--- title: "Premortem / Murphyjitsu" description: "A planning practice that imagines failure in advance, finds the most plausible break point, and revises the plan before reality does it for you." aiSummary: "Premortem / Murphyjitsu is the future-facing revision practice inside Revising Beliefs Under Evidence. Gary Klein's premortem supplies the source-facing method: assume the plan has failed and ask why. CFAR's Murphyjitsu variant adds the inner simulator, a surprise test, and an iterative patch-and-repeat loop. The mechanism (planning fallacy; inner simulator; prospective hindsight; simulated failure as evidence-like pressure), the practice (make the plan concrete; run the failed-future movie; patch the likely failure; repeat), the lineage (Gary Klein's premortem, prospective hindsight, CFAR's Inner Simulator unit and Murphyjitsu technique, Murphy's Law, Kahneman and Tversky's simulation heuristic, planning fallacy research, David Allen's next-action discipline), and cross-references to Update Protocol, Chesterton's Fence, Noticing, Scout Mindset, Feedback Loops, and Entropy." discipline: foundation category: revising-beliefs-under-evidence disposition: living status: working-draft-v0-1 --- Normative *Foundation ยท Revising Beliefs Under Evidence*
01 // Mechanism
## Mechanism Plans fail in ways that were often visible before they failed. You knew you were not going to wake up at 5:30 after going to bed at 1:00. You knew the stakeholder review would take longer than one afternoon. You knew the migration depended on the one person who was about to go on holiday. The knowledge was not explicit enough to change the plan, but it was there. Premortem / Murphyjitsu is the practice of making that implicit prediction speak before reality does. The method is simple. Make a plan. Imagine that the deadline has passed and the plan failed. Ask whether you are surprised. If you are not surprised, ask what went wrong. Patch the plan against the most plausible failure mode. Then imagine the failed future again. Repeat until failure would actually surprise you, or until you admit the plan should not be trusted.
Premortem / Murphyjitsu loop: plan, imagine failure, find failure mode, patch, repeat Plan make concrete Failure would I be surprised? Failure mode what broke? Patch and repeat The plan improves by treating imagined failure as evidence before the deadline.
Inside Revising Beliefs Under Evidence, Premortem / Murphyjitsu is the future-facing update practice. The belief being revised is not a political opinion or a philosophical position. It is the implicit belief that the plan will work. Most people hold that belief with almost no evidence. The plan exists; the calendar has boxes; the steps sound plausible; the conclusion is that the future will comply. The practice puts pressure on that conclusion by asking your inner simulator to show the future in which it did not comply. This matters because explicit planning is often bad at the exact failures experience can already see. The explicit system writes the clean sequence: draft, review, approve, publish. The inner simulator knows the reviewer never answers on Fridays, the draft will uncover a dependency, the approval will trigger a stakeholder objection, and the publishing step has broken twice before. The simulator is not magic. It is your experience compressed into anticipation. Murphyjitsu gives it a specific job. The practice is different from worry. Worry loops on possible failure without improving the plan. Murphyjitsu asks for a concrete failed future, identifies the most plausible cause, and changes the plan. If no change follows, you were not practicing Murphyjitsu. You were rehearsing anxiety in a more intellectual costume, which is a familiar enough trick and not worth dignifying. It is also different from optimism correction by mood. The answer is not to become pessimistic. Pessimism can be as lazy as optimism: "Everything will go wrong" is no more useful than "Everything will be fine." The discipline is specific failure anticipation. Which part fails? Why? What would make that failure less likely? After the patch, would failure still be unsurprising?
02 // Practice
## Practice The diagnostic question is: **"If this plan failed, would I be surprised, and what went wrong?"** If the answer is "I would not be surprised," the plan is not ready. You have found evidence against your current belief that the plan will work.
**Make the plan concrete.** Murphyjitsu needs something the inner simulator can run. "Exercise more" is too vague. "Put running shoes by the bed, wake at 6:30, leave the house by 6:45, run the park loop, shower before the first meeting" is concrete enough. If you cannot visualize yourself executing the plan, you cannot simulate its failure. **Run the failed-future movie.** Imagine it is after the deadline and the plan did not happen. Do not analyze yet. Let the scene appear. What is the most obvious explanation? You forgot? You avoided the hard email? The dependency slipped? The meeting expanded? The social cost was higher than expected? The first answer is often the one the explicit plan politely omitted. **Patch the failure mode.** Add a defense against the most plausible break. Move the meeting. Send the email now. Ask the reviewer before the draft is finished. Put the task before the day's attention gets spent. Make the plan smaller. Get another person involved. The patch should change the expected future, not merely express renewed commitment. **Repeat until surprise.** Run the simulation again. If failure still feels unsurprising, identify the next failure mode and patch it. Stop when failure would actually surprise you, or when you discover that the plan cannot be made credible without changing the goal.
The hardest move is admitting that the plan cannot be patched at the current scale. Sometimes Murphyjitsu reveals that the goal is not wrong but the plan is fantasy. The deadline is impossible. The capacity is not there. The dependency cannot be controlled. The honest update is to reduce scope, renegotiate timing, or refuse the commitment. That can feel like failure before the work has begun. It is usually cheaper than discovering the same fact later with more people depending on you. There is one boundary to hold. Your inner simulator is trained on your experience, and experience is uneven. It is strong in domains where you have lived through enough similar situations. It is weak in domains where you lack examples or where the environment has changed. If you have never shipped a regulated medical device, your inner simulator may produce confidence or fear with equal ignorance. In those cases, Murphyjitsu should send you outward: ask someone who has seen this plan fail before.
03 // In the Wild
## In the Wild A person decided to start running every morning. The plan sounded clean: wake at 6:30, run, shower, work. Murphyjitsu asked him to imagine that two weeks had passed and he had run twice. He was not surprised. The failure was obvious: he would stay up late, wake tired, negotiate with himself, and lose. The patch was not a motivational speech. He moved the run to lunch, blocked the calendar, put shoes at the office, and asked a colleague to join on Mondays. Running at 6:30 was the identity-flattering plan. Running at lunch was the plan that survived contact with his actual life. A team planned to publish a major report on Friday. The plan listed final drafting on Monday, design on Tuesday, executive review on Wednesday, legal on Thursday, publication on Friday. The project lead ran Murphyjitsu. The failed-future movie was instant: legal would find a claim that needed substantiation, and the data owner would be unreachable. They moved legal review before design, pulled the contested claims into a separate appendix, and asked the data owner for sign-off that day. The report still published late, but by one day rather than two weeks. The original plan had treated review as a ceremony. Murphyjitsu treated it as a place where the plan could break. A founder planned to hire a head of operations after fundraising. Murphyjitsu asked: imagine six months from now, no hire, operations still fraying. What happened? The answer came quickly. The founder would keep postponing the role because every candidate looked expensive before the round closed, and after the round closed the company would be too busy to run a good search. The patch was to define the role, identify candidates, and run the first calls before fundraising finished. The plan changed from "hire later" to "make later hiring possible now." The failure had been hidden in the word later.
04 // Closing
Before trusting the next plan you make, visit the future where it failed. If the failure feels unsurprising, the future has already given you information. Use it while the plan can still change.
**Why this tool is in the Inheritance.** Premortem / Murphyjitsu is here because plans carry beliefs about the future, and those beliefs usually go unexamined until they fail. The Update Protocol responds when evidence arrives; this practice asks for a simulated report from the failed future while the plan can still change. Chesterton's Fence slows removal of an existing structure; Premortem / Murphyjitsu stress-tests a proposed sequence before execution. It belongs in Revising Beliefs Under Evidence because the plan's assumed success is a belief too, and it should update while revision is still cheap.
ROOTS
Lineage
## Lineage The primary method name is premortem. Gary Klein's premortem asks a team to assume that a project has failed and generate the reasons why before execution makes those reasons expensive. Prospective hindsight is the closest research base. Deborah Mitchell, Jay Russo, and Nancy Pennington's 1989 work showed that imagining an outcome as already having happened helps people generate more plausible explanations for it than merely considering it as possible. The Center for Applied Rationality later developed Murphyjitsu inside its Inner Simulator unit. The name joins Murphy's Law to an applied-rationality practice that uses anticipated failure as a training partner rather than as fatalism. CFAR's variant adds an experience-trained inner simulator, the question of whether failure would be surprising, and an iterative loop: find the likely break, patch it, and rerun the failed future. Duncan Sabien's 2022 LessWrong version of the CFAR handbook material is the strongest public writeup named by this profile. It distinguishes research support for the components from anecdotal support for the full package. The Inheritance carries both names because they identify different parts of the inheritance. Premortem names the established planning method and its source. Murphyjitsu names the iterative applied-rationality variant this profile teaches. The slash does not claim they are identical; it makes the relationship visible instead of publishing the derivative handle as though it were the source method. The adjacent academic lineage includes Kahneman and Tversky's work on the simulation heuristic, especially the idea that people judge likelihood partly by how easily they can mentally simulate a scenario. Planning fallacy research supplies the recurring problem: people underestimate duration and obstacles even when earlier experience should correct the forecast. David Allen's *Getting Things Done* contributes a practical constraint: a plan needs concrete next actions if the simulator is going to run it as a sequence rather than as an aspiration.
05 // Cross-references
## Cross-references **Within the category.** [The Update Protocol](/inheritance/revising-beliefs-under-evidence/the-update-protocol) records the evidence that would move a belief after reality reports back. Premortem / Murphyjitsu asks for a simulated report from the future and updates the plan before reality has to teach the lesson. [Chesterton's Fence](/inheritance/revising-beliefs-under-evidence/chestertons-fence) is the reform-side companion: before changing a structure, imagine the change failed and ask what hidden function you destroyed. [Double Crux](/inheritance/revising-beliefs-under-evidence/double-crux) can be used when two people disagree about whether a plan is credible: the crux is often the failure mode one person sees and the other has not yet simulated. **Within the Foundation.** [Scout Mindset](/inheritance/watching-your-own-reasoning/scout-mindset) is what makes the failed-future movie tolerable. Without it, Murphyjitsu becomes a threat to the identity of the plan-maker. [Noticing](/inheritance/watching-your-own-reasoning/noticing) catches the felt "of course that would fail" signal before the explicit mind explains it away. [Motivated Reasoning](/inheritance/watching-your-own-reasoning/motivated-reasoning) is the failure pattern Murphyjitsu often exposes: the plan was designed to protect the conclusion that the goal was achievable on the preferred terms. **Across to Reality.** [Entropy](/inheritance/reading-whats-operating/entropy) explains why plans degrade without maintenance. [Feedback Loops](/inheritance/reading-whats-operating/feedback-loops) carries the system-side version: plans should be revised as reality reports back, not defended because they were coherent at the start. The Murphyjitsu variant gives the Foundation-side practice before the feedback arrives. **Limitation.** Premortem / Murphyjitsu is only as good as the simulator it consults. In familiar domains, it can surface failure modes faster than explicit analysis. In unfamiliar domains, it may produce misplaced confidence, misplaced fear, or blankness. The practice should not replace expert review, data, or small tests. It should tell you where to look first.