What Machines Can't Stake
Machines calculate. Humans commit. The gap is not intelligence — it is what you are willing to wager once the evidence runs out.
A language model never decides. It calculates.
Asked for the next token, it computes a probability over everything it has seen and samples one. That sounds like a small distinction. It is the line between everything machines can do and everything humans are for.
The circle keeps shrinking
Deep Blue beat Kasparov. AlphaGo beat Lee Sedol. GPT-4 passes the bar. Each milestone follows the same script: another domain that “required intuition” turns out to require only enough data and enough compute.
So the temptation is to draw a shrinking circle around what is still ours. Consciousness. Creativity. Empathy. Pick your favorite, watch a model do a passable version of it next year, and shrink the circle again.
We are drawing the wrong shape.
The thing machines cannot do is not a capability. It is a stake: binding yourself to an outcome you cannot verify. Marrying someone. Starting a company. Believing a paper is worth writing before anyone agrees. Voting. Praying. Choosing.
Pattern completion does not bind. It outputs and moves on.
A stake is a wager without the evidence
When you commit to something real, you cannot have the relevant information. You do not know who your partner will be in twenty years. You do not know whether the company will work. You have not yet met the obstacles that will test whether you keep going.
You commit anyway. You treat something as certain, knowing it is not.
William James called this “the will to believe.” Not self-deception — practical necessity. When the evidence is genuinely ambiguous and the choice itself helps determine the outcome, waiting for certainty is choosing never to act. The skeptic who refuses to believe until all the facts are in never takes part in creating the facts.
A model can output a recommendation at 73% confidence. It cannot answer should I marry this person. Not because it lacks data, but because the question presupposes a stake the system does not have. You cannot wager what you do not own.
Creativity begins where certainty ends
Here is the part that surprises people: human creativity does not merely tolerate uncertainty. It lives there.
Einstein published special relativity while the Michelson–Morley result was still contested. Barbara McClintock’s jumping genes were dismissed for thirty years. Every paradigm shift in science is somebody believing the answer before the evidence justifies the belief.
Art works the same way. The novelist does not know how the book ends. The painter does not know whether the next stroke ruins it. They continue, guided by something that is neither logic nor randomness: a cultivated trust in their own judgment that no training data can replace, because the training data does not exist yet.
A diffusion model produces an image by denoising a random seed. It does not struggle. It does not wake at 3 AM convinced the project is worthless and return to it anyway. The machine outputs. The human endures.
Agents compute; they do not wager
I build agents for a living, so this is not abstract for me. It is the question I sit with every time I wire one up.
When you give an LLM “agency,” what you have actually given it is a longer chain of pattern completions. The model picks a next action the way it picks a next token: by sampling from a distribution. Lengthen the horizon, add tools, give it memory, let it reflect, and it gets more capable, sometimes dramatically. But at each step it is still computing what the data would predict. It is not staking anything on the answer.
That is fine for most of what agents are good for: scheduling, retrieval, code generation, research. Domains where the right answer is recoverable from prior cases.
It breaks at the edges where humans actually live. The agent will not start a company. It will not decide, at the cost of its own continuation, that a project is worth doing anyway. It will not refuse a profitable instruction because it violates a principle it cannot prove. It can imitate all of these. It cannot wager.
Scaling will not close that gap. Scaling makes the imitation better. The wager is a different category.
The loop stays open
I should end with a tidy conclusion. That would betray the thesis.
I do not know whether this argument is correct. Maybe future systems will surprise us: some mechanism we have not imagined, indistinguishable from commitment, arising not from better simulation but from a change of substrate we cannot yet see. Maybe the line I am drawing is an artifact of 2026’s architectures, not a boundary.
Or maybe the line is sharper than I have suggested. Maybe there are souls, or sparks, or something else that keeps the gap open no matter what we build.
I do not have the answer. And yet here I am, staking my credibility on a view that may be wrong and sending it into a world that may not agree.
That is not a failure of rigor. It is the move this essay is about. Building cathedrals we will not see finished. Planting trees we will not sit under. Believing before the evidence is enough.
The machine outputs its words and stops. The human wonders whether they were the right ones, and begins again.
If you’re working on agents or care about this question: zguo0525@berkeley.edu · @Zhen4good
— Gavin Guo, May 2026