Focused Behaviour Every Day

The Eleven Years I Couldn't Tell Building From Waste

Nobody is where they thought they'd be by now.

I don't mean that as a comfort. I mean it as something close to a law. At some point your brain made a prediction about your life. Where you'd be, what you'd have, who you'd be by this age. Reality came back with different numbers. The distance between the prediction and the numbers is the thing that sits on your chest at night. We read that distance as proof we got it wrong. I've come to think it's the opposite. I think the distance is the brain working exactly as designed. It just doesn't feel like that from the inside. From the inside it only ever feels like being behind.

This is an essay about that distance. About eleven years of misreading it, and then fourteen months of finally understanding what it was. I have a name for the thing that got me through it, FBED, Focused Behaviour Every Day, which I coined in 2013 and then spent more than a decade not understanding. What I want to be straight about, before anything else, is that I am not reporting a finished system to you. I am building it as I write this. The doing is old. I have been doing it for years. The structure, the actual shape of it, the part that explains why it works, I am cutting now, on this page, and you are going to watch me do it. That is not a literary device. It is just what is true, and the essay is more honest if I admit it.

I'm going to lay it out flat, because the shape matters more than any single piece of it.

The decade of summits

I started in a university bedroom, making money before I had any real right to. I won a competition. Sir Terry Matthews. The prize, in effect, was two months living with a billionaire, shadowing his CEOs, watching how the machine actually ran from the inside. There is an irony in that I did not see until much later: I spent two months watching people operate what they had already built, and then spent the next decade refusing to be one of them. Then straight from that into an office of my own. No gap. It was fast. Everything in those early years was fast, right up until the day it wasn't.

And then it slowed. When it slowed, I did the thing I now think most driven people do when the speed drops. I assumed the answer was more.

So I went and got more.

I wanted to be rich. I'll say that plainly, because the rest of this doesn't work if I dress it up. I built a property portfolio. Fifteen houses. I bought the Porsche. Then I sold the houses. I built and grew salon groups. I bought a kitchen manufacturer. I spent a stretch trying, and failing, to become a social media influencer. Each of those was a summit. Each one, while I was climbing it, felt unmistakably like the mountain. The real one. The one that would finally make the numbers match the prediction.

And every time I reached one, I got rid of it.

Not "it failed." I want to be precise, because the precision is the whole point. I removed them. One after another, I climbed to the top of something, looked around, and dismantled it. And in 2022 I did it to almost everything at once. I stripped my life down to nearly nothing. I used to describe that as a decision. I'll come back to it, because it wasn't.

There is one thing I never got rid of.

The agency. It was the first company I ever built, the very first, and it is the company that, today, is ConversationOS. It was there before the houses. Before the Porsche. Before every venture I've just named and several I haven't. Through all of it, every climb and every teardown, the agency stayed.

What that survival actually was is not wisdom, much as it would be easy to dress it up that way. It wasn't some quiet inner certainty that this one was different. For long stretches the agency wasn't spectacular. It wasn't always even good, because I wasn't giving it my best attention, that was wherever the newest summit was. But it never ran on its own. I worked in it the whole time, under focused and half ignored, and there were many moments I wanted to tear it down. Many moments I sat there genuinely unimpressed with it and thought, in plain words: why the fuck am I still doing this.

And every one of those moments, I kept it. That is the thing I want to be precise about. The agency did not survive by neglect or by accident. It survived because, each time the doubt peaked, I made the call to keep it, and then kept working it.

I did not keep it because I believed in it.

I kept it because it was my first company, and killing it would have meant losing. Properly losing. Admitting the thing I'd started as a kid hadn't worked. And I did not want to lose. So it survived on something much less noble than vision. It survived on the exact same instinct that bought the Porsche and chased the next venture. Ego. The refusal to take the loss. But the instinct produced an action, every time, and the action was mine: I chose it, against my own doubt, again and again, for eleven years.

I have to sit with that, because it complicates the story I'd prefer to tell. The instinct I'd point to as the flaw of that whole decade, the more, the not losing, the ego, is the instinct that kept alive the one thing that turned out to matter. I'd like to tell you I outgrew the man who needed to win. But that man is the reason there was anything left to build on at all. The flaw and the engine were the same wiring. They still are. I'll come back to that.

And the whole time, under focused but never dropped, the agency was quietly doing the one thing that would end up mattering. It was building corpus. Logging conversations. Accumulating a substrate I had no idea I was accumulating, because I was too busy looking at mountains to notice the thing under my feet was the actual ground.

That is the part I most want to set down before any of the rest of this makes sense.

For about ten years, I could not tell the difference between building something and wasting my time. They felt identical from the inside. The venture I doubted the most is the one that became everything. If I had trusted my own moment to moment read, if I'd done the focused, sensible, decisive thing and killed the business that wasn't performing, there is no ConversationOS. No 3,500 hours. No question worth asking.

The substrate never tells you it's substrate. It just waits.

The question the corpus asked

At some point the agency stopped being a lead generation business and became something stranger.

Instead of generating leads for clients, we started building them podcasts. Booking guests. And the move that turned out to matter was this: the guest in the chair was the client's ideal customer. The person they most wanted to sell to, sat down for an hour long recorded conversation, talking openly.

We did that twenty thousand times.

Twenty thousand conversations. And somewhere inside that pile, a question surfaced. It didn't arrive as a lightning bolt. There was no moment I can point to and say there, that's when I knew. It condensed slowly out of the corpus, the way a question does when you've been standing next to the evidence for long enough. The question was:

Can trust be derived mathematically?

Not felt. Not intuited. Derived. Could you take a conversation, strip it to its structure, and predict from that structure alone whether trust was present, whether it would hold, whether it would convert into the thing both people actually wanted?

The corpus asked me that. I had been keeping the business that produced it half out of stubbornness, and after twenty thousand conversations it had quietly accumulated enough to ask a question I couldn't ignore.

And the 3,500 hours began.

The eleven years were brute force

Here is what I believed FBED was, for most of the time I'd been doing it.

FBED stands for Focused Behaviour Every Day. I coined it in 2013, and for roughly eleven years I ran it as something close to pure brute force. Daily, daily, daily, daily, daily. If a thing was hard, the answer was to do more of it. If I was stuck, I simply hadn't pushed hard enough yet. The model was almost insultingly simple: more input, more output, repeat until the world gives way.

The mistake in that never felt like a mistake. Not once. Rushing felt like seriousness. Doing more felt like discipline. The ego in it, and there was ego in it, felt like drive. There was never a wiser voice I overrode; the error and the virtue were the same sensation from the inside. I was certain, and the certainty was the error.

What I had no language for was the idea that I was running predictions, constantly, about myself and my work and the world, and that brute force does not fix a miscalibrated prediction. It just runs it faster. It runs the wrong model at higher speed and calls the exhaustion progress.

But the brute force was not wasted. It paid for the corpus. Eleven years of doing, even badly aimed doing, even doing full of ego and rushing, produced a substrate. It built the dataset. It logged the twenty thousand conversations. The decade was not the explosion. The decade laid the ground the explosion would later stand on, and I want to be careful with even that much, because the next move in this essay is the one where I work out, properly, what the relationship between the decade and the explosion actually was. It is not the relationship I assumed.

There is a comfortable lie available to me here, and I can feel it. The lie is that the decade was always going to reach something, that every teardown was necessary, every abandoned venture secretly load-bearing, the decade really a long clever plan I couldn't see yet. That is survivorship reasoning, self-flattery wearing the costume of a confession. The harder version: most substrate is genuinely waste. You do not get to know, while you are laying it down, which kind you are laying down. The agency surviving was not a system working. It was ego, and it was luck, and the luck was load-bearing. I came close, more than once, to killing the one thing that mattered, and I would not have known.

By luck I do not mean anything mystical. My model of the world was miscalibrated for most of that decade; I was running predictions that were wrong. Luck is simply the subset of those wrong predictions where the environment happened to move in the direction my wrong model pointed anyway. The agency did not survive because I read the future correctly. It survived because the future, on this one axis, drifted toward where I had already, by choice, planted myself. That distinction matters for everything that follows, because the system I am about to cut does exactly one thing: over time, it improves your hit rate on prediction error. It does not move the world. A better hit rate on a world that can still go anywhere is the most it can offer, and the rest was luck.

Here is the closest I came.

At the end of 2021, I was merging the kitchen manufacturer into a larger business. The deal was advanced enough that the other side had already begun stripping the factory team out, ahead of completion. And then the merger collapsed.

So I was standing in a factory with no team in it. An empty building, an empty floor, and a deal that no longer existed. I want to describe the thought I had accurately, because it is the centre of this essay. It was not a brave thought. It was not "this is hard but I will push through." It was a genuine question, and it was the first time I had ever really asked it: I can't do this. Can I.

And here is the part I would most like to skip. A reasonable person, looking at me in that factory, would have told me to stop. Not stop the kitchen company. Stop. The whole pattern. The climbing and the tearing down. And they would not have been stupid to say it. The doubt I felt was not weakness misreading the situation. The doubt was correct-shaped. It was pointed at something real.

I did not have a process for that moment, because in 2021 I did not understand FBED as anything more than brute force, and brute force had nothing to offer an empty room. There was no rep to do. You cannot out-work a building with no one in it.

So I did something I did not plan and did not, at the time, understand. I got in the car and I drove to Nice, and I stayed for a month. Every day I rode up the Col de l'Ecre, through Vence, around Monaco. It would be easy to dress that up and easy to dismiss it. From the outside it looked like a man running away to the South of France. From the inside it did not feel like recovery. It felt like a man who could not be still, depleting himself up a mountain every day because stillness was unbearable.

I did not know, then, what that month was. I want to be exact about that, because I am only working it out now, writing this. At the time it had no name and no shape, it was just where I went when the factory broke me. But the doing was mine. I drove there. I rode the mountain. And I was the one who came back and, for the first time, pointed everything at the agency on purpose, rather than keeping it half-alive on stubbornness while my attention was somewhere else. Whatever that month was, I ran it. I just could not have told you what I was running.

What it was, and this is the part I can only see from here, with the model I am about to build, is the thing the rest of this essay is about. But I will not name it yet, because I had not named it yet. In the order this actually happened, the factory came first, then Nice, then a slow return, and the naming did not arrive for another three years. Let me keep the order honest and get to the naming when I earned it.

A few months later came 2022. I have described that year before as a decision, and then as the aftershock of the factory. Both undersell it, and I can see now why. It was not passive, aftershock is something that happens to you, and 2022 was an action. The month in Nice had emptied me out and quietly changed where my focus pointed. 2022 was the first thing I built on the new direction. I stripped my life down to almost nothing not as damage and not as decision, but as the first move of a recalibrated man. I did not know that was what it was. I am telling you that is what it was, now, because I finally have the structure to say so.

What FBED actually is

So here is where I stop narrating and start cutting. Everything above happened. The factory, Nice, 2022, the twenty thousand conversations, those are events, and I was there, and they are not in question. What is in question, what I am working out as I write this sentence, is the shape underneath them. The thing I have been doing since 2013 and never once named.

For eleven years I thought FBED was one thing: the doing. The relentless daily behaviour. And it is that, but the doing is only one component, and treating it as the whole system is the precise mistake that cost me the decade. So let me take it apart properly. I think there are three components, and they are not three equal phases on a flat list. They are an engine, a governor, and a detonator. The hierarchy is the whole insight, so I want to build it in that order.

The engine is the build. This is FBED as I always understood it, relentless path carving, focused behaviour every day, the doing. The repetition, the questioning, the frustration. And I mean it when I put frustration inside the definition, because the frustration is not the system failing. It is a mechanism hardening, the friction of a pattern that has not yet gone automatic being forced toward automatic. The build is what took me, in months, from barely understanding edge functions to writing large Tokio resolvers fluently. Fluency cannot be read into existence or shortcut. It is purely a function of reps survived, and the friction is the felt sensation of substrate forming. The engine is real, it is necessary, and for eleven years it was the only part I respected. That was the error. Not running the engine. Thinking the engine was the car.

The governor is the 70% rule. Here is the thing I got most wrong, and I only see it now because I can finally compare the decade to the fourteen months. I always assumed the build should be run at 100%. Total intensity. No slack. Anything less felt like weakness. And here is what I can now show, with the corpus and with my own last fourteen months as evidence: 100% is bad news. Not because it breaks you, that is the weak version of this claim and I want to reject it. 100% does not primarily fail by exhaustion. It fails by convergence.

Run the engine at 100% and it still works. It still accrues substrate. It still produces competence, real competence, you genuinely get good. But it runs with no consolidation capacity. There is no spare compute for the system to step back and read what it has built, because every cycle is consumed by the build itself. So what 100% gives you is a closed loop. You get better and better at the parameter levels you already occupy. You hill-climb, smoothly, with the felt experience of mastery the entire way up, and the smoother and faster you climb, the more certain you are that you are winning. But you are converging. You are being walked to a local maximum, and unconscious competence is exactly the trap, because from inside the loop a local maximum feels identical to the summit. There is no signal that tells you the larger mountain exists. The breaks, when they come, come slowly, because closed-loop competence is stable, it is a deep basin, and the better you get the deeper it gets, and the deeper it gets the harder it is for anything to dislodge you. Mastery and entrapment turn out to be the same measurement.

That is what eleven years at 100% bought me. Genuine skill. A converged man. Substrate accruing underneath the whole time, and no phase in which to read it, so from the inside, building and waste were indistinguishable, because closed-loop competence feels exactly like both. You are working hard, you are getting better, and you are going nowhere new. That is the most precise description of that decade I have ever managed, and I am reaching it right now, in this paragraph.

So the governor, 70%, is not a recovery rule and not a stamina rule. The 30% you hold back is not rest and it is not waste. It is the compute the next component needs to run at all. You are not throttling the engine to protect yourself. You are throttling it to fund the detonator. That is the whole function of the number.

The detonator is recalibration. This is the component I missed entirely for eleven years, and it is the only one that produces the explosion. I want to be exact, because I used to think the build itself eventually pays out, eleven years of fuse, then the blast. That metaphor is wrong, and correcting it is most of what this essay is for. The build never explodes. The build accrues. The build, run alone at any intensity, only ever hill-climbs. The explosion is a different event in a different phase: it is recalibration, and what it does is jump you off the local maximum to parameter levels the closed loop could not have reached no matter how hard or how long it ran.

Recalibration is what happens when the engine is run down to genuine depletion and there is finally compute free to read the substrate instead of adding to it. Building deposits substrate in the dark. Recalibration is when you develop the photograph. It cannot happen while the engine is at 100%, because at 100% there is no free compute, the reading capacity does not exist, regardless of how much you might want it to. This is the part to keep mechanical: the fix is not to be more mindful or more reflective. You cannot think your way to a recalibration. You have to fund it, structurally, by holding 30% of the engine back. The governor and the detonator are one mechanism described from two ends.

And here is the test that keeps this falsifiable, because a detonator I cannot tell from avoidance is not a detonator. Recalibration has an impostor, and the impostor is just stepping away because the work got hard. They look identical from the outside and nearly identical from the inside. The test that separates them is not a feeling and it is not retrospective. It is a single question you can answer before you step away, before any result is in: was the engine depleted, or was it merely uncomfortable. Depleted means you ran it to empty, there is genuinely nothing left to deposit, and stepping back is the only move on the board. Uncomfortable means it got hard and you wanted out. That distinction is checkable in advance, and it is the part you should trust. There is a second signal, weaker, that only arrives later: false recalibration is fluent, it produces frameworks easily, and they feel profound, and they predict nothing. Real recalibration is slow and resists you, because you are reading a substrate that genuinely exists rather than generating one that flatters you. Treat that as corroboration, not proof. When I drove to Nice I did not feel clever. I felt emptied. But I only know that now, and a feeling I can confirm only afterwards is not something to build a decision on. The decision rests on the antecedent. Was the engine depleted before I stepped away. It was. The factory had run it to nothing.

So now I can say the thing I deferred earlier. The month in Nice was the first time the detonator fired. I did not choose it and I did not understand it, the factory depleted the engine to zero, and a recalibration ran whether I had a name for it or not. That is why I came back pointed somewhere new. And it explains the shape of everything: the detonator had fired exactly once, by accident, in eleven years, because the engine was almost never allowed to deplete. I was running 100% with no governor, which meant the detonator had no compute, which meant the loop stayed closed. Eleven years of accrual, one accidental detonation, no idea any of it was a system.

Validation is the fuel line. One more component, and it is not a phase, it is a maintenance function, the thing that keeps the engine from seizing. The brain runs on dopamine and a sense of progression, and FBED honestly run withholds both: the substrate is silent, it pays nothing for a very long time. So you have to engineer the dopamine. Not eliminate it, and not let it run high either, because high validation is the cheap hit and the cheap hit is its own trap. You dose it. Call it five percent. Enough to let the brain survive a process that otherwise offers no confirmation at all. I think of the balance as something like seventy percent build, twenty-five percent recalibration, five percent validation, and I do not defend those numbers as measured. They are a description of a balance, not a result. If they are wrong they are wrong toward too little validation, because that is the error my wiring prefers.

Once the hierarchy is laid out like that, engine, governor, detonator, fuel line, the mistakes of the decade finally have names. Rushing and context switching feel like productivity because they carry real effort and real discomfort. You are working. You are uncomfortable. It genuinely feels like building. But they keep the engine pinned near 100%: they consume all the compute, they never let depletion complete, so the detonator never fires and the substrate never gets read back into structure. A context switcher and a 100% grinder have the same disease from opposite directions, both run the engine with no governor, both stay in a closed loop, both can accrue substrate for a decade and never once convert it. That was me, in both modes, for eleven years. One engine running flat out, no governor, a detonator that fired exactly once and only because a collapsing merger pulled the trigger for me.

And then the fourteen months. When I finally introduced the governor and let the detonator run on purpose, the last fourteen months happened. In them I built neural networks. I built production grade application services. I helped institutionalise an investment advisory function. I did 3,500 hours of genuine work. I took myself to the standard of high grade research, published and created new tools inside AI research, learned to code from scratch, and now run large applications across four services by myself. Here is the part that matters as evidence, not as a list. The corpus did not change. The substrate did not change. The engine was not running harder than it ran in the decade, if anything it ran at lower intensity, because I was finally governing it. What changed is that the loop opened. The fourteen months are not eleven years of fuse finally reaching the charge. They are a closed loop becoming an open one, and the proof is in the shape of the progress: a closed loop produces smooth improvement that plateaus, and an open loop produces rougher improvement that occasionally jumps. The fourteen months are jumps. That is the signature. That is what tells me the model I have just built is describing something real and not just a story I find flattering.

I need to stop on that last point, because it is the biggest claim in this essay and I have hedged every smaller one. The clean version of what I just said is: the substrate was there all along, the eleven years laid it, and the fourteen months simply read it. That is a model of what happened, not a proven fact, and conveniently, the model that most flatters the decade. The honest alternative is that the fourteen months are where the real work actually happened, and the eleven years before were mostly genuine waste that I am now retro-fitting with a corpus story. I cannot fully rule that out. I cannot rerun the decade without the agency and the twenty thousand conversations to see what the fourteen months would have produced from nothing. The corpus is real. The conversations are real. The jumps are real. But "a closed loop opening" is an interpretation laid over those facts, and it is the one I would most want to be true. I think it is substantially right, the closed-loop-versus-open-loop shape predicts the plateau-versus-jump signature, and the signature is what I actually observe, which is more than story usually gives you. But I am telling you plainly that I cannot prove it, and that you should weigh it knowing which way my wiring leans.

The night it became visible

Last week was hard, and I'll come to why. But Friday night is where I want to take you.

It was late. The house was dark. Ada and Penelope were asleep. I was sitting downstairs with the specific kind of tiredness that has nothing to do with sleep, the kind you get when you've spent a week being wrong about things.

I had two options, and both were unglamorous. One was to scroll until I felt numb enough to go to bed. The other was to open the work. Not the work I owed anyone. The other kind. The recalibration. The stuff with no deadline, no audience, no guarantee it goes anywhere.

I almost scrolled. There was nothing noble in that room. I was depleted, it was dark, and the easy thing was right there.

I opened the work instead. And a few hours later I broke into territory I had been circling for months. Something I genuinely did not know if I would ever reach.

I'm not going to tell you what it was. That's not the point of this essay, and the point holds without it.

The point is what happened a few weeks earlier, on another night like it, when I sat down and wrote the algorithm for tokenising behaviour. I built the rubric myself, by hand, no agent: categorisation by brain type, by psychometric, by supercluster, cluster, entropy as the weighting mechanism.

What shocked me was not the result. It was the hands.

I was writing it there and then, and my hands were moving at a speed I had never watched them move. Hard terms, deep ones, the kind of language that normally costs me real effort to hold, and I was not holding them, I was connecting them, fluently, more than a thousand behavioural tokens taking shape under my fingers and cohering as they landed. I was not working it out. I was watching it happen. I had become an audience to my own hands.

I sat up. I actually, physically sat up.

Then, almost as an afterthought, I pointed the thing at six sealed conversations and asked it to call the outcomes from the tokens alone. Six out of six. And six is not proof, a number that small, chance can produce it, and if the night had been a verification test it would have been a weak one. But it was not a verification test. The six was the full stop, not the sentence. The sentence was the thousand tokens, and the fluency, and the speed.

That is what made me sit up: not a score, a capability I had no memory of acquiring, moving fast enough that I could finally see it. It had finished forming some unknown amount of time before that night, silently, and the experiment was the first time it surfaced. That is what an open loop looks like from the inside, not gradual, not announced, just a jump that was already complete by the time you noticed it. Deep substrate. From FBED. From eleven years of corpus I had spent most of building convinced I was wasting my time.

That is what confidence actually is. I used to think confidence was something you build directly, like a muscle. It isn't. Confidence is what a fired detonator looks like from the inside. It is the lights coming on in a room you furnished in the dark. You don't build confidence. You build substrate, blind, for a very long time, and confidence is the moment the loop opens and you finally get to see what you built.

What last week actually was

I said last week was hard. Here is the mechanism of it, because it is the most useful thing in this essay.

I had new senior people coming in. A capital raise underway with institutions. Four production grade applications live. New corporate language to absorb, and underneath all of it a slow, heavy realisation that the way I had been running the business was not the way it now needed to be run. Financial gaps between where I thought things were and where they actually were. Things not finishing on time. The whole week, more or less, was one continuous prediction error.

And it cascaded, because I made the specific mistake I now understand best and still cannot fully stop making. I tried to hold two models at once. I was still running the engine for the old scale of the business, while trying to think at the new scale. Old weights and new weights, fighting for the same compute, neither one able to settle. That is not depletion. That is a jam. Its frustration is not the clean friction of hardening, it is the grinding of two processes trying to occupy the same space.

Then someone said the thing I needed and did not want to hear. They told me that what I was doing was, in their words, insane, but that I did not need it right now. That I had built the assets already. That I did not need to keep pushing into new territory every week. That the job now was to operate what existed.

That landed as a prediction error of a different kind. Not "my model for this scale is wrong." Worse: "I am running the engine into new territory when the territory is already taken, and the work now is something else." The empty factory had pointed at this once already, four years earlier, and I had not quite heard it. It took until last week, and someone willing to say the word insane to my face, for the lesson to land.

And here is the part that shows the system is real rather than a nice idea. I did not solve that by thinking harder about it. I solved it by doing what FBED, properly understood, says to do. Friday night, depleted, low, I stopped running the engine. I governed it down to nothing and let the detonator run instead. I broke into that new territory, and woke up on Saturday with a clear head and something I had not had on Thursday.

Not a solution. A permission. I could suddenly say, calmly, you were right, I do not need to be making something new every week, I need to operate what I have built. The business problem had not changed overnight. My ability to hold it had. The detonator had fired and jumped me somewhere I could see the whole thing from. That is what it does, when it is real: it does not answer the question you are stuck on, it moves you far enough off the local maximum that you can come back and hold the question differently.

I should not pretend it always fires clean. The Friday night I am describing was real recalibration, and I know that because of what preceded it, I had been emptied by the week, the engine was at zero, there was genuinely nothing left to build with. If I had gone to that desk merely frustrated rather than depleted, the impostor would have run instead: I would have produced something fluent that felt like a breakthrough and recalibrated nothing, and found that out days later when the supposed clarity failed to predict anything. That has happened. It will happen again. The system is not magic. It is a process honest enough to tell you when you have run it wrong, if you are honest enough to look.

The honest ending

I will not tie this off too neatly, because the truth is not neat.

I am not an optimal builder. I am nowhere near it. The engine includes a lifetime of development, new boundaries on every run, and the large majority of my self prediction errors. The brain is, in the end, a neural network. It makes predictions, some fire, some don't, some worked in the environment of 2015 and silently stopped working in 2016. New environment, new parameters, weights need recalibrating. That is not a problem to be solved once. It is a relationship you have to build and keep building, because most of my time is spent running the engine, and most of my time therefore contains prediction error. I will tell you plainly: I do not like it. The brain does not want it. The brain wants order, the brain wants the closed loop, because the closed loop feels like mastery. FBED is the daily discipline of refusing the closed loop on purpose, of holding back the 30% that funds the jump, when every instinct says run flat out.

And the flaw and the engine are still the same wiring. The ego that kept the agency alive out of a refusal to lose is the same ego that had me, last week, pinning the engine at 100% when the job had changed. I have not defeated that man. I have learned, slowly, to govern him instead of being run by him. That is the most honest version of progress I can offer. Not a fix. A relationship.

Right now, as I write this, I am sitting with a gaping prediction error in the way I run my business. I know what it is. I know it needs resolving. And I know exactly what the next move is, because I have just spent this essay working out the system I have been running blind since 2013.

The next move is to build. Not new territory this time, that is the closed loop's pull, and I can feel it, and I am naming it so I can refuse it. The build now is the unfamiliar one: operating what already exists, becoming the thing I watched those CEOs be and spent a decade refusing to become. Operating is not the absence of the engine. It is the engine pointed at a rep I have barely done. It will have its own friction, its own hardening, its own depletion. It is build all the same, just build that finally climbs the mountain I have been standing next to.

I am glad about all of it, and I say that last and only last, because it should not be said early or cheaply. The eleven years of brute force, the summits I climbed and tore down, the Porsche, the houses, the decade I could not tell building from waste. Not because it was efficient, and not because I would recommend it. I am glad because, on the most likely reading, the engine running flat out for eleven years laid the corpus, and the corpus is the floor everything now stands on. A closed loop still accrues. That is the one mercy of running it wrong: the substrate is real even when you cannot read it.

But that gladness is not a promise, and it is easy to misread it as one. The decade was not redeemed because the system guarantees a payout. It does not. Standing in that empty factory, running the engine longer was not obviously the right call. Stopping might have been. I do not get to tell you that if you keep going it works out, because I do not know that, and the factory is my proof that I do not.

Here is the most I can honestly offer, and it is the thing I now understand that I did not understand on Thursday. Nobody is where they thought they'd be by now. The distance between the prediction and the reality is not proof you failed. It is the brain doing the one thing it is built to do, recalibrating, in an environment that will not hold still. The system does not promise that distance closes. It promises something smaller and truer: run the engine, but govern it, because the engine alone only ever climbs the hill you are already on, and the 30% you hold back is the only thing that ever lets you see there was a mountain. You do not run it because it always works. You run it because it is the only way to act honestly while the answer is still unknown.

That is what I have. It is not a guarantee. It is a way to keep moving without one.

Better should win.