/flux

Inversion Method Reference

Purpose: Solve forward problems by working backward from failure. Charlie Munger's inversion mental model — "invert, always invert" (borrowed from Carl Jacobi) — asks not how do we succeed but what would guarantee failure?, then systematically avoids those paths. Nassim Taleb's via negativa generalizes this: improvement by subtraction (remove what makes things fragile) rather than addition. Inversion fires best when forward analysis is saturated, when survivorship bias clouds the picture, or when goal-positive framing has masked obvious failure modes.

Scope Boundary

• flux inversion: invert the goal, enumerate failure-guarantees, derive avoid-list. Produces failure-mode list with avoidance actions, not full RPN scoring.
• flux reframe (default): full DEEP pipeline. Inversion is one tool in CHALLENGE; this subcommand isolates it.
• flux challenge: First Principles and Assumption Reversal. Reversal flips assumptions; inversion flips the goal.
• flux shift: perspective rotation across stakeholders/timeframes. Shift rotates viewpoint; inversion negates the objective.
• flux analogy: source-domain mapping. Analogy borrows shape; inversion subtracts.
• omen (elsewhere): pre-mortem with RPN/AP scoring of failure modes. Inversion generates failure scenarios; Omen quantifies and ranks them. Inversion is upstream — feed Omen.
• flux (Subtract mode, elsewhere): iterative subtraction dialogue. Inversion is a single-pass goal-flip; Flux Subtract is multi-turn pruning.
• void (elsewhere): YAGNI scope cutting. Void removes work-not-yet-done; inversion surfaces failure-paths to avoid.

Workflow

ENTER    →  state the goal in positive form
         →  capture context, constraints, desired outcome

INVERT   →  rewrite the goal as its negation
         →  "succeed at X" → "guarantee failure at X"

ENUMERATE →  list ≥10 ways to guarantee that failure
          →  use prompt bank: behaviors, structures, decisions, omissions

CLUSTER  →  group failure-paths by mechanism (technical, social,
            economic, cognitive, temporal)

AVOID    →  derive avoid-list — concrete actions / non-actions
         →  each failure-path → at least one avoidance commitment

VIA NEG  →  apply Taleb subtraction — what to *stop doing* often
            outranks what to start doing

DELIVER  →  hand to Omen (RPN/AP scoring), Magi (decision), or
            user (avoidance commitments)

Munger's Inversion Prompts

• If we wanted to guarantee this fails, what would we do?
• What does the failure case look like in detail? Walk it forward.
• Who has tried this and failed? What did they do? (Survivorship-bias antidote.)
• What would a saboteur recommend, knowing the team's blind spots?
• What is the team currently doing that resembles items on the failure list?
• If success requires N things to go right, what is each thing's failure mode?

Via Negativa (Taleb) Prompts

• What can we stop doing that would make this more robust?
• What dependency, if removed, reduces fragility?
• What feature/process/policy adds risk without proportional value?
• What "best practice" are we doing because it is conventional, not because it works?
• Which addition has, historically, made similar systems worse?

Failure-Mode Categories (enumeration scaffold)

Aim for coverage across categories — single-category failure lists betray a blind spot.

What-Not-To-Do Anti-Bias

Forward planning is positive-action biased: teams list what to do, not what to stop. Inversion's main value is extracting the negative commitments — items the team will not do, paths it will not take, additions it will refuse. These are typically underweighted in planning artifacts and overweighted in postmortems.

Anti-Patterns

• Inversion-as-rhetoric — saying "let's invert" then continuing forward analysis. Inversion requires actually rewriting the goal as its negation.
• Stopping at 3 failure-paths — the obvious ones surface first. Push to ≥10 to reach non-obvious paths.
• Single-category enumeration — listing only technical failures (or only social ones). Force coverage across all 6 categories.
• Conflating with pre-mortem — inversion generates failure scenarios; pre-mortem (Omen) quantifies them. Don't skip the handoff.
• Avoidance-list without commitment — listing "don't do X" without an owner or trigger creates theater, not behavior change.
• Survivorship-blind inversion — citing only the failure paths the team already discusses. Seek failure paths from others' postmortems.
• Inverting the wrong granularity — inverting "succeed at the project" is too coarse. Invert specific decisions, milestones, or assumptions.
• Treating via negativa as nihilism — the goal is not to remove everything, but to identify the fragile-additions whose removal increases robustness.
• Skipping the avoid-list step — failure-paths without avoidance commitments are entertainment.

Handoff

• To Omen: failure-path list as candidates for RPN/AP scoring and pre-mortem ranking. Inversion is the generation engine; Omen is the prioritization engine.
• To Magi: when avoidance commitments conflict with proposed actions — decision needed.
• To Flux (Subtract mode): via negativa candidates — iterate on what to stop doing.
• To Spark: failure-paths sometimes invert into feature ideas (the avoidance becomes the differentiator).
• To Void: avoidance commitments that imply scope cuts.
• To analogy: failure-paths that resemble known patterns from other domains — borrow their countermeasures.
• To User: final avoidance commitments with owners and triggers, alongside the original forward plan.