The 0.92 Mirror
The one number that can help you decide if you should build, buy, or rent your AI.
There’s a story they tell in advertising about a hand-lettered sign outside a farmhouse in Devon. Simple, direct and perfectly logical to the farmer who wrote it.
It read: Clotted Cream and Horse Manure.
To the farmer, this was efficiency. Two items for sale; one sign.
But to the families driving past, it was a stomach-turning disaster. Where he saw products, they saw unappetizing proximity. The image of spreading cream on a scone while thinking about stable waste overwhelmed whatever intentions the farmer had.
The farmer had the context. He knew his dairy was spotless. But he lacked the judgment to know how the world would make sense of his words.
Curiously, there’s a parallel to the mistake most of us make every time we open ChatGPT.
The Mirror
A few weeks back, Anthropic released its fourth Economic Index. Buried in the data was a key number worth knowing: 0.92.
Researchers found a near-perfect correlation between the sophistication of a human’s prompt and the quality of the AI response. Bring high-school depth, get high-school strategy back. Bring PhD-level nuance, meet you there.
(call it the 0.92 Mirror.)
It means that AI, in its generic form, isn’t a tutor that upgrades your thinking. It’s a high-fidelity reflection of your current depth. If you’re a B+ marketer, the mirror gives you B- marketing. Off-the-shelf systems cannot, by their nature, reflect reasoning you don’t already possess.
Risking belaboring the example… AI can turn “Cream and Manure” into “Organic Dairy and Fertilizer Available.” It polishes the vocabulary, but it does not fix the proximity. It obeys the 0.92 law: it reflects your blind spots in high definition.
It’s still a disaster. Just one with better vocabulary.
Now imagine we’re not dealing with a playful parable and instead it’s your 1,500-word business strategy…
Break the mirror
To escape the 0.92 Mirror, we have to stop treating AI as a magic box and start teaching humans the system of inputs that results in this mirror. It’s reflecting the intersection of three distinct parts.
The model is the engine. Claude, GPT-5, Gemini. A vast, high-fidelity surface capable of reflecting almost anything. You can adjust the dials (temperature, reasoning budgets, and more) to change how sharply it reflects. More computation polishes the glass; lower temperature reduces the noise. But these are adjustments to fidelity, not to the underlying physics. Whether the reflection is blurry or 4K sharp, it’s still showing you what you brought to it.
The system prompt is the brain. The methodology. The invisible architect. The set of instructions and judgment patterns that tell the model how to think about a problem.
The user prompt is the context. The specific, messy reality of your business. The fuel for a single task.
The trap
The reason the 0.92 Mirror traps so many professionals is that they’re relying on generic AI.
When you open a standard ChatGPT window, the model is the universal average, and the system prompt itself a vanilla average of the entire internet. Designed to be safe, helpful, and generic. A C-student by design.
This creates a massive gap. To get expert output from that generic mirror, you have to compensate by providing expert-level judgment in your prompt.
The question is: Are you actually an expert in both?
If you’re a founder trying to architect a complex tax strategy using a generic model, you’re asking yourself to be a world-class tax strategist and a world-class observer of your own business.
If you aren’t that tax expert, you’re the farmer.
You have the cream and the manure—the context—but you lack the judgment to know how to present them. You’re writing a sign without knowing the rules of sign-writing. Building a high-fidelity reflection of your own limitations.
The move
The strategic shift is to stop trying to be the architect of every “brain” you use. Separate the labor:
Rent the system prompt. Find the expert proxy, an AI agent where the methodology has been built by someone who actually knows the domain. You’re borrowing their years of pattern recognition. Plugging into a brain that already knows not to put the cream next to the manure.
This frees you to author the user prompt. To be the sovereign of the particular. You focus entirely on the 10% the AI can never know: the specific reality of your situation (context).
When you pair expert methodology with your specific context, the mirror finally breaks. You’re no longer reflecting your own average. You’re reflecting the synthesis of an elite mind applied to your specific truth.
The Choice: Decide if you’re the farmer or the expert…
The 0.92 Mirror surfaces a choice. We have to decide who we are in the room before we decide which tool to pick up.
If you are not the expert, rent the brain and own the context. Stop trying to prompt a generic AI to think like a master craftsman in a language you don’t speak. If you’re a CEO trying to architect technical SEO strategy, you’re the farmer. You lack the judgment to see when the cream is touching the manure. Find an expert proxy—an AI agent already built by a strategist. Provide your unique context, and rent the methodology. You’re no longer limited by your own 0.92 reflection.
If you are the expert, the mirror is your invitation to scale. Build the brain by documenting your judgment. Because you possess the depth, your prompts already reflect deep results. Your job is no longer to do the work, but to build the vessel for it. Capture the tacit patterns and gray-area judgments that make you an A-player and codify them into a digital agent. Build a vessel for your own genius so it can operate without your hands on the keyboard.
When the expert brain meets your specific context, the math flips. The output is no longer a reflection of one person’s limitations. It’s a synthesis of elite judgment applied to specific truth.
The Enterprise Lesson
For the organization, the lesson is starker: Never attempt to “build AI” in a domain where you don’t already have an internal A-player to calibrate it. Without that north star to verify results, you’re simply building a high-speed version of the farmer’s mistake. You’ll spend millions to build a high-fidelity reflection of your own mediocrity.
In the age of the 0.92 Mirror, the master craftsman doesn’t work harder to get average results. They choose their engine, provide their context, and step back to perform the only task that remains: the joy of being the editor-in-chief of their own scaled genius.