Limits and Methods
We've all had the feeling: in conversation with a deep expert and suddenly you're so deep in the weeds of their expertise you really don't know what's being said or what it means basically you just nod along, buoyed by their enthusiasm, unable to contribute or even assess the truth of what they're saying. Not usually a problem, since people don't generally out-and-out lie; though the cases where that does happen (Theranos) it can become a very bad thing.
But how can we with our limited knowledge know the difference between sublime truth and utter malarkey?
We're about to ask that question at global scale.
The two papers here - presented as a pair because the second hinges upon the first - consider the world we already live in, with relative superintelligence: systems that are smart enough that only a few humans can judge their accuracy.
Every day that pool of humans grows smaller. At some point (and this point will vary from topic to topic) it will vanish completely.
What do we do then? Do we trust these systems, even though we can't understand them enough to put their outputs to the test?
How do we trust those systems?
This isn't the first time we've asked these questions. They go back a long, long way.
It's the question addressed in the first of these papers, "The Limits of Truth: Verification in Relatively Superintelligent Systems".
But trusting the answer is only half the problem. The other half is finding it.
For three and a half centuries, science has relied on a simple, powerful contract: show your work. It said: do this, and you should see what I saw, turning a private claim into a public fact. In the 20th century, the "Methods" section of a scientific paper offered a way to verify the claims being made, turning the paper into an instrument of trust.
That contract is now breaking.
The second paper in this pair, "Methods," asks what happens when the generator of new knowledge (the latent space) can no longer show its work. We are already seeing the first results in mathematics that arrive without a path: counterexamples to century-old conjectures that simply exist, born in a black box.
"Methods" argues that we are witnessing the death of the scientific paper as a bundle of claim and narrative, and the birth of something older and stranger: a loop of oracular discovery and mechanical verification. Opening with a letter from Madras in 1913, it asks what replaces the method when the method is missing. And it offers a surprising answer: the record.
Together, these two papers map the new territory. "Limits" first tells us where the line is. "Methods" tells us what we must build to operate on the far side of it.
The Limits of Truth: Verification in Relatively Superintelligent Systems
Every human being has a discernment horizon: the line past which they can no longer judge whether an answer is right, because checking the work is itself beyond them. As machine intelligence rises, every output crosses more horizons, until outputs arrive that have crossed them all. This paper follows verification past that line. Verification survives the crossing: a proof checker works identically on the output of a mind of any size, and the asymmetry on which the loop economy rests - checking is cheaper than producing - holds at any capability gap. What does not survive is discernment, the capacity to understand what was verified and to judge whether it was worth wanting. The two come apart, and the four human duties on which the post-Watershed framework depends are all duties of discernment. Past the horizon, a relative superintelligence speaks two languages, the formal and the oracular, and neither will be completely understood. The economics of this series reaches its own boundary at the same line: currency requires fungibility, fungibility requires assessment, and what cannot be assessed at receipt is not currency but relic. Humanity has met this asymmetry before and built durable institutions around it; the apparatus of oracle, college, and rite was loop governance under a capability gap that could not be engineered away. The instrument that survives the crossing is the record.
Methods
The methods section is the part of a scientific paper that lets a stranger attempt what its authors did: do this, and you should see what we saw. The newest results in mathematics have no methods section. A counterexample that ended an eighty-seven-year-old conjecture was drawn from the latent space of a language model - the published record holds the object and its certificate, and no path at all - certified afterwards by a proof kernel that was never inside the discovery. The result is certified; the method is missing. This paper argues that the condition is general: post-Watershed, discovery becomes oracular while verification becomes mechanical, and the scientific method reconstitutes as a translation protocol between the two languages identified in the previous paper - forcing whispers from the latent space into the formal channel, or discarding them. The structure is old. Ramanujan and Hardy ran it by hand; Poincaré described its mechanism in 1908. What is new is that the generator scales as far as compute allows and the discriminator at the bottom of the loop has no discretion to corrupt - which matters, because the auditor always could be bribed, and the reproducibility crisis was science discovering that about itself. Vulnerability is not eliminated by the closed loop; it is conserved, and migrates upstream, coming to rest in the question. What replaces the methods section is the record.
These papers are part of the "Verification Design" series; each comes with its own verification record of adversarial audit and changes. I invite your own critiques and comments - which will also be added to the verification record.