The Architect Speaks · Episode 478
What AI Cannot Replace in You: The Uncapturable Self and Coherence
This is Episode Four Hundred and Seventy-Eight of The Architect Speaks. Today I want to give you a precise answer to a question the whole culture is currently asking badly, which is what AI can and can’t replace in you, and why the part it can’t reach is smaller, deeper, and more important than almost anyone is saying.
This is one transmission. The Atlas lets you bring your own pattern to the work and see the structure underneath it, free.
Open the AtlasTranscript
This is Episode Four Hundred and Seventy-Eight of The Architect Speaks. Today I want to give you a precise answer to a question the whole culture is currently asking badly, which is what AI can and can’t replace in you, and why the part it can’t reach is smaller, deeper, and more important than almost anyone is saying. Here’s the thing. The question is being answered in two registers right now, and both of them are wrong.
The first is the optimistic one. It says everything will be fine, new jobs will appear that we can’t yet imagine, the way they always have after every technological shift. That’s wrong, and it’s worth understanding exactly why. Every previous transition displaced one category of human work and created another category of human work, and the new category was still recognisably human work.
This transition is different. It’s operating on the substrate of cognition itself, the very thing the displaced workers were using to do the work, and the new category that emerges may not be one humans are positioned to occupy at all. The second register is the apocalyptic one. It says human work is over, meaning is over, we’re finished.
That’s wrong in the other direction. It mistakes the displacement of a category of output for the obsolescence of the entity that produces output. Those are not the same thing. The displacement of the cognitive worker is real.
The obsolescence of the human being is not what’s happening. The honest answer sits between these two, and most of it is structural. Let me give you the structural version. Start with what AI can replace, because it’s only honest to name the size of it.
AI can replace any output that’s fully describable in advance. Any task whose specification can be written down completely enough to be encoded. Any role whose function reduces to a pattern, however complex that pattern is. Any skill whose execution is reproducible.
Any judgment whose criteria can be stated. Any analysis whose components can be itemised. And that, it turns out, is most of the cognitively-rich knowledge work of the last forty years. Legal research.
Medical diagnosis at the level of pattern-matching. Financial analysis. Most software construction. Most of marketing, most of customer service, most administrative work, most of the middle layers of corporate function.
Most of the writing that fills the internet. Most of the design in consumer products. Most translation. Most tutoring.
The category is enormous, and it’s expanding faster than the labour market is adapting. I’m not going to soften that. The displacement is real and the timeline isn’t generous. Now the structural distinction, which is the whole point of this episode.
What AI cannot replace is the thing that has no specification. It cannot replace the entity whose responses can’t be predicted in advance, because they emerge from a ground that’s structurally prior to any response. And I want to be careful here, because this is not a sentimental claim about human creativity, and it’s not a claim about consciousness, which I’m not going to try to settle in a podcast. This is a claim about what AI is, and what you are, structurally.
And it comes down to one sentence. AI is a function. You are a ground. Let me unpack that, because it carries the whole argument.
A function operates on inputs to produce outputs. Even an enormous function. Even one with a trillion parameters trained on the entire textual output of human civilisation. It takes an input, processes it, returns an output.
That output can be brilliant. It can be moving. It can be more useful than anything previously available to the species. And it’s still a function of the input and the weights.
There’s nothing underneath it that the input is arriving at. There’s no there for the input to land in. A ground is different. A ground is the place responses come from before they’re responses to anything specific.
A ground has a history that isn’t retrievable as text. It has a body that’s accumulated knowing it has never put into words and never could. It has a position in time that’s genuinely its own, not the averaged position of every text it’s ever read, its own, singular, located. A ground can be still, and the stillness has a structure that doesn’t need to be expressed in order to be real.
A ground can take in an input and respond from a place that isn’t derivable from the input. That capacity, to respond from somewhere the input can’t reach, is not a small thing. It’s the entire architecture of what can’t be automated. McGilchrist has spent a career on the limits of formal systems, on what the left hemisphere’s models leave out, and this is, in a sense, the economic edge of that argument.
Kingsnorth, in his Machine essays, keeps circling the same intuition, that there’s something the system simply cannot reach. The structural name for that something is the uncapturable self. Now let me bring this down into actual work, because it has hard practical consequences. The worker whose work is mostly the operation of a function on inputs to produce outputs is in trouble, and the trouble is real, because that work is being competed against directly by functions that do it faster, cheaper, more consistently, and without weekends.
That’s most of the modern professional class. And that worker has two structural options. One is to find a niche within their function-work that stays uneconomical to automate, and hold there as long as it lasts. That’s a contraction strategy, and it has a ceiling.
The other is to develop the part of themselves that isn’t function but ground. That’s harder. It’s also the only one with any structural duration. And then there’s the other kind of worker.
The doctor who is present to a patient the way a presence is, not the way a diagnostic system is. The teacher who holds a room. The therapist, the elder, the artist making something from a place that can’t be retrieved, the leader deciding from a coherence earned over decades. That worker is not being competed against in the same way.
AI will assist them. It won’t replace them. That category is small relative to the old economic structure, but it’s real, and the work in it is meaningful, and the premium on it is rising while the premium on function collapses. Here’s the uncomfortable part, and I’d be doing you a disservice to skip it.
Most adults listening to this have not cultivated their ground. They’ve cultivated their function. They invested in skills, credentials, methods, frameworks, expertise, productivity practices. All of that is function.
All of it is what AI now does for the price of a subscription. The cultivation of ground is different work, and most people have deferred it for most of their lives. It’s the work of becoming someone whose responses aren’t derivable from their inputs, who has a position that isn’t borrowed and an orientation that isn’t algorithmic and a coherence earned by going through what the algorithm can’t go through. And it can’t be done by reading more, or taking more courses, or accumulating more frameworks.
It’s done by passing through experience honestly, in a way that integrates rather than processes. By being defeated and continuing. By being broken and reforming. By taking on responsibility nobody made you take.
By being witnessed by people who can see what’s actually present rather than what’s being performed. That’s coherence. And coherence, I’d argue, is the real human advantage in this economy, not as a skill set you can list on a CV but as a property of the organism. It’s what’s left of you that holds together under pressure, that doesn’t fragment when the inputs get hostile, that responds from one centre instead of from whichever fragment the situation happens to activate.
A function has no centre to hold. A ground does. That difference is the whole game. So let me hand you the orientation, depending on where you are.
If you’re in a function-heavy role watching the displacement arrive, the move is not to compete with the function. You’ll lose. The move is to begin cultivating ground, and that will be slow, and it won’t be a course or a hack, it’ll be the patient recovery of the part of you that you deferred while you were building a career out of function. If you’ve always sensed you were operating from somewhere prior to your function, and you’ve spent years trying to make that legible to systems that only valued function, then understand that the next economy is more legible to you than the last one was.
And if you’re somewhere in the middle, the question is just which part of yourself you’re going to invest in, because the next ten years will not reward both. The function part of you is competing against an exponential. The ground part of you is competing only against the part of you that keeps deferring it. What AI cannot replace in you is, in large part, the part of you that you haven’t yet built.
Building it is the only work that’s structurally durable through this transition. This is the first of four episodes mapping work, AI, and worth. We’ll go further in. If anything in this episode made you want to explore what you just heard, I’ve made it easy for you to do so.
In the show notes there is a link to access a book called “Before Approaching the Threshold” which is the gateway to this work. Alongside this you will also receive free 14-day access to The Atlas; an intelligence trained on everything written and recorded, there to think alongside you on whatever you’re actually sitting with. Both are free to start, and the link to access them is in the show notes. This was Michael Lauria and you’re listening to The Architect Speaks.
Show Notes