The two previous posts in this series made a case for the adversarial mindset and then got honest about what it costs. This one is about something different: how it actually works. Not as a concept. As a practice.

Because there is a version of adversarial thinking that is mostly aesthetic. The hoodie, the “think like an attacker” framing, the red-tinted LinkedIn post. Identity wearing the costume of a skill.

And then there is the version that is a structural role. One you can enter and exit deliberately. One with a method. One that does not require you to be a particular kind of person, because it is not about personality. It is about function.

That distinction matters more than people want to admit.

It Was Never About the Individual#

Red teaming was not invented in a security operations center.

It started in the military. During the Cold War, both the US and the Soviet Union used independent groups whose specific job was to challenge their own war plans. Not advise. Not improve. Challenge. Find the assumptions the planners were too close to see.

Then business adopted it. Pressure-testing strategy before it became expensive. Stress-testing M&A decisions before they became irreversible. The premise was the same: the people who built the plan cannot see its weaknesses as clearly as someone who was not involved in building it.

Cyber came later. AI red teaming is happening now. The label changes. The concept does not.

But if you want to trace this idea to its actual origin, you need to go further back than any of that.

The term “Devil’s Advocate” comes from the Catholic Church. The Advocatus Diaboli. Before someone could be canonized a saint, the Church appointed a specific person whose only job was to argue against it. Find the flaws. Challenge the evidence. Make the case for the other side.

This was not an insult to the candidate. It was not a personal attack. It was a structural role.

The institution understood something most organizations still do not: unchallenged consensus produces bad outcomes. It does not matter how good the people are, how senior the room is, or how thorough the process feels. If everyone is converging on the same answer and nobody is assigned to break it, you are not finding truth. You are confirming assumptions.

Red teaming is the same idea. Not a person. Not a personality type. A role. Someone whose job in that room is to think like the adversary. To ask what everyone else is not asking. To find what the main thinking missed.

The 10th Man#

If nine people in a room reach the same conclusion, it becomes the tenth person’s institutional duty to argue the opposite.

Not because they think the nine are wrong. Not because they are contrarian by nature. Because someone has to find what everyone else missed. Because the most dangerous moment in any planning process is not when disagreement is high. It is when agreement feels obvious.

The Bin Laden raid is the most documented recent example. Before the mission, the CIA ran a dedicated red team. A separate group, kept apart from the main planning effort, whose specific task was to argue that Bin Laden was not in that compound.

Not to stop the mission. To make the case survive scrutiny.

To find every weakness in the intelligence before the commitment was made. To test the assumptions before they became irreversible. The goal was not to prove the operation wrong. The goal was to ensure the operation could withstand being proved wrong, and still hold up.

That is the definition: test and retest all available information for weaknesses and potential inaccuracies. Not nihilism. Rigor.

The question the adversarial thinker is always asking is not “is this wrong?” It is “what would have to be true for this to fail?”

The Framework#

There is no checklist. No step-by-step. No framework to memorize and paste into a deck.

But there is a sequence. And if you follow it properly, you will arrive somewhere a generic plan cannot reach.

Take a mission. You have been asked to infiltrate an online far-right community. You need presence inside it. You need trust. You need information that does not come from news headlines.

Most people think they already know enough to start planning. They do not.

Step One: Learn - Until You Could Survive Contact#

Not until you feel informed. Until you could hold a real conversation with a suspicious member and not get burned in the first three exchanges.

That gap is enormous.

Most people’s mental model of a target comes from media coverage. Headlines, stock photos, surface characterization. For the far-right example: shaved head, Nazi tattoo, stock flag image. That same photograph has appeared in hundreds of articles. It is almost entirely useless as an operational model.

Because real environments do not look like their media representations. Slang shifts. Internal culture is specific. Symbols mean different things to different factions. Clothes matter. Locations matter. What you say in the first message matters more than you think.

Bellingcat published a piece on how not to interpret far-right symbols precisely because the public model was so wrong that it would get someone burned on contact.

The failure mode is stopping when you feel like you understand the topic. Learning is not done when you can write a paragraph about a group. It is done when you could survive contact with someone who actually lives inside it.

This applies everywhere. It is not about far-right forums. It is about knowing your target well enough that your plan is built on reality, not on what you assumed the reality was.

Step Two: Understand - Empathy Is Not Sympathy#

This is the step most people either skip or fake.

They are not the same thing.

Sympathy is feeling what someone else feels. That is an emotional state. It is not what you need here.

Empathy is being able to model how someone else thinks. That is a cognitive skill. And it is the one that matters for adversarial work.

You do not need to share the grief, the rage, the ideology, or the worldview. You need to understand the logic well enough to predict what they do next.

You can hold both at once: “I understand exactly why this person ended up here, and I completely disagree with it.” That combination is not contradiction. That is the goal. Not feeling it. Modeling it.

The moment you stop seeing the adversary as a threat category and start seeing them as a person, the model gets accurate.

They are not monsters. They are not superhuman. They have pressures, blind spots, ego, fear. They cut corners. They get tired. They care about things that have nothing to do with you. They make decisions based on incentives, not ideology alone.

You cannot predict a monster. You can predict a person. So that is what you model.

Step Three: Invent - “How” Is a Spectrum#

Because of steps one and two, you can now build something a generic plan would never reach.

The obvious move for the infiltration mission: create a fake profile, post some opinions, try to blend in.

The problem with the obvious move is that every trained member of a suspicious community knows how to test it. Five messages in, they will probe your history, your consistency, your knowledge of specifics that a fake account cannot fake for long.

The bold move: build something real. A fashion and lifestyle brand, aesthetically aligned with the community, real enough to have a social presence, real enough to sell products. A natural reason to exist in those circles that does not require maintaining a lie in every conversation.

Notice what makes this work: it is not more complex. It is more grounded. It is actually simpler to maintain a real identity than a fake one. The KISS principle applied adversarially.

This is the point. Adversarial thinking does not always push toward the unique, the novel, the complex. Sometimes the basics work because they are the basics. The creative leap is not always about invention. Sometimes it is recognizing that a simple thing, done properly, is more durable than an elaborate thing, done impressively.

The question at this step is never “if this is possible.” It is “how do we make this work.” If is binary. How is a spectrum. The moment your default becomes “that won’t work,” you have stopped thinking adversarially and started being a blocker.

You are here to say how. Not if.

Step Four: Convince - “Yes, And” Not “Yes, But”#

A good idea that cannot survive a skeptical room is not a finished idea.

Getting here is the part most practitioners skip because they think the work is done when the plan is built. It is not. The plan that exists only in your head, or only in a document, has not been tested yet.

Before you take it to anyone, ask: if I were the adversary of this plan, what would I hit first? Where does it depend on an assumption that might not hold? Where does it require cooperation from someone who has not agreed yet?

Red team your own plan before anyone else does. Not because you want to kill it, but because the version that survives you is the version that can survive the room.

And then, when you get into the room: “yes, and” instead of “yes, but.”

Leadership’s first reaction to the fashion-brand idea is: “Why not just a fake profile?” That is a reasonable question. It feels like the simpler answer. The instinct is to argue against it.

Do not.

“Yes, and” sounds like: “Yes, a fake profile is the fast, easy version - and that is exactly why every trained community member knows how to spot it in the first five messages.” Same point. No fight. You acknowledged what they said, you built on it, and you made your case without making it a debate.

The second technique: get ahead of the objection. Say it yourself before they do. Name the weakness in your own plan and address it. The person who identifies the flaw in their own argument before the room does is the person the room trusts.

The failure mode at this step is emotional. By the time you reach step four, you have learned the target deeply, you have built a model of how they think, and you have invented something you believe in. Pushback feels personal.

It is not. The plan is not you.

The Version That Becomes a Problem#

There is a version of the adversarial mindset that looks like adversarial thinking but is not.

It is the person who raises an objection in every meeting. Who never agrees with anything. Who has a problem with every plan. Who slows everything down not because they see real risk but because disagreement has become their brand.

That is not adversarial thinking. That is ego.

Adversarial thinking has a target. It is deployed in response to a specific question. It is scoped. It is time-limited. When it is done, you close the session and return to builder mode.

The person who cannot turn it off is exhausting to work with. They are not doing the work. They are performing the identity.

The practical version: decide in advance when adversarial mode is needed. Enter it deliberately. Exit it deliberately. Debrief. Reset. Do not let the mindset follow you into every meeting, every relationship, every conversation about something that does not require it.

Left of bang thinking is a skill. The failure mode is when it never goes right of bang. When you are always in anticipation mode, always threat-scanning, always preparing for what comes next, and never actually present for what is happening.

Know the difference between the person who brings adversarial thinking when it is needed and the person who has made not agreeing into a personality.

This post is the third in the Adversarial Mindset series. The first covered what the mindset is and where it comes from. The second covered what it costs you. This one is about how to use it without being used by it.

“To everything there is a season, and a time for every purpose under heaven.” - Kohelet (Ecclesiastes) 3:1