After a company closes, there is an accounting that runs without being asked. It adds up the losses: the valuation, the product, the team, the time. The list is real and it is worth facing. What it almost never includes is the column on the other side, because the cognitive load of the loss makes the gains invisible. This is not a motivational observation. It is a measurement problem with a specific mechanism, and it produces a systematically wrong input for the most important decisions of the next six months.
The honest version of the audit does not skip the loss column. It completes the balance sheet.
What the human capital research actually predicts
Labour economics distinguishes between skills that are tightly context-specific, useful only inside a particular firm or role, and skills that are general and durable, transferring across contexts as human capital. The research by Heckman and colleagues on what predicts long-run outcomes points consistently at the second category: not domain-specific technical knowledge, but what they call non-cognitive skills. Persistence under ambiguous conditions. The capacity to regulate behaviour and emotion under pressure. The ability to read a situation with incomplete information and make a decision anyway. Comfort with asymmetric risk. These compound with practice rather than depreciate with a company's result.
Running a startup puts an unusual volume of pressure on exactly these muscles, more pressure per unit of time than most other work environments produce. The founder who operated a company for three years, regardless of how it ended, did a large volume of reps on the skills that the research identifies as durable. The outcome does not undo the reps.
Why the accounting runs wrong
There are two reasons the audit is systematically incomplete after a shutdown, and neither is about weakness of character.
The first is what Seligman's research on explanatory style calls the global attribution pattern: the move from "this venture failed" to "I failed, and the failure is a general verdict on my competence." When you are inside that pattern, the inventory of what worked collapses. The frame that should hold it has been replaced by a frame that finds nothing to hold. The global attribution is not only a mood problem; it is an epistemically wrong model that makes the skill record invisible.
The second layer comes from Kahneman's work on availability: the mind samples its own evidence by how vivid and recent it is, not by how representative it is. The most recent months of a failed company are loaded at high resolution: a fundraise that did not close, a churn event, a last conversation with a team member. The two years of decisions and skill-building that preceded them are not. The audit is not just emotionally biased; it is sampling from the wrong part of the timeline.
Three evidence types the audit should include
Running the honest version means separating what you built from the fact of the outcome. The outcome is the company, which closed. The skills built during the company's life are a different category, and they can be read from a different part of the record.
First: decisions that were right. A company that ends in failure still contains a large number of correct calls: early hires that worked, product bets that held, partnerships that did not blow up, hard conversations that resolved cleanly. Those decisions reflect a competence that was real when they were made. They belong on the balance sheet. The availability bias does not let them surface without effort; you have to go looking for them deliberately.
Second: the general skills the role built. Managing a team, even a small one, builds management competence that transfers. Running a fundraise, even an unsuccessful one, builds the capacity to compress a complex story, read a room under pressure, and hold a relationship through an adversarial dynamic. Diagnosing a product problem in real time, with real money at stake, builds pattern recognition that classroom study does not produce. These are not consolation prizes. They are the durable assets the research identifies as predictive.
Third: domain knowledge that is now yours. Whatever problem the company was attacking, you accumulated a level of insight into it that most people do not carry. Whether that knowledge is relevant to the next move is a separate question. The point is that it exists, it is specific, and it belongs in the inventory.
What the audit does not let you off the hook for
The competence audit is not an alibi, and running it honestly means including what was not built as well as what was. The skills the company needed that you did not have, the judgment calls that were wrong with the information available at the time, the pattern of decisions that in retrospect had a clear shape: those belong in the same inventory. The point of running it is not to arrive at a conclusion that feels good. It is to arrive at an accurate input.
The decision about what to do next, whether that is founding again, joining a team as an operator, or taking a deliberate pause before deciding, is a decision made with an asset base. The decision made with an accurate read of that base is a different decision from one made by someone who believes they are starting from nothing. The research does not say the assets guarantee any particular outcome. It says that ignoring them produces a miscalibrated starting point, and the decisions downstream of a miscalibrated starting point tend to compound the error.
You are not starting from nothing. That is not a motivational claim. It is what the balance sheet says when the accounting is complete.