01
Recognise six recurring illusions
High profit is not high capital efficiency; consolidated growth is not competitive growth; earn-outs do not automatically protect value; book assets are not realisable value; accounting compliance does not prove investment quality; short-term outcomes do not prove decision quality.
These are retrieval entrances, not aphorisms. Each requires facts, counterexamples, time and outcomes.
02
Study failure for breakpoints and success for systems
Failure amplifies breakpoints in strategy, valuation, diligence, structure, integration and execution, but rarely has one cause. Separate known risk, evidence gaps, weighting errors, execution variance and external change.
Success usually combines multiple conditions. Identify controlled risks, merely adequate conditions, genuine advantages and outcomes attributable to timing or luck.
03
Preserve the decision scene
Retain objective, knowns and unknowns, alternatives, rationale, rejected options, transaction design, key changes and outcomes. Hindsight must not rewrite the information set.
Human pages organise conflict, choice and outcome; machine indexes organise stable IDs, dates, evidence, relations and boundaries.
04
Publication requires evidence and de-identification
Public-company facts may be named with announcement and outcome dates; historical publications retain publication dates; participated projects require confidentiality and role review; current work publishes only abstract problems and general methods.
One case can support several judgments, and a judgment must face several cases and counterexamples. Retrieval follows relationships, not serial numbering.