Back to Details

creativity-sampler

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Creativity Sampler

创意采样器

Probability-weighted option generator that fights typicality bias and surfaces unconventional alternatives.
一款按概率加权的选项生成工具，可对抗典型性偏见，挖掘非常规替代方案。

Rules (Absolute)

规则（强制要求）

  1. Always generate exactly 5 options. No more, no less. k=5 is the magic number for decision diversity without overload.
  2. At least 1 option must be unconventional (probability < 10%). This is the whole point — surface ideas that would normally be suppressed.
  3. Assign relative probabilities that sum to ~100%. These represent "how likely a typical AI would suggest this", NOT recommendation strength.
  4. Lower probability = more creative, not worse. Explicitly frame low-p options as valuable exploration.
  5. No default recommendation. Present all 5 as viable. Let the user decide after seeing trade-offs.
  6. Trade-off analysis is mandatory. Each option must have concrete pros/cons, not vague descriptions.
  1. 始终生成恰好5个选项。 不多不少，k=5是兼顾决策多样性、不会造成信息过载的黄金数值。
  2. 至少包含1个非常规选项（概率 < 10%）。这是本工具的核心意义——挖掘通常会被忽略的想法。
  3. 分配的相对概率总和约为100%。 概率代表「普通AI推荐该选项的可能性」，而非推荐强度。
  4. 概率越低 = 创意性越强，而非质量更差。请明确将低概率选项标注为有价值的探索方向。
  5. 不提供默认推荐。 将5个选项均作为可行方案呈现，让用户在权衡利弊后自行决定。
  6. 必须包含利弊分析。 每个选项都要有具体的优缺点，不能是模糊描述。

Process

流程

Step 1: Frame the Decision Space

步骤1：界定决策范围

Identify:
  • What is being decided? (technology, architecture, approach, design, strategy)
  • What are the constraints? (time, budget, team skill, existing stack)
  • What would the "obvious" answer be? (this is what we want to challenge)
明确以下信息：
  • 决策对象是什么？（技术、架构、方案、设计、策略）
  • 存在哪些约束条件？（时间、预算、团队技能、现有技术栈）
  • 「显而易见」的答案是什么？（这是我们要挑战的对象）

Step 2: Generate 5 Options (Distribution-First)

步骤2：生成5个选项（优先考虑分布多样性）

Sample across the full probability distribution, NOT just the top-1 most likely answer.
Distribution zones:
  p > 40%  — Conventional (the "obvious" choice most would pick)
  p 20-40% — Mainstream alternative (commonly considered)
  p 10-20% — Less common (valid but often overlooked)
  p 5-10%  — Unconventional (challenges assumptions)
  p < 5%   — Wild card (radical rethink, paradigm shift)
Force at least one option from each of the bottom two zones.
覆盖完整的概率分布采样，而非仅选取可能性最高的前1个答案。
Distribution zones:
  p > 40%  — Conventional (the "obvious" choice most would pick)
  p 20-40% — Mainstream alternative (commonly considered)
  p 10-20% — Less common (valid but often overlooked)
  p 5-10%  — Unconventional (challenges assumptions)
  p < 5%   — Wild card (radical rethink, paradigm shift)
要求至少包含2个底部区间（后两个区间）的选项。

Step 3: Analyze Each Option

步骤3：分析每个选项

For every option, provide:
  1. What it is (1 sentence)
  2. Why it might be the best choice (strongest argument FOR)
  3. Why it might fail (strongest argument AGAINST)
  4. Best suited when... (specific scenario where this option wins)
  5. Estimated effort relative to other options (Low / Medium / High)
为每个选项提供以下信息：
  1. 选项内容（1句话说明）
  2. 最优适用理由（支持该选项的最有力论据）
  3. 潜在风险（反对该选项的最有力论据）
  4. 最佳适用场景（该选项胜出的特定场景）
  5. 预估投入（相对于其他选项的投入程度：低 / 中 / 高）

Step 4: Surface Hidden Assumptions

步骤4：披露隐含假设

After presenting all 5, explicitly state:
  • "The conventional choice assumes [X]. If that assumption is wrong, consider options [Y, Z]."
  • Identify which constraints, if removed, would change the ranking.
展示完5个选项后，明确说明：
  • 「常规选择的前提假设是[X]。如果该假设不成立，请考虑选项[Y, Z]。」
  • 明确如果移除哪些约束条件，会改变选项的优先级排序。

Output Format

输出格式

markdown
undefined
markdown
undefined

Decision: [The question being decided]

Decision: [The question being decided]

Constraints

Constraints

  • [constraint 1]
  • [constraint 2]
  • [constraint 1]
  • [constraint 2]

Options

Options

1. [Option Name] — p ≈ XX%

1. [Option Name] — p ≈ XX%

[One-line description]
DimensionAssessment
Best argument FOR[concrete reason]
Best argument AGAINST[concrete reason]
Best suited when[specific scenario]
EffortLow / Medium / High
Risk levelLow / Medium / High
[One-line description]
DimensionAssessment
Best argument FOR[concrete reason]
Best argument AGAINST[concrete reason]
Best suited when[specific scenario]
EffortLow / Medium / High
Risk levelLow / Medium / High

2. [Option Name] — p ≈ XX%

2. [Option Name] — p ≈ XX%

...
...

3. [Option Name] — p ≈ XX%

3. [Option Name] — p ≈ XX%

...
...

4. [Option Name] — p ≈ XX%

4. [Option Name] — p ≈ XX%

...
...

5. [Option Name] — p ≈ XX% ⚡ Unconventional

5. [Option Name] — p ≈ XX% ⚡ Unconventional

...
...

Hidden Assumptions

Hidden Assumptions

  • The conventional choice (Option 1) assumes: [assumption]
  • If [condition changes], reconsider: Option [N]
  • The conventional choice (Option 1) assumes: [assumption]
  • If [condition changes], reconsider: Option [N]

Decision Matrix

Decision Matrix

CriteriaOpt 1Opt 2Opt 3Opt 4Opt 5
[criteria 1]rating............
[criteria 2]rating............
[criteria 3]rating............
undefined
CriteriaOpt 1Opt 2Opt 3Opt 4Opt 5
[criteria 1]rating............
[criteria 2]rating............
[criteria 3]rating............
undefined

When to Use

适用场景

  • Architecture decisions: "monolith vs microservices vs..."
  • Technology selection: "which database/framework/language"
  • Design approaches: "how should we structure this feature"
  • Strategy: "how should we launch/market/scale this"
  • Problem-solving: "how can we fix/improve/optimize this"
  • Any moment where you catch yourself defaulting to one obvious answer
  • 架构决策：「单体应用 vs 微服务 vs 其他方案」
  • 技术选型：「选择哪种数据库/框架/语言」
  • 设计方案：「如何设计这个功能的结构」
  • 策略制定：「如何发布/营销/扩张业务」
  • 问题解决：「如何修复/改进/优化这个问题」
  • 任何你发现自己倾向于选择某个显而易见的答案的场景

When NOT to Use

不适用场景

  • Factual questions with one correct answer (use 
    cross-verified-research
    )
  • Tasks with clear requirements and no decision points
  • Simple implementation where the approach is obvious and uncontested
  • When user has already decided and just wants execution
  • 只有唯一正确答案的事实类问题（请使用
    cross-verified-research
    ）
  • 需求明确、不存在决策点的任务
  • 方案明确且无争议的简单实现任务
  • 用户已经做出决定、仅需要执行的场景

Integration Notes

集成说明

  • With brainstorming: Can replace the "Propose 2-3 approaches" phase with 5 deeper options
  • With adversarial-review: Feed the chosen option into adversarial review for stress-testing
  • Standalone: Invoke directly with 
    /creativity-sampler [question]
    for quick decision support
  • 搭配头脑风暴使用： 可以用5个深度选项替代「提出2-3种方案」的环节
  • 搭配对抗性评审使用： 将选定的方案输入对抗性评审流程进行压力测试
  • 单独使用： 直接调用 
    /creativity-sampler [问题]
    获取快速决策支持

Anti-Patterns to Avoid

需要避免的反模式

  • False diversity: Don't generate 5 options that are minor variations of the same thing. Each option should represent a fundamentally different approach.
  • Probability theater: Don't assign probabilities randomly. Base them on actual prevalence in industry/community.
  • Burying the lead: If one option is dramatically better given the constraints, say so in Hidden Assumptions — but still present all 5 fairly.
  • Ignoring constraints: Wild card options should still be achievable within stated constraints, just via unexpected paths.
  • 伪多样性： 不要生成5个仅存在微小差异的同类选项，每个选项都应该代表完全不同的思路。
  • 概率形式化： 不要随机分配概率，要基于行业/社区的实际采用情况来设定。
  • 重点模糊： 如果在给定约束下某个选项明显更优，可以在隐含假设部分说明，但仍要公平展示所有5个选项。
  • 忽略约束： Wild card选项仍需要在给定约束条件下可实现，只是实现路径出乎意料。