Catastrophizing: One Bug to “The Company Is Over” in Four Seconds
What it actually looks like
The email says one customer is cancelling. By the time you reach the second paragraph you have already been through the sequence: if they left, others will leave, if others leave the revenue does not hold, if the revenue does not hold you cannot pay the two people who joined on the strength of your promise. The email is nine lines long. You have run a full company failure inside it.
A bug report comes in. Not a serious one. Your first physical response — before you have read what it says — is the drop in the stomach, and it is the same drop whether the report turns out to be a typo in a tooltip or an actual data problem. The alarm does not grade its inputs.
Your language goes absolute. Everything is broken. Nobody is converting. This always happens. You would notice these words instantly in someone else's sentence.
You open the analytics at 11pm to check whether anything else has gone wrong. Nothing about that check can help. You do it anyway, and then again twenty minutes later.
And the tell that separates this from ordinary worry: two days on, when the customer turns out to have been a bad fit and the bug turns out to be trivial, you do not feel relief in proportion to the panic. You feel nothing much. The alarm never gets corrected — it just moves to the next thing.
Who this happens to
This is common among first-time founders, and it is not a defect of temperament so much as a shortage of data.
Judging how bad something is requires having seen a comparable thing before and watched how it ended. Someone on their third company has a shelf of resolved crises to draw on: they have seen a churned customer that meant nothing and one that meant something, and they can tell them apart within a minute. On your first, every incident is unprecedented, so every incident gets sized against imagination rather than memory.
It is common in founders under real financial pressure. This matters, because it is what makes the response reasonable rather than irrational. When the runway is short, the leap from a small loss to a serious one is not a distortion — it is arithmetic. The trouble is that the same machinery keeps running afterwards, at the same volume, once the pressure has eased.
And it is common in people who are conscientious to a fault. If you are the person who checks the deploy twice, you have a system tuned to catch problems early. That system is genuinely why some things have not gone wrong. It is the same system that cannot leave the room when there is nothing to catch.
What sets it off
Bad news arriving with no context attached. A one-line cancellation with no reason given, a two-star review, a message from an investor that just says "can we talk". Ambiguity is the accelerant — an unexplained event has to be filled in, and the system that fills it in has a strong house style.
Anything at night. The same customer email at 9am produces a task; at 11pm it produces a spiral. Tiredness removes the part of you that argues back.
Money events. An invoice going unpaid, a payment failing, a number lower than the month before. These attach directly to survival, which is where the alarm is loudest.
Silence. No signups today, no reply to the proposal, nobody in the support inbox. Nothing has happened, which means nothing contradicts the story, which means the story runs unopposed.
And a specific founder trigger: shipping something. In the hours after a release, the system that has been holding still finally has an opening, and every unfamiliar number in the dashboard looks like the beginning of a consequence.
Why it keeps happening
Because the system generating these predictions is doing its job, and its job was never accuracy.
It was built to make false alarms cheap and missed threats fatal. Assume the rustle in the grass is a predator a thousand times and you lose a thousand seconds; be wrong once in the other direction and there is no thousand-and-first time. That trade is correct for a physical threat, and the machinery has not been told that its inputs are now emails. Related to this is a well-supported finding: negative information is weighted more heavily than positive information of the same size. One cancellation lands harder than one signup, and both are one customer.
Then the loop that keeps it in place, which is where most of the damage lives. The prediction arrives with the full physical signature — heart rate, tight chest, urgency. That signature is evidence. It feels like the body confirming the forecast, when it is only the body responding to the forecast. Feeling something strongly is not information about how likely it is.
And the reason it never learns: the catastrophe does not arrive, and no correction gets filed. Nobody sits down at the end of the month and notes that eleven predicted disasters produced zero disasters. The alarm has no incentive to update, because it was never scored. It goes on being loyal, fast, and almost always wrong.
Meanwhile, the arithmetic runs in the background. The panic is not free: it costs the afternoon, degrades the decision you make next, and — this is the part worth being honest about — makes the wrong call more likely, because the thing you reach for during a spiral is whatever ends the spiral fastest. Founders discount their product, fire a good person, or abandon a working strategy at exactly this temperature. The bug was small. The response to the bug is where the actual damage is done.
What actually helps
Separate the event from the story, on paper. Two columns. Left: what actually happened, in the flattest possible language — one customer of forty cancelled, no reason given. Right: what your mind added — churn is starting, we are not going to make it. The gap between the columns is the whole problem, and it is not visible until it is written. Ten seconds of writing, and it works when arguing with yourself does not.
Ask the two decatastrophizing questions. What is the realistic worst case here — not the worst imaginable, the realistic one? And what would I actually do if it happened? The second question is the one that does the work. Almost every founder catastrophe has a survivable answer, and naming the answer hands you back the thing panic takes: the sense that there is a move.
Get out of the imagined future through your senses. When the spiral has started, arguing has already stopped working. Name five things you can see and four you can hear, slowly. It is not calming for its own sake — it moves attention from a forecast to a room, and the forecast cannot run without your attention.
Refuse to decide anything at that temperature. Make it a standing rule: no pricing change, no personnel decision, no strategy reversal within two hours of a spike. Write the decision down and revisit it in the morning. Most of what you would have sent, you will not send.
Keep a short record of predictions that did not happen. One line each: what you were sure would happen, what happened. Ten entries in, you will have something the alarm has never had — a scoreboard. This is the only thing that eventually changes its calibration, and it takes weeks, not minutes.