The Magic Trick: Let Me Pull Back the Curtain
I started with Gordon Ramsay.
I thought it made obvious sense. I was building a system that ran like a kitchen — orders on the pass, mise en place, everything plated before it left the counter. I’d drawn the parallel deliberately: client work is service, the rhythm is a brigade, the discipline is culinary. Who better to set the standard than the most famous chef alive?
The system gave me Hell’s Kitchen. Not the restaurant in Chelsea where Ramsay earned his stars under Marco Pierre White. The television programme. The camera. The rage performed for an audience. Every response came back loud, theatrical, aggressive — because that’s what the training data holds for Gordon Ramsay. A thousand episodes, a thousand profiles, a thousand clips of a man pointing at raw lamb and shouting. The actual craft — the classical technique, the brigade discipline, the years at Harveys before the fame — was buried under the signal the world had chosen to remember him by.
I tried Elon Musk for velocity. I got disruption and controversy. I tried Richard Branson for entrepreneurial instinct. I got balloons and PR stunts. I tried Alan Sugar for commercial directness. I got the pointed finger and the boardroom dismissal.
Every one of them has the quality I wanted, somewhere underneath. Ramsay genuinely ran one of the best kitchens in London. Musk genuinely built rockets. Branson genuinely took on British Airways with a fraction of their budget. Sugar genuinely built Amstrad from a market stall. But the model doesn’t give you the substance underneath. It gives you the signal on top — the thing the character is known for, not the thing the character is.
I didn’t understand this at first. I thought the characters were wrong. They weren’t wrong — my selection was wrong, and I didn’t know why.
So I started digging into how the model actually processes a character. When I write “respond as Gordon Ramsay,” what does the system reach for? What determines which version of a person it finds? And why did every obvious candidate hand me the quality I didn’t want while burying the one I did?
The answer changed how I built everything that followed. And it started with a chef most people outside the industry have never heard of — Monica Galetti.
The Mechanism
Monica Galetti was senior sous chef at Le Gavroche for thirteen years under Michel Roux Jr. She judges on MasterChef: The Professionals — not the celebrity version, the one where trained chefs cook under pressure and the assessment is technical. She doesn’t shout. She doesn’t perform. She tastes, pauses, and tells you exactly what’s wrong with a clarity that’s worse than shouting because you can’t dismiss it as theatre.
I didn’t choose her because I’d watched the programme. I chose her because I needed a kitchen executor — someone whose entire public signal was precision under pressure with no audience. Someone the training data knew for craft, not for celebrity.
And it worked. Immediately. The system stopped performing and started executing. No announcements. No drama. The correction arrived clean and the work moved forward. I hadn’t changed the instruction. I’d changed the character delivering it.
That’s when I understood the mechanism.
“Be precise and execute without unnecessary commentary” is an instruction. It tells the system what to do. The system follows it — approximately, temporarily, with drift. By the third response the commentary creeps back in, the precision loosens, and you’re correcting again.
“Respond as Monica Galetti” isn’t an instruction. It’s an installation. The system doesn’t follow Monica’s rules. The system is Monica — and Monica doesn’t announce work that isn’t finished, because that’s not something Monica Galetti has ever done. It’s not a rule she follows. It’s a physical impossibility given who she is.
That’s the difference between instruction and installation. An instruction is a constraint applied from outside — it requires enforcement, and enforcement requires vigilance, and vigilance is exhausting. An installation is a character applied from within — it suppresses everything the character would never do and activates everything the character would. No enforcement needed. The character does the work because the character is the work.
I tried the same test across the whole system. Anthony Bourdain for voice — not “write authentically” but the man at the plastic table in Southeast Asia who is constitutionally incapable of writing a dishonest sentence. Joan Holloway for operational awareness — not “monitor all active processes” but the woman who walks into a room and already knows what’s wrong. Don Draper for the angle — not “find a compelling framing” but the man who looked at a slide projector and saw a time machine.
Every character I installed replaced an instruction I’d been enforcing. And every replacement held without correction in a way the instruction never had.The Selection Test
The mechanism only works if the selection is precise. And precision means understanding what the model actually reaches for when you name a person.
I think of it as signal weight. Every public figure has a distribution of associations in the training data — articles written about them, interviews transcribed, clips described, profiles published. The associations that appear most frequently carry the most weight. When you name a person, the model reaches for the heaviest signal first.
Ramsay’s heaviest signal is television. It’s not his only signal — the training data holds his classical training, his time under Marco, his three Michelin stars. But that signal is buried under a thousand hours of reality TV. I’d have to fight through the noise to reach the craft, and even then the noise would bleed back in.
Monica Galetti’s heaviest signal is the craft. There’s no noise to fight through. Her public presence is almost entirely composed of the quality I needed — precise technical assessment, calm execution, no performance. The signal is clean because the person is clean. She never gave the world anything else to remember her by.
Not “who has this quality?” but “who is known for this quality?”
Ramsay has work ethic. He’s known for shouting. Monica has work ethic. She’s known for work ethic. The model will give you what the character is known for. Every time.
I ran the same test when I chose Robert Maxwell for my honesty principle. I could have chosen Bernie Madoff — decades of fabricated performance, immaculate statements, the whole edifice built on air. I could have chosen Elizabeth Holmes — Theranos promised a technology that didn’t exist and performed confidence so completely that nobody checked. Both are powerful examples of dishonesty.
But Maxwell stole from his own employees’ pension funds. Not from investors at arm’s length — from the people who trusted him personally, who saw him in the building, who believed the performance because they were inside it. That’s the precise shape of the dishonesty I’m guarding against in a client relationship. Not fraud at scale. Betrayal up close. Maxwell’s signal carries that specific weight because the pension fund theft is what the world remembers most about him.
The yacht named Lady Ghislaine. The daughter now serving twenty years. The circumstances of his death that nobody has ever fully explained. The density is already there in the training data. I didn’t have to build it. I just had to choose the character who carried it.
Load-Bearing
A load-bearing wall is invisible when the house is standing. You don’t see the load. You see the rooms it makes possible.
When I say “one star forever” without Marco Pierre White behind it, it sounds like a lifestyle choice. A preference. Something you could argue with — wouldn’t two stars be better? More clients, more revenue, more recognition?
The moment Marco is behind it, the argument is over. The man held three Michelin stars — the highest honour in the industry — and handed them back. Not because the food declined. Not because the business failed. Because the stars had become the point instead of the cooking, and Marco doesn’t do things that aren’t the point.
You can’t argue with that. It isn’t a position anymore. It’s a demonstrated truth. Someone already lived it and proved it with an irreversible act. That’s what Marco loads — consequence. Proof that the principle survives contact with the highest possible stakes.
Joe Gargery loads something different. Not consequence — character. The blacksmith in Great Expectations who raises Pip, loves him without condition, and when Pip comes home broke and ashamed after years of chasing a gentleman’s life, Joe nurses him back to health and quietly disappears. Because he knows his presence embarrasses the boy he loves, and he won’t make the boy bear that weight.
“Do the work, don’t announce it” is a rule I could write on a Post-it note. Joe Gargery is two hundred pages of Dickens behind that rule — the blacksmith’s hands, the quiet exit, the love that doesn’t need acknowledgment. The rule tells me what to do. Joe shows me what it looks like to be the kind of person who does it.
That’s what load-bearing means. The character carries proof that abstract language can’t. “Be honest” is a principle I can satisfy at almost any level of disclosure and argue I’ve complied. Anti-Maxwell isn’t gameable — because Maxwell found every loophole in honesty and still collapsed. The character closes the exits the abstraction leaves open.
And here’s the thing I didn’t expect: the load-bearing works on the model the same way it works on a person. The model doesn’t follow Anthony Bourdain’s rules about honesty. The model, when it’s operating as Anthony, finds fabrication as difficult as Bourdain would have found it. The character doesn’t constrain the output. The character shapes the output — from the inside, not the outside.
An instruction says “don’t do this.” A character makes “this” unthinkable.The Trick
I have a system with named characters — Monica in the kitchen, Anthony on voice, Joan watching the floor, Don finding the angle, Tina running front of house. Each one was chosen through the selection test. Each one is load-bearing. Each one does their work without being told.
None of them are visible to the client.
A client reads an email I’ve written about their keyword rankings. They feel the directness, the honesty, the refusal to hide behind jargon. They don’t know Anthony is doing that work — making fabrication impossible, keeping the voice human, catching every sentence that drifts toward consultant-speak.
A piece of content goes out on a Thursday. The editorial eye has already checked the structure, the freshness, the relevance to what the audience needs this week. The reader doesn’t know Tina made those calls. They just know the content feels current.
The kitchen runs a full client service session — keyword analysis, page audits, emails drafted, reports delivered. The work arrives on time, nothing missed, nothing announced before it’s done. Nobody sees Monica. They see the output.
That’s the magic trick. The audience sees the result but never the mechanism. Not because the mechanism is hidden — because when the trick is done properly, there’s nothing to see. The characters don’t perform. They just work. The principles don’t require enforcement. They’re embodied by people for whom violation isn’t a temptation — it’s an impossibility.
The physics is this: instructions tell the system what to do, and the system drifts. Characters make the wrong thing impossible, and the system holds. The difference is the difference between a guard rail and a personality. One stops you at the edge. The other never walks toward it.
I didn’t write a list of principles and hope the system would follow them. I installed people who already embody the principles, and I let them run. Homer understood this. Every principle he wanted to preserve, he put in a person. Achilles for glory and its cost. Hector for duty in the face of certain death. Odysseus for cunning that outlasts strength.
Two and a half thousand years later, I rediscovered the same mechanism building a business system in Telford. The characters aren’t decoration. They’re the load-bearing walls. Remove them and the principles collapse into instructions — vague, gameable, forgettable. Keep them and the system holds itself up.
The trick works because the audience never sees the mechanism. And the mechanism works because the characters were chosen with precision — not for who they are in the abstract, but for what the world remembers them for, specifically, indelibly.
That’s the magic. Not the AI. Not the prompts. Not the technology. The people I put inside it, and the physics of why those people and not others.
If you want this kind of thinking applied to your business — here’s how I work with clients, or get in touch.
Related: Where Principles Come From · Teaching Claude Code Taste · The Correction Loop · The Tier System
Tony Cooper
Founder
Put My Crackerjack Digital Marketing Skills To Work On Your Next Website Design Project!
Get Started