On Making Decisions

This is a living document. I update it when I change my mind.

Before you decide

• Classify the decision before optimizing it.
  • Is it reversible or irreversible? Recoverable or catastrophic?
  • Reversible + low downside + fast feedback → decide quickly and learn through action
  • Irreversible + high stakes + slow feedback → slow down and demand better evidence
  • Applying irreversible-decision rigor to reversible decisions causes paralysis
• Start with a model, then gather relevant information.
  • Define what variables matter, what outcomes are possible, what evidence would change the conclusion
  • Information gathering should improve the decision, not just increase knowledge
  • Stop when it no longer changes the decision, or when the search becomes justification
• Actively search for disconfirming evidence.
  • What would prove this wrong? What would a credible skeptic say? What assumptions might fail?
  • Intelligent people are better at defending beliefs than revising them. Seek contradiction deliberately

Managing risk

• Avoid ruin before chasing upside.
  • Avoid situations where a single mistake destroys optionality or losses compound faster than learning
  • Prefer decisions where failure is survivable, future attempts remain possible, and mistakes teach something
  • Long-term success comes from staying in the game long enough for good bets to compound
• Prefer asymmetric opportunities.
  • Limited downside, large or nonlinear upside, multiple attempts allowed
  • Even mediocre probability becomes attractive when the payoff structure is asymmetric
  • High probability + capped upside + meaningful downside often becomes a trap
• Think in probabilities rather than certainties.
  • "This will succeed" → "This likely has a 60-70% chance"
  • Consider ranges instead of single forecasts. Examine expected outcomes and tail risks
  • Good decision-makers remain comfortable under uncertainty rather than false certainty

Mind and emotion

• Use intuition selectively.
  • Intuition is compressed experience. It works when you've seen many examples and feedback is reliable
  • It fails when the domain is new, feedback is slow, or strong emotions are present
  • Combine structured reasoning with earned intuition
• Treat emotional conviction as information, not proof.
  • Strong feelings can represent genuine pattern recognition or temporary states like fear, fatigue, or excitement
  • When conviction feels overwhelming, pause and ask whether it arises from experience or current state
• Evaluate decisions across different states of mind.
  • Revisit major decisions after sleep, after exercise, during calm reflection, during normal stress
  • If a decision only feels correct in a particular emotional state, it may not be robust
• Let major decisions mature over time.
  • Does the reasoning become clearer over days? Does it remain stable across contexts?
  • Good decisions become simpler with time. Weak ones rely on urgency or pressure
• Match speed to reversibility.
  • Speed when decisions reverse easily, learning comes through experimentation, opportunities move fast
  • Deliberation when consequences are long-lasting, errors are costly, outcomes unfold slowly

Evaluating outcomes

• Separate decision quality from outcome quality.
  • Outcomes contain randomness. A sound process can produce poor results, and poor reasoning can occasionally succeed
  • Evaluate the reasoning, the evidence available, the quality of assumptions
  • Think in long sequences, not isolated events
• Let behavior outrank narratives.
  • Consistent follow-through, repeated sacrifices, behavior under pressure
  • Statements of intent are weaker evidence than patterns of behavior
• Principles should constrain behavior, not decorate it.
  • A principle is real only when it applies even when inconvenient and imposes short-term costs
  • If a rule appears only when it benefits you, it is rationalization, not principle

What to optimize for

• Pursue ambitious problems carefully.
  • Large problems attract stronger collaborators, create more meaningful impact, offer greater asymmetry of reward
  • Pair ambition with realistic milestones, learning loops, and incremental progress
  • Vision without feedback becomes fantasy
• Be explicit about what you are optimizing for.
  • Meaning, mastery, financial security, freedom, credibility, relationships
  • Misalignment occurs when people pursue one goal while believing they are pursuing another
• Protect early intuition but test it through reality.
  • Preserve the intuition, test it through small experiments
  • Avoid turning untested intuition into large commitment. Curiosity + disciplined testing
• Value experiences that remain meaningful without recognition.
  • Would this still feel worthwhile if nobody knew about it?
  • Internal satisfaction sustains motivation far longer than recognition alone