The Spy Who Knew Too Much: An Ex-CIA Officer’s Quest Through a Legacy of Betrayal
Book reflection
The Ones Who Wouldn’t Let Go
I’ve been on a spy biography kick for the better part of a year: WWII through the Cold War. Operation Mincemeat. The Billion Dollar Spy. The Sisterhood. The Recruiter. A dozen others. I tell myself it’s intellectual curiosity about a particular era. And it is. But the real pull is something more specific: I’m fascinated by how these minds work.
Not the tradecraft. Not the dead drops and the brush passes. The cognitive machinery underneath. How an intelligence officer processes mountains of information, much of it deliberately false, and arrives at a conclusion. Or gets diverted from the truth. Or holds a conviction for decades that everyone around them insists is wrong, and turns out to be right. Or wrong. And sometimes can’t tell the difference until it’s too late, or too late to matter.
No book I’ve read captures that more completely or more troublingly than Howard Blum’s The Spy Who Knew Too Much, which I finished last month.
Pete Bagley and the Problem of Knowing
Tennent “Pete” Bagley was a rising star in the CIA’s Soviet Bloc division. In 1962, in Geneva, he interviewed a KGB officer named Yuri Nosenko who was offering his services. Something felt wrong from the start. Nosenko’s claimed knowledge of KGB procedures was shallow, his motives murky, his story riddled with inconsistencies. Bagley came away convinced: Nosenko was a double agent, sent by the KGB to plant disinformation.
The CIA eventually disagreed. Nosenko was rehabilitated, given a salary, an apology, and permission to stay in the US. Bagley’s career quietly derailed. He retired, increasingly isolated, still carrying a conviction nobody wanted to hear.
Most stories end there. The stubborn analyst, broken by the institution, fading into irrelevance.
Bagley didn’t fade. He kept pulling threads.
What Behavioral Economics Sees
There’s a framework I find myself applying when I read these histories — Transform the Human Experience, or THX, which draws on the work of Kahneman, Tversky, and the behavioral economists who followed them. A lens for understanding how humans actually make decisions under uncertainty, as opposed to how rational-actor models assume we do.
Applied to the intelligence world, it’s uncomfortably clarifying.
Prospect theory holds that losses loom larger than equivalent gains. We feel the pain of losing something more acutely than the pleasure of gaining the same thing. The conventional read of Bagley is that loss aversion trapped him: his reputation was the reference point, his career’s derailment the loss, and no amount of rational calculation could pull him away from the quest to recover it. He couldn’t let go because letting go meant the loss was permanent and the loss was his fault.
Fair enough. But prospect theory also tells us that people operating in the loss domain become risk-seeking. They’ll take bets that a gains-domain person would never take. The CIA under William Colby was sitting on its gains, protecting its credibility, risk-averse about reopening a concluded case. Bagley, already in the loss domain, had nothing left to protect. He could afford to keep going precisely because the institution had already taken everything.
Anchoring explains how he got so dug in. Before Bagley ever interviewed Nosenko the second time, James Angleton had handed him the Golitsyn file. Golitsyn was a prior KGB defector whose account overlapped suspiciously with Nosenko’s. That briefing set the anchor. Every subsequent data point got evaluated relative to it. When Nosenko contradicted himself, Bagley saw confirmation. When Nosenko’s story held up under pressure, Bagley saw sophistication. The anchor didn’t just influence his judgment, it structured the entire frame through which evidence arrived.
That’s what makes disinformation so effective in the right hands. The KGB wasn’t just lying. They were engineering signals calibrated to activate these cognitive patterns in particular analysts. Operation Mincemeat, the WWII deception that fed the Germans a false invasion plan through a planted corpse, worked on exactly this principle. You don’t defeat a good analyst by giving them false information. You defeat them by giving them false information that confirms what they already half-believe. The anchor does the rest.
Confirmation bias runs underneath all of it. One account of Blum’s book notes that Bagley’s broad deception theory might be confirmation bias, interpreting evidence in ways that confirm suspicions rather than considering alternatives. A fair critique. But sometimes the person with confirmation bias is confirming something true. The bias and the accuracy aren’t mutually exclusive. And the institution that dismissed Bagley had its own confirmation bias running the other direction — toward the conclusion that there was no mole, because a mole meant catastrophic institutional failure.
Both sides were anchored. Both were loss-averse in their own domain. Only one was right.
The Books I Keep Returning To
The Billion Dollar Spy, David Hoffman’s account of CIA asset Adolf Tolkachev in Moscow, runs the same dynamic in reverse. A Soviet scientist walks into the US Embassy volunteering to spy, and the CIA’s first instinct is suspicion. Is he genuine? Is he a provocation? The Nosenko case had poisoned the well. Institutional memory of being burned by a double agent made officers more likely to dismiss a genuine one. The anchoring from one case distorted judgment in the next.
The Recruiter, Douglas London’s memoir of his CIA career, puts flesh on the organizational pressure underneath. The institution doesn’t just make bad calls, it creates environments where certain kinds of knowing become career risks. Being right in the wrong direction is often worse than being wrong in the right one. Bagley knew this. He was living it.
What strikes me across all these books is that the analysts and officers who got it right shared something no cognitive framework captures well. They built relationships. They stayed curious about people, not just information. They treated former enemies as humans with complex motives rather than as data points to be categorized and filed.
The Cemetery
Bagley spent decades working the problem. Barred from CIA archives, most of the principals dead by the time the Soviet Union collapsed and he reconstructed the case. So, he collaborated with former KGB operatives, men who had been his adversaries, because they were the only ones left who knew what had actually happened.
The truth, when it came, didn’t come through analysis. It came through trust built across ideological lines.
One of those former Soviet contacts led Bagley to a cemetery. To the grave of John Paisley, the CIA official whose suspicious death had reignited Bagley’s quest decades earlier. Paisley, Bagley had long believed, was the mole.
Standing at that grave, led there by a former enemy, Bagley had something the institution never gave him: confirmation. Not official. Not on the record. Not the kind that restores a career or rewrites a file. A quiet arrival at the truth he’d carried for thirty years.
What the Rational Actor Would Have Done
The rational-actor model would have told Bagley to quit. The expected value of continuing — discounted by probability of success, institutional barriers, the passage of time, personal cost — was negative. Any clean calculation said let it go.
The rational actor would have been wrong.
Prospect theory explains why Bagley couldn’t stop. Anchoring explains how he got so dug in. Confirmation bias explains some of what he saw and some of what he missed. None of it explains the cemetery.
That part required something the frameworks don’t account for: the willingness to humanize people on the other side of the ideological line. To build enough trust with a former adversary that he’d walk you to the grave of the man you’d been hunting for thirty years.
The frameworks map how we get trapped. Getting free took a relationship, a former enemy, and thirty years of everyone telling him he was wrong.