AgentSkillsCN

scale-game

在极端尺度下(放大或缩小 1000 倍)进行测试,以揭示那些在常规尺度下被掩盖的根本真理。

SKILL.md
--- frontmatter
name: scale-game
description: "Test at extremes (1000x bigger/smaller) to expose fundamental truths hidden at normal scales."
last_updated: 2026-01-10
tools_required: []
agent_type: main_agent

Scale Game

Test approaches at extreme scales to find what breaks and what survives.

[GOAL]

Validate an approach by testing it at extreme scales to expose hidden assumptions and limits.

[CONTEXT]

Core principle: Extremes expose fundamental truths hidden at normal scales.

Use when:

  • Uncertain about scalability
  • Edge cases unclear
  • Validating architecture for production volumes
  • "It works in dev" needs verification

[PROCESS]

  • Pick dimension - What could vary extremely?
  • Test minimum - What if 1000x smaller/faster/fewer?
  • Test maximum - What if 1000x bigger/slower/more?
  • Note what breaks - Where do limits appear?
  • Note what survives - What's fundamentally sound?

[SCALE DIMENSIONS]

DimensionTest At ExtremesReveals
Volume1 item vs 1B itemsAlgorithmic complexity limits
SpeedInstant vs 1 yearAsync requirements, caching needs
Users1 user vs 1B usersConcurrency issues, resource limits
DurationMilliseconds vs yearsMemory leaks, state growth
Failure rateNever fails vs always failsError handling adequacy

[EXAMPLES]

Error Handling

  • Normal: "Handle errors when they occur" works fine
  • At 1B scale: Error volume overwhelms logging, crashes system
  • Reveals: Need to make errors impossible or expect them

Synchronous APIs

  • Normal: Direct function calls work
  • At global scale: Network latency makes sync calls unusable
  • Reveals: Async/messaging becomes survival requirement

In-Memory State

  • Normal duration: Works for hours/days
  • At years: Memory grows unbounded, eventual crash
  • Reveals: Need persistence or periodic cleanup

[IMPORTANT]

  • Test both directions (bigger AND smaller)
  • What works at one scale fails at another
  • Use insights to validate architecture early
  • Don't assume "should scale fine" without testing

See Also

  • [[when-stuck]] - Dispatch to right technique
  • [[simplification-cascades]] - Find unifying abstractions