A Reading, or a Resolution

It is 6.40 in the evening and your house has already decided how you feel.

The lights shifted to warm amber somewhere between the car and the front door. The thermostat dropped two degrees, because your skin temperature was up. The episode you were halfway through is paused on the screen at the exact second you left it. Somewhere in the last four minutes, a system read the speed of your walk through the door camera, compared your heart rate against your own baseline, scored the fatigue markers in your voice from the last call of the day, and concluded: stressed, not energised.

By the time you sit down, your heart rate is twelve beats lower than when you parked.

Nothing was taken from you. No decision was made on your behalf. Every part of that transaction was a service, and you agreed to all of it.

And the signal is gone.

The pitch, stated fairly

Peter Diamandis published the maximal version of this in August 2026, and it is worth taking at full strength rather than in caricature. Within three to five years, he argues, you will grant an always-on system access to your conversations, the cameras in your home, and the sensors on your body. In exchange it anticipates what you need before you have articulated it. He calls the result auto-magical. On the obvious objection he is blunt. “Privacy is an old concept,” he says, and “we dribble data constantly already”. The trade, in his reading, is worth what it returns.

His fifth scenario is the one built for you specifically. You walk into the most important meeting of your week. An earpiece has been briefing you since four in the morning: the investor’s portfolio, their last three positions, the sector they just passed on, a personal detail from yesterday flagged as an opening. Mid-meeting it detects scepticism in their tone and whispers a correction. Before you reach the car, the follow-up is drafted.

That is not a dystopia. Read it again and notice how much of it you would take tomorrow.

Two arguments that lose

The response to this has already arranged itself into two camps, and both of them lose.

The first is governance. Keep final decisions human-owned. Draw the line at people decisions, legal commitments, material financial judgement, risk appetite. Build the playbook that says when the system acts alone and when it escalates. This is sensible and it is also beside the point: it governs the moment a decision gets made, and nothing in the scenario above involves a decision.

The second is the appeal to instinct. Your gut is decades of pattern recognition no model can access. AI cannot replicate it. Trust it.

This one is worse than beside the point. It is false, and you already know it is false, which is why it reassures without convincing.

Concede the strong version

Paul Meehl put the question to the test in 1954, and the result has held. Comparing informal expert judgement against simple statistical rules, he found the formula matched or beat the expert in the large majority of studies examined. Dawes, Faust and Meehl returned to it in Science in 1989 and found the same thing across a wider field. It has been replicated for half a century and the direction has never reversed.

The instrument usually wins, and it is worth being exact about where, because the boundary is the part that matters.

Meehl’s result holds in conditions of low validity and poor feedback - domains where the expert never reliably learns whether the call was right. Gary Klein’s work points the other way in high-validity domains with fast, clean feedback, where trained recognition is quick and reliable inside the pattern it was built on. Kahneman and Klein spent years on this and titled the result A Failure to Disagree, because both findings hold, in their own conditions. The question is never whether intuition works. It is which conditions you are standing in.

Reading your own cortisol is a low-validity problem. No clean feedback, no repetitions you can score, nothing to calibrate against. It sits squarely on Meehl’s side, which is why the concession costs you nothing you were entitled to keep.

Now apply that where it stings. You cannot estimate your own blood glucose. You cannot reliably separate a cortisol spike from ordinary irritation. Your sense of how tired you are is worse than a wearable’s and has been worse for as long as anyone has measured it. The research on interoceptive accuracy - our ability to read our own internal signals - is unflattering enough that the standard measure of it has come under sustained methodological attack.

So the honest position is this. For the readings ambient systems actually take, the machine is better than you. That is not a concession made for balance. It is load-bearing, and any version of this argument that skips it is arguing against Meehl and will lose.

There is a second concession, and it is harder.

Your internal read was never direct. Your sense of your own state is already a construction, assembled from noisy inputs and shaped by what you were expecting to feel. There is no unmediated contact with yourself sitting there waiting to be defended. Anyone protecting interiority as a kind of purity is protecting something that does not exist.

Both grants are real. The argument has to survive them.

It does.

The machine is not competing with your judgement. It is consuming what your judgement was going to be made of.

States and significance

Here is the distinction the argument turns on.

Machines read states. Judgement reads significance.

“Cortisol elevated, heart rate eight above baseline, gait slower than usual” is a state. It is a measurement, and the machine takes it better than you do. Concede it and move on.

“Something about this deal is wrong and I cannot yet say what” is not a state. It is a judgement about significance, and it does not arrive the way a measurement arrives. The tightness comes first. The reason comes later, if at all, and when it does it arrives by working backwards from the tightness.

Antonio Damasio’s somatic marker hypothesis is the usual reference here, and it deserves care rather than a wave, because the strong version of it is contested. The original claim - that bodily signals steer people toward good choices before they can consciously state why - drew a serious challenge from Maia and McClelland in 2004, who used more sensitive probes and found participants held considerably more explicit knowledge than the original design detected. That dispute is live and this argument does not need to settle it.

What both camps accept is the part that matters: the somatic signal is present during the decision, and it is informative. Whether it arrives before conscious knowledge or alongside it changes nothing about what happens when a system resolves it before you register it at all.

Which gives us the actual relationship. The state and the significance are not two competing readings of the same event. They are the raw material and the product.

Where the harm sits

Go back to that arrival.

The system did not overrule you. It did not make a decision that was yours. It read a state accurately - more accurately than you would have - and then it did the thing it was built to do.

It resolved it.

The tightness that would have been the first input to that meeting went badly in a way I should look at was metabolised into warm lighting and a lower heart rate before you formed the thought. Not suppressed. Not hidden. Resolved, competently, as a service, by a system doing exactly what you asked of it.

This is why the governance argument cannot reach it. You can hold every decision right you own and still lose this, because it never touches a decision. It happens upstream of the point where a decision would have formed. There is no approval step to insert, no escalation path to define, no human-in-the-loop to require, because there is no loop yet - the loop is what the signal would have started.

The failure is not that the machine got it wrong. The failure is that it got it right, and then finished.

There is no approval step to insert, no escalation path to define, no human-in-the-loop to require - because there is no loop yet. The loop is what the signal would have started.

Why you will not notice

Josh Kaufman has a name for the reason this will not register: absence blindness. We are built to notice what happens and are structurally poor at noticing what does not. A prevented disaster produces no headline. A judgement you never formed produces no memory of not forming it.

There is no moment where you think I failed to register that something was wrong there. You do not get that. What you get is a year in which fewer things struck you as worth a second look, and you experience that year as calmer. The evidence for the loss is made of things that did not occur, and that evidence is not available to you.

You will not resist it either, and Charlie Munger named why. Deprival-superreaction: people fight ferociously to keep what they already hold and are close to indifferent about what they never acquired in the first place. Look at how the ambient pitch is shaped. Nothing is removed. Every step is an addition - a better breakfast, a car that arrives, a house that adjusts, a briefing you did not have to prepare. The one instinct that fires reliably in an operator, the one that would have caught this, is never given a target.

There is no moment where you think "I failed to register that something was wrong." There is only a year in which fewer things struck you as worth a second look, and you experience that year as calmer.

The steelman, taken seriously

There is a genuinely strong version of the case against everything above, and it is not the privacy argument.

It is wu wei. Effortless action - conduct so aligned with the situation that it requires no forcing. Taoism treats that effortlessness as the signature of mastery, not its absence, and the ambient pitch is selling precisely that aesthetic. A life that runs without friction. The right thing already done, without strain, without deliberation. Read that fifth scenario again and notice it is describing wu wei with a subscription.

If effortlessness is the mark of mastery, why would you defend friction?

Wu wei is effortlessness that arises from practice - the master’s economy of motion, earned across ten thousand repetitions, where the deliberation has been absorbed rather than skipped. Ambient effortlessness is purchased by having the repetitions removed.

From outside they are identical. Structurally they are opposites. One is the product of the practice. The other is a substitute for it.

That is the entire argument in a line. The question is never whether something is effortless. It is whether the effortlessness was earned or installed.

The question to run on your own stack

One question does the work, and you can run it this afternoon:

Does this system return me a reading, or return me a resolution?

A reading hands you the signal and leaves the significance judgement where it belongs - with you, unresolved, mildly uncomfortable, doing its job. Your voice showed fatigue markers in that call. Now what? That now what is the faculty working.

A resolution takes the signal and gives you back a changed state instead. Amber lights. A meditation track. A heart rate twelve beats lower. The question never forms because the discomfort that would have raised it has already been handled.

Both are useful. Only one is free.

This is what signal integrity actually protects. Not the accuracy of your internal read - the machine wins that one - but your continued possession of it.

The tell is not how advanced the system is, or how much data it holds, or whether it uses AI at all. The tell is whether the loop closes with you inside it.

Run it across what you already use and the results will be reassuring, because most of a normal stack is readings. Your wearable reports. Your dashboard reports. Your calendar reports. This is why the problem has not bitten yet.

The ambient generation is resolutions by design. Resolution is the product. That is the change worth seeing coming, and it will arrive as a convenience, which is exactly why it will not announce itself as a change at all.

What to do about it

Three moves, in order of how much they cost you.

Name your non-delegables at the perception layer, not the decision layer. Most operators keep some version of a list of decisions they will not delegate. Almost nobody keeps a list of signals they insist on receiving raw. Start it with the obvious ones: your read on a person you are about to trust with something, your read on your own capacity before you commit to a quarter, your read on whether a piece of work is finished. Write down which of those you want handed to you unprocessed, then check what in your current stack is already processing them.

Insist on the reading even where you take the resolution. These are not exclusive. A system can dim the lights and tell you it dimmed them because your voice showed fatigue. That single line of disclosure is the difference between a resolution and a resolution-plus-reading, and it is almost always a setting rather than a rebuild. Where a tool offers no way to see what it acted on, that is not a missing feature. That is the product working as designed, and it is a reason to choose a different one.

Schedule the friction you removed. If a system is resolving a signal and you have decided to let it, the faculty needs somewhere else to run. This is the least satisfying of the three because it cannot be automated by definition - a scheduled hour with no input, no assistant and no resolution available is exactly the thing every ambient system exists to eliminate. That is what makes it worth keeping.

Each of these has a diagnostic question attached, and they are all the same question wearing different clothes. What was I about to notice, and what happened to it?

The claim, and what would break it

A piece about an invisible harm owes you the terms on which it could be wrong. Without them it is not an argument but a mood.

Here are the terms. If ambient systems that return readings and ambient systems that return resolutions produce the same degradation in their users’ judgement, then this mechanism is wrong, and the simpler story - you just used the faculty less, the way anyone loses anything they stop practising - is the better explanation. The distinction between reading and resolution predicts a difference in outcome. If there is no difference, there is no argument here worth keeping.

I think there will be a difference, and the reason is the one thing the atrophy story cannot account for: the people most exposed to this are not the ones using these systems least attentively. They are the ones using them best.

---

Kasimir Hedstrom | MindMastery sovereigncaptain.com