The Senior Seminar in Abject Failure
A commencement address for the people about to inherit the institutions that just broke
The chairs are already out. Somewhere this month, a field house is filling with families holding programs they’ll fold into fans, since the ventilation in a building designed for basketball was never meant for two thousand people in gowns in June. A name gets read. A young woman in a mortarboard that doesn’t quite sit right crosses a stage, takes a hand, moves the tassel left to right, and the entire ceremony exists to certify one sentence: she learned what we taught her, and we are vouching for it.
I want to talk to that young woman, and to the thousands like her this season, about the one seminar that wasn’t on her transcript.
I coach boys’ gymnastics, so I think about certification differently than most people in that field house do. The sport I teach, six events of it, floor, pommel horse, vault, parallel bars, still rings, and high bar, is under the chalk and the calluses and the medals a multi-year course in falling on purpose. Before a kid does anything you’d pay to watch on the high bar, he spends a year learning what his body should do at the top of a release when his hands miss the bar, which they will, repeatedly, in front of an audience. He learns the same thing flying off the end of the pommel horse and coming down short off the rings. There’s no medal for that part. Nobody’s grandfather drives four hours to watch a twelve-year-old boy finally stop reaching for the floor with a stiff arm and instead tuck and roll through a missed dismount the way I’ve told him six hundred times. But that skill, the ugly one, the one with no ceremony attached, is the one that keeps him in the gym long enough to earn the pretty ones. The medals are a lagging indicator of how well a boy learned to fall.
I’m also a sixty-year-old graduate student. Two degrees in progress, an MBA and a PhD, both started decades after the age when a transcript says you’re supposed to start them. So I’ve sat in a lot of classrooms recently, and I can report with some authority on what is not in any university catalog in this country. There is no required senior seminar in abject failure. There is no syllabus for the thing that is actually going to happen to the graduate in the gown, somewhere in her first eighteen months out: the launch that craters, the model that was confidently wrong, the organization she joined for looking unsinkable, which turns out to have been quietly taking on water the whole time she was studying how to get hired by it.
So consider this the makeup course. It’s pass/fail; the reading covers the last twelve months, and it’s long since the last twelve months were generous. One detail, before the syllabus, that I want you to hold onto for the whole hour: every single failure on this list was committed by people who had the credentials. Not by dropouts. Not by amateurs working out of a garage. By the certified, the audited, the board-approved, the Ivy-trained, the people with exactly the kind of diploma the young woman in the gown is about to receive. That’s not incidental to the course. That is the course.
Unit One: The diploma certifies the paperwork, not the truth underneath it
Start where the credential is densest, because that’s where the rot was best hidden.
In September 2025, a company called First Brands filed for bankruptcy. You have never heard of First Brands, but you have held its products. It makes the FRAM oil filter, the Raybestos brake pad, the Anco wiper blade, and the unglamorous parts in the bin at the auto store. When it filed, it listed liabilities of somewhere between ten and fifty billion dollars against a business that booked roughly five billion in sales a year. Sit with that ratio for a second. Then consider that federal prosecutors went on to indict the founder and his brother on conspiracy, wire and bank fraud, and money laundering, alleging that the same collateral had been pledged to more than one lender at the same time. By February 2026, the company had filed notices to close four Ohio plants and cut 1,267 jobs.
Six weeks before First Brands filed, a subprime auto lender, Tricolor, collapsed into Chapter 7 on the same trick. The U.S. Attorney for the Southern District of New York later unsealed indictments against the founder and three other executives and called the company, in plain English, a “financial crimes enterprise,” saying fraud “became an integral component of Tricolor’s business strategy.” JPMorgan took a $170 million charge between the two failures. Jamie Dimon, asked about it, said the line that traveled: “When you see one cockroach, there are probably more.”
Now, the people in this field house, with the finance concentrations, the brand-new CPA results, and the offer letters from firms whose entire product is the word trust: this unit is addressed to you specifically, so lean in.
Every one of those deals had paperwork. Audited financials. Credit ratings. Loan tapes reviewed by warehouse lenders whose job, whose only job, is to look at the loan tapes. The certifications were issued by credentialed professionals who do exactly what their credentials certify they can do. And the fraud went through for years anyway, until something unrelated stumbled and the lights came on by accident. The audit didn’t catch it. The audit dressed it. Your diploma will accurately certify that you can produce the document. It will not certify that the document is true, and it will offer you no protection at all from being the person whose name is on it when the truth shows up late and angry.
The case to sit with the longest here isn’t even the fraud. It’s 23andMe. A company that, at its peak, had collected the genetic information of more than fifteen million people, around eighty percent of whom had agreed to let it be used for research, including myself, my spouse, and all four of my children, two biological and two adopted, in order for our family to understand more about where we all came from. It filed for bankruptcy in March 2025. More than two dozen state attorneys general objected to what happened next on the entirely sane ground that a human genome is not the same category of object as a forklift. The court ran the auction anyway. The DNA sold for $305 million to a nonprofit run by the same executive who’d run the company into the ground in the first place. The legal instrument that moved fifteen million genomes was Section 363 of the Bankruptcy Code, the ordinary, unremarkable “sell it free and clear” provision. The privacy ombudsman, as provided for in the law in a case like this, could write a report. He could not stop the sale.
Nothing malfunctioned. That’s the part I need the graduate in the gown to understand before she signs anything. The bankruptcy system performed flawlessly. It took a distressed asset and found it a buyer using the exact procedure it would use for a warehouse lease, since the people who built that procedure, all of them credentialed, never imagined feeding fifteen million genomes through it. An institution is a procedure with a building around it. A procedure does not have a category called unthinkable. It will process the unthinkable with precisely the same calm it brings to the routine, and it will generate clean paperwork the entire time.
Unit Two: Redundancy is not safety, and the backup is usually a second copy of the problem
The first thing any competent field teaches you is to have a backup. Two engines. Two data centers. A spare. The 2025 to 2026 academic year was, among other things, a long and very expensive seminar in why that sentence is only half finished.
On the night of October 19, 2025, an automated system inside Amazon Web Services deleted the DNS records for one of its own core databases in Virginia. The system that was deleted was a safety system. It had a redundant twin, built for exactly this, so that if one copy failed, the other would take over. The two of them raced each other, and the cleanup routine that was supposed to tidy up after the race deleted the live record instead. For roughly fifteen hours, Snapchat was down. So were Fortnite, Venmo, Ring doorbells, United Airlines’ booking system, and the United Kingdom’s tax website. The redundancy didn’t fail, despite being redundant. The redundancy was the failure. Two safety processes built to protect each other collided, and the safety routine deleted the thing it was supposed to guard.
Three weeks later, Cloudflare did it differently and landed in the same crater. A permission change in the database caused a query to return duplicate rows, doubling the size of a configuration file that Cloudflare’s own bot-detection software loads into memory. The file blew past a memory limit. The memory limit existed for safety. The traffic-routing software started crashing in a loop, and X went down, and ChatGPT, and Spotify, and Discord, and Zoom, and, with a symmetry you could not put in fiction, Downdetector, the website you visit to confirm that the thing that’s down is in fact down. Two weeks after that, Cloudflare went down a third time and conceded that the fixes it had promised after the second outage were not yet fully in place.
A configuration file that exists to tune the system. A memory limit exists to keep the system safe. The tuning broke it, and the limit was the wall it shattered against. When the graduate in the gown builds her own version of this, and she will, the question to ask is never “Is there a backup?” Everyone has a backup. The question almost no one asks, the question that separates the people who survive their own infrastructure from the people who write the postmortem, is: Does the backup fail independently of the thing it’s backing up? If it fails the same way at the same time for the same reason, it was never redundant. It was a duplicate of the single point of failure, and the org paid extra to feel safe about it.
Unit Three: The warning almost always exists; the failure is that no one owns the response
In late January 2025, an Army helicopter and a regional passenger jet collided in the air near Reagan National Airport, and sixty-seven people died. When the National Transportation Safety Board released its final report a year later, the finding that should have ended careers was this: the airspace contained a helicopter route placed too close to a runway approach, and the agency responsible for that airspace had its own data showing 15,214 dangerously close events between helicopters and commercial planes in that exact corridor over three years. Eighty-five of them were genuine near misses. The warning was not missing. The warning was not subtle. The warning had been sitting, fully formed, in a federal database for years, waiting for a person whose job was to read it and act.
You will spend your entire career being told, implicitly, that gathering the information is the work. It is not. Gathering the information is the cheap part, the part that generates dashboards and quarterly decks and the comfortable feeling of diligence. The expensive part, the part with no class grades and no dashboard measures, is the institutional act of looking at something you already know and changing what you do in response. Sixty-seven people died in the gap between the data and the decision. That gap is not empty. It is crowded with competent, credentialed people who all had access to the warning and not one of whom owned the response. Try, for your whole career, not to be a resident of that gap. It has more people in it than you’d ever guess, and most of them have very good degrees.
The Department of Government Efficiency belongs in this unit, and it belongs here as the largest single illustration of the principle in the entire reading. Set aside, for a moment, every argument about whether government should be smaller. That’s a real argument, and it is not this one. This one is about what happens when a credentialed operation decides the warning is meant for others.
DOGE was created by executive order in January 2025, with a promise of $2 trillion in savings. The number fell to one trillion, then to a hundred and fifty billion, then to a public “Wall of Receipts” that, on inspection by reporters with calculators, listed an $8 million contract as $8 billion, counted single contracts two and three times, counted contracts that had ended under the previous administration, and counted credit-line ceilings as if they were realized savings. An independent, nonpartisan analysis estimated that the effort’s actions cost taxpayers roughly $135 billion in a single fiscal year. At the same time, self-reported savings could not be verified by any external auditor, including right-leaning ones, at face value.
But the part for this unit, the part about owning the response, is the firings. DOGE fired, then scrambled within roughly twenty-four hours to rehire several hundred employees responsible for the U.S. nuclear weapons stockpile, some of whom had already lost their email access before anyone could tell them they still had jobs. It fired, then, within forty-eight hours, tried to reverse, with USDA staff working the bird-flu response at the height of an outbreak that had already killed 148 million birds. The warnings here were not buried in a database. They were the literal job titles of the people being fired. Nuclear weapons. Avian influenza. The operation moved fast enough that the title on the badge was not read before the badge was deactivated. The courts, for their part, repeatedly found the operation had acted unlawfully, and the unit quietly ceased to function eight months before its own charter expired, its director’s denial of its demise contradicted by the reporting within a single day.
I’m not telling you this to score a political point. I’m telling you this for a simpler reason. The graduate in the gown will be handed, at some point, a mandate, a deadline, and a feeling of urgency, and the most seductive sentence in professional life is: “We don’t have time to check.” DOGE is what that sentence costs at scale. The warning was the job title. Nobody read the badge.
Unit Four: The correction exists, and it arrives far too late to be a save
There is a particular kind of comfort, especially for the idealistic graduate, in the belief that systems are self-correcting. That if something goes badly enough wrong, a mechanism, a court, a regulator, an election, a guardrail, will engage and set it right. The reassuring thing is that this belief is true. The devastating thing is the timing.
In April 2025, the president imposed the steepest tariffs the United States had levied in more than a century. Global markets fell in the worst week since the early days of the pandemic. The constitutional structure that exists precisely to constrain that kind of unilateral economic action did, in the fullness of time, constrain it: the courts ruled the tariffs illegal, and in February 2026, the Supreme Court affirmed it, six to three. Roughly $166 billion had to be refunded to more than 300,000 businesses. The check worked. It worked for eleven months and one market crash after the damage was inflicted.
The same year produced the longest government shutdown in American history, forty-three days, and then, four months later, a longer partial one, seventy-six days. The federal budget process is itself a safety mechanism. The requirement that elected people affirmatively decide to keep the lights on is supposed to force deliberation. It performed exactly as designed. It produced precisely the deadlock it is structurally capable of producing, twice in one year. Several hundred thousand federal workers went without pay during the gap. Roughly forty million people on food assistance watched their November benefits hang in the balance on an emergency court order.
Here is the lesson, and it is the one your idealism is least likely to accept on graduation day. The presence of a correction in a system does not guarantee that it will arrive in time to matter. A guardrail that engages after the car is already in the ravine is a genuinely good thing to have installed, for the sake of the next car. It is not safe for the car in the ravine. The graduate in the gown should build the kind of work and choose the kind of institutions where correction is fast and close to failure. And she should be ruthlessly honest with herself, for the rest of her career, about how rare that kind of work actually is, and how much of professional life is conducted in systems whose corrections are real, and slow, and arrive for the next person.
The unit I am not going to resolve, since resolving it would be a lie
There is one more, and I am going to break the shape of this seminar to handle it, since keeping the shape would require me to lie to you, and you’ve been lied to enough by the time you reach a stage in a gown.
On June 12, 2025, Air India Flight 171 lifted off from Ahmedabad and came down thirty-two seconds later into the hostel block of a medical college a little over a mile from the runway. Of the 242 people on board, 241 died. One man, in seat 11A, walked out of it. At least nineteen more people died on the ground. It was the first fatal crash and the first hull loss in the history of the Boeing 787, an aircraft that had flown without a fatal accident since 2011. The preliminary report found that both engine fuel cutoff switches moved from RUN to CUTOFF within seconds of liftoff. The cockpit voice recorder caught one pilot asking the other why he had cut off the fuel, and the other saying he hadn’t. A year later, investigators on two continents do not agree on what happened, and the families do not have an answer.
I am not going to extract a tidy lesson from that, because there isn’t one, and a commencement speaker who manufactures wisdom out of two hundred and sixty deaths is doing exactly the thing this whole seminar is warning you against: dressing the absence of an answer in the costume of one. Sometimes the failure does not yield a takeaway on your timeline, your career’s timeline, or your lifetime. Sometimes the most rigorous and honest thing a credentialed person can say is, “We do not yet know,” and the people who deserve to know are still waiting. That sentence is not a failure of analysis. It is the highest form of it. The course does not always pay you back in insight. Sometimes it just takes, and asks you to keep the names in mind as it does, and to refuse the comfort of a moral you didn’t earn.
The final question is one.
So here is what the field house ceremony will not say, which means I have to.
Every failure in this seminar was the work of the successful. I want to be precise about that, since it is the entire argument, and the graduate in the gown is about to walk directly into it. Amazon employs some of the best infrastructure engineers money can buy, and its safety automation deleted its own database. First Brands had audited financials, produced by accountants with the same credentials as half this room is about to earn. The FAA had the data, all 15,214 events of it. The bankruptcy court had decades of settled, expert procedure, and it used that settled craft to auction fifteen million genomes with a clean docket. People staffed DOGE credentialed enough to be trusted with the federal payment system, and they deactivated the badges of the people who guard the nuclear stockpile without reading the badges. In every single case, the thing everyone trusted, the credential, the audit, the redundancy, the procedure, the constitutional check, was either the precise thing that failed or the precise reason no one was watching the thing that did.
Your diploma is real. You earned it, and you should be proud of it, and I mean that without a flake of irony. It certifies, correctly, that you learned what they taught you, and that isn’t nothing; it is, in fact, a great deal. But there is exactly one thing it cannot certify, the one thing no speaker at any podium in any field house this May or June will say into the microphone, and it is the entire content of the senior seminar that wasn’t on your transcript: being good at the success part does not inoculate you against the failure part. It never has. It only changes the shape the failure takes and moves it somewhere your training has specifically taught you not to look.
So look there. That is the whole final exam, one question, and you will retake it for the rest of your working life. When the year goes wrong, and it will, with witnesses, the only thing that will distinguish you from the very credentialed people in this seminar’s reading list is whether you are the kind of professional who keeps her eyes on the tired metal even when the inspection schedule, signed and stamped by someone with your exact degree, says the metal is fine.
Nobody grades that question. There’s no ceremony for getting it right, the same way there’s no medal for the twelve-year-old boy who finally learned to fall off the high bar without breaking himself. It just quietly determines everything that happens after the chairs are folded and put away.
Congratulations. Welcome to the part of the curriculum that was always the point.
For a deeper dive into colossal, abject failure, please check out the research paper that was the basis for this essay:
When Systems Cracked: Monumental Failures From May 2025 to May 2026
Author Note
Grace Ann Hansen is an independent researcher and writer, and an MBA & PhD graduate student in health informatics and artificial intelligence. She is also a published author, a professional musician, a gymnastics coach, and a queer transgender woman living in Sioux Falls, South Dakota. She corrects all her papers and articles with Grammarly, because even though she has deep thoughts, she has shallow patience for punctuation. She uses Anthropic’s Claude in Research mode for source location and verification on cited factual claims; all interpretation, argument, and prose are her own. Correspondence concerning this article should be addressed to Grace Ann Hansen at grace@graceannhansen.com.