The Samurai and the Machine Gun
Why AI is breaking the institutions that delivered your judgment and how to still be standing when the sort is done
"When the facts change, I change my mind. What do you do, sir?
- John Maynard Keynes
"It is difficult to get a man to understand something when his salary depends on his not understanding it."
- Upton Sinclair
Ken Griffin runs Citadel. He has been, for as long as anyone in finance has been paying attention, an AI sceptic. He did not believe the technology would replace his analysts’ judgment.
He went home one Friday and said he was depressed.
He had watched Citadel’s masters and PhDs spend months on work that agents now did in hours. Not the analysts’ calculations. Their judgment work. The thing he had been telling himself, and his board, and the market, would remain inside the four walls of his firm because his people were exceptional.
His phrase: when you see work that used to be man-years done in days inside your own four walls, there is no theoretical position left to retreat to.
I have been thinking about this admission. The interesting thing is not that Ken Griffin updated. Many people are updating. The interesting thing is what he was actually witnessing.
He was not watching AI get smart. AI had been getting smart for years. He was watching the analyst’s judgment come apart from the institution that had been wrapped around it. The judgment was real. Considered alone, it could be replicated. The thing that had been making the work expensive was something else. It was the firm and the seat and the comp structure. It was the years of grinding. It was the credential that got the analyst into the seat in the first place. The institution and the judgment had travelled together for so long that nobody had to ask which was doing the work. The proxy had become the thing.
He just found out.
Six months ago I published an essay called Your Brain Is Not a Computer.
The argument was that human cognition is different in kind from algorithmic cognition. That we run on stories, embodied, narrative, contextual and that the institutions of the last forty years have been engineering this out of us in the name of the testable and the legible. AI is the accelerant. The thing that gets scarcer as the model improves is the cognitive architecture that the model cannot reproduce.
I still believe all of that.
What I had not yet said is that this cognition has been delivered, for two generations, through institutions that are now coming apart faster than I was prepared to admit. The story-thinking is real. The architecture is real. But the delivery system has been doing more of the work than anyone has been pricing, and the delivery system is dissolving in real time.
Your Brain Is Not A Computer (YBNC) told you your cognition was different in kind from the machine’s. It is.
This is the harder essay, about the institutions that have been delivering that cognition for two generations, and what happens now that they are coming apart.
The MBA is the cleanest version of the story, because almost everyone reading this has either earned one, paid for one, or is watching someone they love decide whether to spend two years and three hundred thousand dollars on one.
The MBA was never the knowledge. The knowledge in an MBA programme is in the textbooks, and the textbooks are free, and the textbooks have been free for thirty years. What an MBA actually sold was a credential as a sorting signal, a costly, time-consuming, hard-to-fake stamp that told an employer this person had survived a particular filter. The filter was the product. The textbook was the alibi.
This worked for seventy years because there was no cheaper way to verify the candidate. An employer faced with a stack of resumes did not have the time or the tools to evaluate each candidate’s actual judgment. The credential was the proxy. The credential and the judgment travelled together so reliably that the proxy became the thing.
It is no longer the proxy.
The Wall Street Journal ran a piece this spring titled There Is a Fire Sale on MBAs. Applications are down. Schools are discounting. The unemployment rate for workers under thirty-five with a master’s degree is at a twenty-year high. The plush, recently-built campuses, the most apparently solid pieces of the credential factory are sitting on the most fragile underlying business model in the system. The credential they were selling no longer carries anything the employer cannot verify another way.
McKinsey is the same story at a higher price point. The product was never the deck. The product was contextual judgment delivered into the C-suite by three twenty-somethings with the right credential, the right haircut, and the right partner standing behind them. The judgment was the thing. The apparatus around it was what got it through the door.
This spring McKinsey, BCG, Accenture, and Deloitte all agreed to embed OpenAI’s forward-deployed engineers directly inside their client teams.
The labs are not competing with the consultancies. The labs are being invited in. By the time the consultants realise they have automated eighty percent of their own intellectual property, they will have already trained the models that replace them.
This is what YBNC was missing.
The cognition is intact. The institutions are coming apart. The question of whether you were ever doing the work or whether you were riding inside an institution and mistaking the institution for the work is about to be answered in public, person by person, over the next ten years.
The frame I keep returning to comes from a conversation Gavin Baker had on Invest Like The Best. He had been watching The Last Samurai, the Tom Cruise film from twenty years ago, and he had asked everyone at his firm to watch it too.
The conceit. Tom Cruise plays a Civil War veteran hired to train Japanese peasants to fight the samurai. He ends up fighting on the samurai’s side. At the end he is killed by a peasant with a machine gun.
The samurai had a lifetime of pattern recognition. He had spent every year of his life learning a particular kind of cognition, embodied, contextual, hard-won. The peasant had none of that. The peasant had a machine gun.
The machine gun won.
Baker’s reading is the one I want to hold onto, because it is the version of the argument that survives both YBNC and The Samurai & Machine Gun (TSMG).
The machine gun is here. It is going to be here for the rest of our lives. The lifetime samurai who refuses to pick it up dies. The peasant with the gun but no judgment also loses, eventually, once the patterns start repeating and the situations get strange enough that pattern recognition matters more than rate of fire. The durable position is the composite — the samurai who has picked up the machine gun. Decades of cognition, plus operational fluency with the tool that was sold as the replacement for cognition.
The composite is rare. It requires holding two things together that feel mutually exclusive, the deep human cognition YBNC was defending, and the operational fluency with the technology YBNC was framing as the threat to it. The samurai who treats the machine gun as beneath him dies. The peasant who treats the lifetime of cognition as obsolete loses the next round. The person who holds both fluent in story-thinking and fluent in directing the agent is, for the next ten years, the most valuable and the rarest worker in the economy.
So the question TSMG leaves you with is not the question YBNC left you with.
YBNC asked whether you still had the cognitive architecture that the machines cannot reproduce and whether the institutions of the last forty years had been eroding it out of you, and whether you could get it back.
TSMG asks something harder.
For most of your career, for most of mine, we have not had to know what we were actually selling. The credential and the judgment travelled together. The seat and the work were one thing. The deck and the insight were one thing.
The institutions are coming apart now.
If you were the institution, if your value to the people around you was the credential, the brand, the seat, the apparatus that wrapped around you, you are about to find out whether you ever had anything else to deliver. Many will discover that the answer is less than they assumed.
If you were doing the actual work, the contextual judgment, the narrative cognition, the pattern recognition that twenty years of effort earns you, you are going to be fine. Better than fine. The premium on the work is going up, not down, as the institutions around it collapse.
But only on one condition.
You have to pick up the machine gun.
The samurai who refuses it dies in the field, regardless of how good a samurai he was. The brilliance of the cognition does not save you if you will not direct the tool that has been put in front of you. The veteran has the advantage only if the veteran wields the gun. Otherwise the gun walks the field, and the field is taken by whoever was holding it.
I wrote YBNC to tell you your cognition is real. It is.
I am writing TSMG to tell you the institutions you have been delivering it through are not.
The next ten years are going to sort the proxy from the thing, in every profession on earth. The question is not whether you can survive that sort. The question is whether you can do it on purpose.
You can.
But you have to know which one you were.
Next week, I’ll tell you what my personal plan…
"In times of change, learners inherit the earth, while the learned find themselves beautifully equipped to deal with a world that no longer exists."
- Eric Hoffer
Great essay, Ahmed. As usual, you are speaking to many of us on a level we are generally not accustomed to. And increasingly rarely find. I am still learning how it works, but I picked up the MG. I have been fighting the credentialists my whole career. This is excellent work, and I look forward to the plan.
brilliant tying it all together.