaller weight and muted color so \u201Cprocess\u201D doesn\u2019t compete with the main answer.\n\n##### View steps (H5)\n\n- [x] Permission and context checks\n- [x] Resolve workspace path and open files\n- [ ] Generate or edit according to your instructions\n\nAbove is a **task list**: checked = done, unchecked = in progress\u2014good for steps and checklists.\n\n---\n\n## Code and math examples\n\nInline code like `cowork.plan()` and `cowork.edit(file, diff)` sits in the sentence; fenced code keeps monospace and line height for readability:\n\n```python\n# Spark Design concept (pseudo-code)\ndef cowork_session(context: WorkspaceContext, user_goal: str):\n    plan = cowork.plan(goal=user_goal, context=context)\n    for step in plan.steps:\n        if step.kind == \"edit_file\":\n            cowork.edit(step.target, step.diff)\n        elif step.kind == \"run_command\":\n            cowork.run(step.command, cwd=context.root)\n    return cowork.summary()\n```\n\nInline math: from idea to delivery can be written as $\\mathbb{E}[delivery] = f(idea, context, time)$.  \nDisplay math (when KaTeX is enabled) can be centered and on its own line for longer derivations.\n\n---\n\n## Summary\n\n- Use **H1** only for the main title; **H2 / H3** for sections; **H4\u2013H6** for steps or log-style subheadings.\n- **Blockquotes** for \u201Cthinking / premise\u201D; body text for the actual answer.\n- **Task lists** for progress and checklists; **tables** for structured comparison; **code** and **math** for clarity and rigor.\n\n*Spark Design is designed so AI-native UI moves from \u201Ca generic chat demo\u201D into \u201Cthis file and this product you are editing\u201D with production-ready tokens and components.*\n";