My new book, The Coming AI Subconscious, is built around a simple claim: the real deliverable isn’t “help.” It’s disappearance.
But to say that cleanly, we need to use the right economic language.
Attention is not a tax. Attention is a premium-priced resource—an executive override, a manual control loop, the most expensive thing you can spend.
The word “tax” belongs to the subconscious.
The Subconscious Tax: Invisible, Real, and Worth It
Your biological subconscious is not free. It has a budget.
While you sleep, you pay a tax for breathing, digestion, thermoregulation, immune scanning, tissue repair, and all the background “keep-me-alive” work you never attend to. You don’t notice it, but it’s real: energy. Hydration. Nutrients. Metabolic load.
That’s what makes it tax-like.
It’s always on.
It’s mostly invisible.
It’s paid by everyone.
And it funds the kind of baseline reliability that allows you to stop thinking about it.
There’s a political analogy here that matters: you pay taxes for national defense so you don’t have to maintain your own private army. The tax is shared, relatively small per person, and absurdly cheaper than doing it alone.
That’s the point: the subconscious is expensive in absolute terms—but cheap compared to the alternative.
Attention Costs 100X More Than the Subconscious Tax
If subconscious tax is the quiet budget that keeps things running, attention is the emergency budget you only want to spend when you must.
Attention requires:
constant monitoring,
interruptions,
context loading,
decision fatigue,
error-checking,
manual approvals,
exception handling.
Attention is not “some effort.” It’s a cognition burn-rate. It steals from your highest-leverage thinking. It fragments your day. It depletes you.
That’s why I treat attention as roughly two orders of magnitude more expensive than the subconscious tax. Not as a lab measurement— as a structural truth about how humans behave.
And that’s also why, the moment a prediction machine becomes reliably “good enough” for a category of work, I stop attending to it.
Not because I’m careless.
Because attention is irrationally expensive.
If the Customer Must Attend, You’re Selling a Copilot
Here’s the test that cuts through all marketing:
If the customer must
check it,
correct it,
re-explain context,
approve steps,
babysit exceptions…
…then you are not selling a subconscious system.
You’re selling a co-pilot.
Co-pilot products are conscious work. They add a new managerial loop. The user becomes a supervisor. The tool becomes an intern. And the “AI magic” becomes an attention sink.
A subconscious product earns the right to vanish. It does the work without requiring the user’s mind to remain in the loop.
The Threshold Isn’t Accuracy. It’s Regret Below Tolerance.
People don’t stop attending because your system hits a particular accuracy percentage.
They stop attending when the expected regret of not watching becomes lower than the attention cost of watching.
That’s the honest threshold.
Low reliability feels like toddler-with-scissors supervision. You can’t look away because the downside is too unpredictable.
Past the threshold, language changes. It’s no longer described as a tool. It’s described as infrastructure. It becomes “just how it’s done.”
That’s subconscious adoption.
Disappearance Doesn’t Mean “Free.” It Means “Cheap Enough to Ignore.”
This is where product teams get confused.
A subconscious system still costs something—just like digestion and breathing still cost something.
But the point is that the subconscious tax is cheap enough, stable enough, and predictable enough that you stop spending attention.
So the product spec isn’t “feature completeness.”
It’s two questions:
- What is the subconscious tax of this system (energy, compute, subscription, maintenance)?
- How quickly does the system reduce required attention to near-zero?
A system can be imperfect and still become subconscious if regret stays below tolerance and the user doesn’t have to babysit it.
The Real Roadmap: Reduce the Attention Surface Area
If disappearance is the deliverable, your roadmap changes.
You stop celebrating “more capabilities” and start measuring “fewer interventions.”
Track things like:
How many times per week does the user need to step in?
How predictable are failure modes?
How recoverable are mistakes after the fact?
How often does the system demand new context for the same work?
How often does the user feel compelled to “just check”?
The winning systems are the ones that steadily convert attention into subconscious tax.
That conversion is the business model.
That conversion is the adoption curve.
That conversion is the coming era.
The Coming AI Subconscious Is an Economic Shift
We are moving from an economy where humans spend attention to produce outcomes…
…to an economy where humans pay a subconscious tax to get outcomes without attending.
And because attention is so costly—because it’s the premium resource—humans will surrender it the moment regret falls below tolerance.
That’s the future I’m pointing at.
Not more tools.
More disappearance.