The Model
The broker had twenty-two years of closed transactions. He could drive past a house and price it within three percent. His read on a neighborhood carried more weight with local banks than their own appraisals. Two decades of contact with the same market, the same negotiation patterns, the same failure modes had compressed into judgment that operated below conscious thought.
I was twenty minutes into the demo when his face changed.
He had spent those decades building something that lived inside his nervous system and nowhere else. Listings priced by instinct, negotiations read by feel, and not a single system to carry any of it beyond him. When he took a vacation, the signal went dark. Training a junior agent meant watching his methodology degrade at every handoff. When a client called at midnight, the answer had one address.
"That's just like Go High Level," he said.
It was not. He had no model for what he was looking at. The gap between what he knew and what his business could carry without him was structural, not a marketing problem or a hiring problem. I watched it repeat for the next two years.
Eight hundred demos across industries and price points. Always the same constraint.
A consulting partner billing four hundred dollars an hour whose practice went silent when she took a week off. She was the signal her firm transmitted, so when the source disappeared, the revenue followed.
A fitness coach in Austin with three hundred thousand followers who spent every morning converting his expertise into sixty-second videos. Nothing stored. Nothing compounded. Tomorrow he started from zero or the channel went dark.
An agency founder in London with forty employees and not a single system that could carry her judgment from brief to deliverable. Four handoffs between her and the client, each degrading the original intent until what shipped barely resembled what she meant.
The signal could not transmit without the source.
M = S × K × E
M is what happens when one person's encoded expertise captures the production layer of a market. Matthew Gallagher built Medvi, a telehealth company, from his house in Los Angeles with $20,000 and a dozen AI tools. Two employees. $401 million in revenue in the first year, on pace for $1.8 billion this year. Serge Gatari encoded his high-ticket sales methodology into TryCook AI and now has 348 founders running his closing system at scale. Eric Simons spent seven years building developer tools, encoded that expertise into Bolt.new, and went from zero to $40 million in annual revenue in five months with 35 people. David Holz built Midjourney to $500 million a year with zero venture capital and $4.7 million in revenue per employee. Maor Shlomo built Base44 alone, no funding, and sold it to Wix for $80 million six months later. These are early returns on a model that compounds.
How the Variables Relate
| Element | Name | Role |
|---|---|---|
| T | Truth | The gate. You pass it or the formula doesn't apply. |
| S | Situational Awareness | Operating variable. Human intuition, timing, reading terrain. |
| K | Knowledge | Operating variable. The signal source. What you actually know. |
| A | Architecture | Phase 1 of E. Designing what gets encoded before encoding it. |
| E | Encoding | Operating variable. The transfer: Architecture → Encoding → Infrastructure. |
| I | Infrastructure | Phase 3 of E. Where encoding lives once it has a home. |
| L | Leverage | The outcome. What compounds when S × K × E is running. |
The Gate and the Signal: Truth and Knowledge
Every signal degrades. Noise is inevitable; fidelity is the battle. But there is a category of degradation that originates at the source, one Shannon never modeled. A well-built channel carrying a corrupted signal produces a convincing lie at scale. I call the accumulated distance between signal and source brand debt, and it compounds silently until the structure fails under its own growth.
At five hundred followers, the divergence is invisible. At five hundred thousand, every new follower is a potential auditor. The internet is a fossil record, not a whiteboard, and AI has made cross-referencing and pattern-matching a thirty-second operation. The base rate of honest signal transmission is dropping while the tools to detect the drop accelerate.
Truth is the gate because it sets direction before anything else runs. You either have signal integrity or you don't — and there is no partial credit. High truth alignment means the source and the transmitted signal converge over time, gaining credibility under scrutiny. Low truth alignment means they diverge, and the divergence accelerates precisely when you can least afford it, at maximum amplification. Truth is the interest rate on the entire model. Every other variable compounds in the direction it sets.
Knowledge is the source material. If the source is empty, there is nothing for the interest rate to compound on.
The Operating Variable: Encoding
Most businesses fail at AI implementation for the same reason most buildings collapse: nobody mapped the structure holding the weight.
Encoding is where one person's expertise crosses into a structure that carries it. It runs in three phases. The first is Architecture: answering the question most operators never ask — what are the final outputs this system exists to produce? Not the processes. The deliverables and client-facing artifacts the business transmits into the market. Without that map, nothing worth encoding gets captured. The second phase is the transfer itself: explicit rules that translate into code and automations, and tacit principles that translate into system instructions. The third phase is Infrastructure: where the encoding lives once it has a home.
The Outcome: Leverage
Leverage is not something you build directly. It is what appears when S × K × E is operating correctly over time. Code and media are permissionless and infinitely replicable. AI flips a constraint that held for the entire history of professional work: one person's expertise can now serve an entire industry at once. Dario Amodei predicted a billion-dollar company run by one or two people would exist by end of 2026. It arrived in April. Two employees. $1.8 billion run rate. That is leverage as outcome — not a variable you set, but a compound that emerges when everything else is running.
The Environment: Situational Awareness
Situational awareness is the discipline of reading how fast the gap between AI capability and industry adoption is narrowing. When capability is high and adoption is low, the window stands wide open. When adoption catches up, positions lock.
Every major researcher who has published on this in the last eighteen months points at the same shift, even where they disagree on timelines and risk: the window is one to three years. Anyone encoding right now is inside it. If you start two years from now, you will face an incumbent with two years of compounding data, encoding, and infrastructure running ahead of them.
S constrains every other variable, because the deepest knowledge and the most precise architecture are wasted if you build for a market that moved while you were planning.
How the Model Operates
The elements are not equal. They occupy three distinct roles.
Truth is the gate. Binary. You either have signal integrity or you do not. If Truth fails, the model does not produce a weak result — it produces a corrupted one. You pass through the gate before the formula applies, or you do not pass at all. The only question that matters: is the signal clean at the source?
S, K, and E are the operating variables. Situational awareness determines when and where you act. Knowledge is the signal source — what you actually know, the depth of pattern built through closed loops. Encoding transfers it into a system that carries it without you, in three phases: Architecture first, then the transfer itself, then Infrastructure. These three are where the work happens.
Leverage is the outcome. Infrastructure is where the encoding lives once it has a home. Leverage is what the compound generates when the feedback loops have been running long enough. You do not build leverage the way you build knowledge or encoding. It emerges from the three variables operating together. What that emergence looks like, and what remains genuinely unpredictable about it, is where the model leads.
The model loops, and each cycle sharpens every variable: K deepens, E refines, S sharpens, and the compound widens with every cycle a competitor has not yet started running.
Time is the variable no amount of capital, talent, or technology can buy back. The line that governs what compounds and what expires is simple: hard to do, or hard to get? AI compresses everything that is hard to do: building software, producing content, running analysis, maintaining integrations. These were moats against the scarcity of intelligence, and intelligence is the one form of scarcity that is ending. What AI does not compress is what takes years to accumulate: proprietary data generated through operations, network density built through adoption, regulatory permission earned through compliance, and the encoded expertise that only exists because someone spent a thousand hours closing feedback loops before AI existed. If your advantage is bottlenecked by intelligence, you are on borrowed time. If it is bottlenecked by years, you are building something that lasts.
Capital, regulation, network effects, competitive response, and black swan events kill real businesses. The model does not pretend otherwise. It maps the variables you can affect while the ones you cannot are in motion, and it maximizes your position regardless of which randomness occurs. When the environment shifts in your favor, you are the only one with the infrastructure to capture it.
Ninety-five percent of businesses report zero value from AI. That failure rate is the model's thesis. If almost everyone fails at AI implementation, the five percent who figure out encoding have a multi-year compounding advantage that widens with every cycle the ninety-five percent are still failing. The failure rate is not the problem. It is the moat.
The model does not claim completeness. You can pass the gate, sharpen all three variables, and still fail because regulation shifted, capital dried up, or a competitor with deeper pockets entered six months later. The model maps what you control. If you score above 30 on every audit in this paper and fail to gain measurable traction within twenty-four months, the model is missing a variable. That is a falsifiable claim, and the standard I hold this work to.