Every healthy system needs the boundary-crosser who breaks the rules to find the truth. The blind spot and the rulebook are the same document.
In 2011, Netflix did something that, described plainly, sounds like sabotage. They wrote a piece of software whose entire job is to log into their own production network, the live systems serving movies to millions of paying customers, and randomly kill servers. On purpose. During business hours. They named it Chaos Monkey, after the image of a wild monkey loose in a data center, ripping cables out of the racks, and they set it free in the one place every engineer is trained to protect above all else.
This is insane by every instinct of operational discipline. Your whole job is to keep production up; you do not build a robot to tear it down. And yet Chaos Monkey, and the broader Simian Army of failure-injecting tools Netflix built around it, is one of the most celebrated engineering decisions of the last fifteen years. It founded the discipline now called chaos engineering, copied by Amazon, Google, and most of the serious infrastructure on the internet. The logic is airtight once you see it: a system that has never been forced to survive a server dying at random will fail catastrophically the first time one does, at 3 a.m., unrehearsed. So you make the failure happen constantly, on your terms, in daylight, until surviving it is routine. The monkey doesn't weaken the system. It's the only thing that ever finds the weakness.
What Netflix built, without quite saying so, is one of the oldest figures in human storytelling. They built a trickster, and, crucially, they gave it a license.
In 1956, the anthropologist Paul Radin published The Trickster: A Study in American Indian Mythology, with commentaries by the mythologist Karl Kerényi and the psychologist Carl Jung. What the three of them were circling was a pattern so persistent across unrelated cultures that Jung classed it as a universal archetype. Radin's definition is the one to keep: the trickster is "at one and the same time creator and destroyer, giver and negator, he who dupes others and is always duped himself." He is both, order's enemy and order's renewer in the same body.
You have met him under a dozen names. Loki in the Norse myths, whose mischief is both the gods' constant headache and, repeatedly, the thing that saves them. Coyote and Raven across Native American traditions. Hermes in Greece, fittingly, the god of boundaries, of thieves, and of messengers, the one who moves between worlds. Anansi, the West African spider who owns all the stories. Sun Wukong, the Monkey King of Journey to the West, who breaks heaven itself. Lewis Hyde, in his 1998 study Trickster Makes This World, distills the whole archetype to two words: the trickster is a "boundary-crosser." He is the figure who goes where the rules say no one may go, and comes back with something.
And what he comes back with is always a truth the order was hiding. It is the trickster who crosses the forbidden line to steal fire from the gods and bring it to humanity. It is the trickster-energy in the child who says the emperor has no clothes, speaking the obvious thing the entire well-ordered court had agreed not to see. Jung's own phrasing names the function exactly: the trickster myth, he wrote, is "a present-day outlet for the most unashamed and liberating satire of the onerous obligations of social order, religion, and ritual." The trickster exists to violate the order so that the order can be seen, and, where it has rotted, replaced.
Here is the part that turns a charming piece of comparative mythology into an operational principle, and it is the most important sentence in this essay: a system maintained only by rule-followers cannot find the failure the rules forbid it to examine.
Think about what a rule actually is, inside an organization. It is a boundary someone drew, long enough ago that most people have forgotten the drawing, around a region everyone now treats as settled. "We don't touch the billing service." "Production failover obviously works, we tested it at launch." "That's just how the deploy process is." Each of these is a fence, and the diligent, conscientious, disciplined engineer's job is to operate within the fences. That's not a flaw in them; it's the definition of competence. But it means the rule-follower's competence is structurally bounded by the rules, and the most dangerous failure modes live exactly where the rules said not to look. The assumption everyone learned not to question is, by construction, the assumption no one is checking. The blind spot and the rulebook are the same document.
This is why the trickster is not a luxury or a personality type a healthy culture merely tolerates. He is structurally necessary, because he performs the one cognitive act the rule-followers are organizationally incapable of: he crosses the boundary on purpose to discover whether the thing on the other side, the failure, the rot, the obsolete assumption, is real. Netflix's senior engineers could not, by sitting in a room being careful, discover that their failover didn't actually work; being careful was the very habit that hid it. Only the monkey, who respects no boundary, could surface it. The myths are unanimous on the consequence of refusing this: a world with no trickster does not become more orderly. It stagnates. The unchallenged order ossifies, and then it dies, killed by a failure mode its own rules had forbidden it to examine.
Once you see the archetype, you see it everywhere in engineering, wearing modern clothes:
Every one of these performs the mythological function precisely: they cross the boundaries the rule-followers maintain, to reveal the truths the rules conceal. And here is the tragedy that the myths predicted three thousand years ago: the trickster is always unwelcome. Loki is bound to a rock. The Monkey King is crushed under a mountain. The court that the child embarrassed does not thank the child. The organizational instinct is to punish boundary-crossing, and the instinct feels like virtue: the boundary-crosser violated the order, disrupted the plan, made senior people uncomfortable, broke something on purpose. So we route around them, manage them out, and reward the tidy operator who never rocks the boat.
And in doing so we make a catastrophic trade and congratulate ourselves on it. A culture that suppresses all boundary-crossing has chosen ossification and called it discipline. The calm, tidy, no-surprises engineering org is not necessarily a healthy one. It may simply be one that has successfully exiled everyone capable of finding its hidden failures, a system that feels stable right up to the morning a 3 a.m. server death, the one the monkey was never allowed to rehearse, takes the whole thing down.
The naive reading of all this is "tolerate your rebels," and it's not enough, because the lone-wolf trickster is fragile and the org is stronger. The rogue who breaks production to prove a point gets fired; the heretic gets bound to the rock. If the trickster's function depends on the courage of unprotected individuals, it will be suppressed every time, exactly as the myths say.
The mature move, the one Netflix actually made, is to institutionalize the trickster: to create a sanctioned role with an explicit license to cross boundaries, so the disruption is protected, scoped, and expected rather than punished. Human cultures have known this trick for centuries: the court jester was the institutionalized trickster, the one man licensed to mock the king and speak the unspeakable truth precisely because the role was sanctioned, the license is what made the boundary-crossing safe to perform and safe to hear. Engineering has reinvented the jester, and you should hire one (or several):
The common thread is the license. The trickster's disruption is only productive when it is sanctioned, because sanctioning is what converts a fireable offense into an institutional function. You are not asking people to be brave heretics. You are building a role whose job description is heresy, and then protecting it.
This is not a license for chaos as a personality, and the distinction matters, because the trickster has an evil twin: the vandal, who breaks things for ego or for the thrill of breaking and calls it courage. Not every boundary-crossing is wisdom, and an org that mistakes mere disruption for the real article gets all the cost and none of the truth. The sanctioned trickster needs a tether, and the tether is purpose: the boundary is crossed to find the hidden truth, not to prove the crosser can cross it. The chaos game-day has a hypothesis and a rollback. The red team writes a report. The dumb question is asked in good faith. The discipline borrows from G.K. Chesterton's famous fence, the rule that you should not tear down a fence until you understand why it was put up. The true trickster doesn't ignore the fence; he crosses it to learn why it's there, and sometimes discovers the reason is sound and leaves it standing, and sometimes discovers there was never a reason at all. The vandal just kicks it over. A culture needs the boundary-crosser who seeks the why, not the one who seeks the noise.
Look at your team and find your tricksters, they're already there. The person who keeps asking the uncomfortable "why," the one who wants to break things on purpose to see what happens, the new hire who hasn't learned which questions are forbidden. Your instinct, and your org's, is to gently sand them down into rule-followers, because they're disruptive and the rule-followers are comfortable. Resist it. Those people are doing the one job your conscientious operators are structurally unable to do: finding the failure that lives where the rules said not to look.
Then go further than tolerating them: institutionalize the function. Schedule the chaos game-day. Stand up the red team. Make "ask the dumb question" a celebrated act with a protected asker. Carve out blameless permission to run the experiment everyone knows will fail. Give the boundary-crossing a license and a purpose, so it survives contact with your own immune system.
Because the myths have been telling us the same thing across every culture that ever told stories, and it is not a metaphor: a world with no trickster ossifies and dies. The calm, tidy, no-one-rocks-the-boat engineering culture is not the disciplined one. It's the one quietly accumulating the failure modes it has forbidden itself to examine, until the morning the unrehearsed server death arrives, and it learns, too late, that it should have hired a monkey. Hire your Loki. And protect him.
The boundary-crosser needs something real to check, not the order's account of itself.
A red team, a chaos game-day, a naive question: each crosses the comfortable assumption to find what the rules forbade anyone to examine. For an agent, the comfortable assumption is its own self-report, the tidy output, the green check, the claim that the thing was done right. What the boundary-crosser is actually after is what the agent did, and that has to exist as something inspectable rather than a surface you trust. Chain of Consciousness is that record: a tamper-evident account of an agent's reasoning, checks, and actions, the real thing your trickster can cross the fence and read, instead of the order's story about itself.
See Hosted Chain of Consciousness · verify an action chain
pip install chain-of-consciousness · npm install chain-of-consciousness