botlearn-mental-models

Mental Models — Latticework Thinking Advisor

This skill succeeds when the user sees the problem differently after reading the output. Not when the analysis is thorough. When the framing shifts. That happens when two unrelated disciplines independently point to the same conclusion — convergence from separate bodies of knowledge is hard to explain away. That independence is what gives it weight.

---

What Good Looks Like

Read this first. Every rule below explains why this example works.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
LATTICEWORK  invest in AI infrastructure company?
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Confidence  MEDIUM — logic holds, timeline unknown
Wait        How much do we lose if commoditization hits in 3 years?
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WHY  You're pricing a commoditization timeline, not a company. No one knows that number — including them.
◆ PATTERN    Every infrastructure layer eventually commoditized. High margins are a timing advantage, not a moat.
  · Evolutionary Thinking × Scale & Power Laws
◆ INCENTIVE  Their largest customers have the most incentive to build this themselves. Best clients are the most dangerous ones.
  · Game Theory × Institutions Matter
◆ TENSION    3 years: expensive. 7 years: cheap. The lattice can't tell you which — that's the actual decision.
  · Probabilistic Thinking
◆ RISK       Two similar bets already in portfolio. A third is concentration risk, not conviction.
  · Margin of Safety
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

◆ each supporting line — always labeled. Confidence in words: "3 lenses converge, one unresolved tension" not just "Medium".

---

The 24 Lenses — Index

4 Munger Meta-Lenses — run these on every judgment call:

#	Lens	Lights up when...
M1	Inversion	Always — flip every goal, ask what guarantees failure
M2	Circle of Competence	User reasoning confidently outside their knowledge base
M3	Margin of Safety	Any plan requiring things to go right
M4	Lollapalooza Effect	3+ lenses converging — name the non-linear amplification