tldraw 创始人 Steve Ruiz 探讨为何无限画布是人机协作的强大基础，并分享他从美术到构建画布工具的历程。
Steve Ruiz, founder of tldraw, discusses why the infinite canvas is a powerful primitive for human-AI collaboration, sharing his journey from fine art to building a canvas that powers many startups.
要点 · TL;DR
无限画布是实现多智能体与人类实时协作的关键界面。 Infinite canvas is a pivotal interface for real-time multi-agent and human-AI collaboration.
好工具在精确与灵活间取得平衡，支持安全实验和快速迭代。 Good tools balance precision and flexibility, enabling safe experimentation and rapid iteration.
画布上的 AI 智能体受益于视觉区分、状态表示和空间上下文。 AI agents on canvas benefit from visual differentiation, state representation, and spatial context.
核心观点 · Key points
无限画布是人机协作的强大基础，支持实时多智能体和多人协同工作。 The infinite canvas is a powerful primitive for human-AI collaboration, enabling real-time multi-agent and multi-human work.
好的工具在精确性和灵活性之间取得平衡，允许安全实验和快速迭代。 Good tools balance precision and flexibility, allowing safe experimentation and rapid iteration.
画布惯例（如捏合缩放和撤销）至关重要；违反它们会让工具感觉有缺陷。 Canvas conventions like pinch-to-zoom and undo are critical; breaking them makes a tool feel broken.
画布上的AI智能体受益于视觉区分、状态表示和空间上下文。 AI agents on canvas benefit from visual differentiation, state representation, and spatial context.
像“精灵”这样的隐喻加速了智能体系统的构思，使复杂的编排变得直观。 Metaphors like 'fairies' accelerate ideation for agentic systems, making complex orchestration intuitive.
反共识 · Contrarian takes
画布不仅用于设计；它是智能体（包括AI）的通用协作环境。 Canvas is not just for design; it's a universal collaboration environment for intelligences, including AI.
当前画布上的AI智能体“糟糕但惊人”——精度差但展示了未来潜力。 Current AI agents on canvas are 'shitty but amazing'—bad at precision but show future potential.
画布上的多智能体编排模仿视频游戏AI，而非传统软件模式。 Multi-agent orchestration on canvas mirrors video game AI, not traditional software patterns.
像“make real”这样的画布工具早于vibe coding，证明视觉提示是自然界面。 Canvas-based tools like 'make real' predate vibe coding, proving visual prompting is a natural interface.
将AI称为“虚拟协作者”是错误的；它们更像“精灵”——非人类、古怪且有限。 Framing AI as 'virtual collaborators' is wrong; they are more like 'fairies'—non-human, quirky, and limited.
本期章节 · Chapters（共 28）
引言：无限画布与AIIntroduction: The Infinite Canvas and AI
史蒂夫背景：从美术到设计Steve's Background: From Fine Art to Design
早期职业生涯与原型学习Early career and learning prototyping
状态机与状态设计器State Machines and State Designer
箭头问题与完美箭头The Arrow Problem and Perfect Arrows
开源与社区反响Open Source and Community Response
反思Framer内容创作Reflection on Content Creation at Framer
发现绘图工具需求Discovering the need for a drawing tool
痴迷难题与学习Obsession with hard problems and learning
工具作为决策辅助Tools as decision-making aids
创新与惯例的张力Tension between innovation and convention
识别应用差异Identifying differences in apps
Dive人才网络广告Dive talent network ad
独特细节层次Unique level of detail
画布范式趋势Canvas paradigm trend
设计工具的更多可能More possibilities in design tools
工作流作为画布子集Workflows as a subset of canvas
多人画布潜力Multiplayer canvas potential
Figma收购影响Figma acquisition impact
画布编码vs传统工具Coding on a canvas vs. traditional tools
画布作为AI协作环境Canvas as a collaboration environment for AI
自动补全与早期演示Autocomplete and early demos
将AI比作精灵Framing AI as fairies
可视化智能体状态Visualizing agent state
多智能体编排与涌现行为Multi-agent orchestration and emergent behavior
智能体设计中的微观决策Micro decisions in agentic design
隐喻作为设计工具Metaphor as a design tool
实时人机协作挑战Challenges of human-AI collaboration in real-time