Files

T

Antigravity Agent c24165b8bc refactor(topics): 멀티 에이전트용 지식 재편 — _Common(공통 기본기) + Domain_* 구조

에이전트 8종(대화형/프로그래머 C·S/디자이너/설계자/기획자/QA/PD/PM)에게
[공통 기본 능력 + 롤별 Specialty] 2층으로 지식을 주입하기 위한 재분류.
문서 내용·포맷은 무수정, 폴더 이동만 (6,372개 문서 수 보존 확인).

- Topic_Programming → Domain_Programming (내부 구조 보존)
- Topic_Graphic → Domain_Design
- Topic_Business → Domain_Product
- Topic_General → Domain_General
- _Common 신설: Math(구 Topic_Math_Specialty), Reasoning(구 General/From_Thinking & Reasoning),
  Reasoning_Creativity(구 General/From_창의성), Communication(Poetic_Blog_Writing + From_writing)
- 타 도메인의 From_* 폴더는 유지 (출처 표기일 뿐, 이미 도메인에 맞게 분류된 문서)
- 빈 폴더 정리 (memory/procedures)
- 에이전트→폴더 매핑은 workspace의 .astra/agent-knowledge-map.json (9개 에이전트)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-11 11:05:56 +09:00

3.8 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Mental Operations Synthesized

매 한 줄

"매 think-about-thinking 의 toolkit". 매 Polya (1945 How to Solve It), Bloom's Taxonomy, Dual Process Theory (Kahneman), 매 lineage 의 synthesis — 매 generic mental moves 매 cross-domain transferable.

매 핵심

매 the operations

Decompose: split problem into sub-problems.
Abstract: drop irrelevant details, keep essence.
Generalize: extend known to wider class.
Specialize: take general → solve concrete instance.
Analogize: structurally-similar known problem.
Invert: solve the negation, the dual, the inverse.
Verify: test boundary cases, sanity checks.
Iterate: refine via cycles.

매 modern frame

System 1 (Kahneman): pattern-match, intuitive.
System 2: deliberate, sequential, costly.
매 expert 의 chunking 의 System 1 expansion.

매 응용

Algorithm design (CLRS-style).
Debugging (binary search of state space).
Strategy / business problems (MECE, issue trees).
LLM prompting (Chain-of-Thought = explicit System 2).

💻 패턴

Polya's 4-step (problem solving)

1. Understand: restate problem in own words.
2. Plan:       map to known technique (analogize).
3. Execute:    carry out plan, check each step.
4. Look back:  verify, generalize, simplify.

Decompose: tree

Problem: Build chat app
├── auth (sub: oauth flow, session, refresh)
├── transport (sub: WS, fallback to SSE)
├── persistence (sub: schema, migrations)
└── UI (sub: list, composer, attachments)

Inversion (Charlie Munger / Carl Jacobi)

Original: "How do I make this app fast?"
Inverted: "How would I make it slow?"
  → list 50 ways → avoid each

Analogy mapping

Known: Dijkstra (shortest path on graph)
New:   "find cheapest API call sequence"
Map:   nodes=states, edges=API calls, weight=cost
→ apply Dijkstra

Specialize → Generalize bootstrap

1. Solve N=1 case by hand.
2. Solve N=2.
3. Spot pattern.
4. Conjecture for N.
5. Prove by induction.

Verify: boundary tests (CS / engineering)

def median(xs): ...
assert median([5]) == 5                       # single
assert median([1, 2]) == 1.5                  # even
assert median([]) is None                     # empty
assert median([1, 1, 1]) == 1                 # duplicates

매 결정 기준

Stuck-state	Operation
Too big to grasp	Decompose
Too messy / many details	Abstract
No idea where to start	Analogize
Direct attack failing	Invert
Don't trust the answer	Verify

기본값: Polya 4-step + decompose-first. 매 invert when stuck.

🔗 Graph

부모: Problem_Solving
변형: First-Principles · MECE
Adjacent: Chain-of-Thought

🤖 LLM 활용

언제: prompt-design (CoT = these operations made explicit), self-review of complex tasks. 언제 X: routine pattern-match tasks (System 1 sufficient).

❌ 안티패턴

Always System 2: 매 exhausting + slow.
Always System 1 on hard problems: 매 systematic errors.
No verification step: 매 plausible-but-wrong.

🧪 검증 / 중복

Verified (Polya — How to Solve It; Kahneman — Thinking Fast and Slow; Bloom's Taxonomy).
신뢰도 B+.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — Mental Operations FULL synthesis

3.8 KiB Raw Blame History