← 作品← Works← Works

实验档案Experiments ArchiveExperiments Archive

当下探索Current ExplorationCurrent Exploration

我现在在探索什么?What am I exploring now?What am I exploring now?

我现在在探索什么?What am I exploring now?What am I exploring now?

方向DirectionsDirections

AI 工作流设计AI workflow designAI workflow design基于 Codex 的实验Codex-based experimentationCodex-based experimentationAgent 系统Agent systemsAgent systems人机协作Human + AI collaborationHuman + AI collaboration内容工作流Content workflowsContent workflows研究工作流Research workflowsResearch workflows迭代循环Iteration loopsIteration loops个人知识系统Personal knowledge systemsPersonal knowledge systems

背景ContextContext

当下的探索聚焦于:AI 如何在不取代人类判断的前提下,支持思考、制作、测试与迭代。The current exploration focuses on how AI can support thinking, making, testing, and iteration without replacing human judgment.The current exploration focuses on how AI can support thinking, making, testing, and iteration without replacing human judgment.

重点不是把 AI 当作新鲜事物来使用,而是设计这样的工作流:人的意图与机器的协助,能以实际、可检视的方式协同工作。The focus is not on using AI as a novelty. It is on designing workflows where human intention and machine assistance can work together in practical, inspectable ways.The focus is not on using AI as a novelty. It is on designing workflows where human intention and machine assistance can work together in practical, inspectable ways.

挑战ChallengeChallenge

AI 工具让生产更快,但只有速度并不能带来更好的工作。AI tools make production faster, but speed alone does not create better work.AI tools make production faster, but speed alone does not create better work.

挑战在于设计这样的系统:实验变得更容易,同时判断依然可见:The challenge is to design systems where experimentation becomes easier while judgment remains visible:The challenge is to design systems where experimentation becomes easier while judgment remains visible:

  • 什么应该由人来决定?What should the human decide?What should the human decide?
  • 什么可以让机器起草、测试或比较?What can the machine draft, test, or compare?What can the machine draft, test, or compare?
  • 反馈应该从哪里进入循环?Where should feedback enter the loop?Where should feedback enter the loop?
  • 如何让想法更容易被验证?How can ideas become easier to validate?How can ideas become easier to validate?
  • 工具如何支持品味,而不是把它磨平?How can tools support taste instead of flattening it?How can tools support taste instead of flattening it?

思考ThinkingThinking

最有趣的 AI 工作流并不是全自动流水线。The most interesting AI workflows are not fully automated pipelines.The most interesting AI workflows are not fully automated pipelines.

它们是协作系统:人的判断设定方向,AI 扩大可能选项的范围,迭代变得更便宜。They are collaborative systems where human judgment sets direction, AI expands the field of possible options, and iteration becomes less expensive.They are collaborative systems where human judgment sets direction, AI expands the field of possible options, and iteration becomes less expensive.

这让测试更多想法、比较不同方案、发现规律、改进最终结果,都变得更轻。This makes it easier to test more ideas, compare alternatives, notice patterns, and improve the final result.This makes it easier to test more ideas, compare alternatives, notice patterns, and improve the final result.

观察ObservationObservation

AI 不取代判断。AI does not replace judgment.AI does not replace judgment.

它降低的是实验的成本,让想法更容易被测试、验证和改进。It lowers the cost of experimentation, making ideas easier to test, validate, and improve.It lowers the cost of experimentation, making ideas easier to test, validate, and improve.

为什么重要Why This MattersWhy This Matters

这次探索把同一套「搭建系统」的问题带进了新的语境。This exploration extends the same system-building questions into a new context.This exploration extends the same system-building questions into a new context.

媒介变了,但底层的关切依然熟悉:The medium has changed, but the underlying concerns remain familiar:The medium has changed, but the underlying concerns remain familiar:

  • 信息如何流动?How does information move?How does information move?
  • 决策如何变得更清晰?How do decisions become clearer?How do decisions become clearer?
  • 工具如何降低复杂度?How can tools reduce complexity?How can tools reduce complexity?
  • 系统如何保住人的意图?How can systems preserve human intent?How can systems preserve human intent?

已有原型Visible PrototypesVisible Prototypes