Why Intellectuals Despair

I was reading a bleak Substack post recently about the Dark Night of Mathematics, which predictably sparked a massive, despairing thread on Hacker News. Highly intelligent people are claiming that AI is robbing pure discovery of its soul. They argue that something ineffable and sacred is being permanently stripped from the act of research. Look closely at what they actually mean by soul, however, and you find something much less mystical. The human brain is essentially a legacy processor, and what intellectuals experience as the "soul" of their work is just the biological heat generated by inefficient, high-friction search algorithms. When a system like FunSearch successfully automating pure mathematics suddenly deprecates that cognitive search, the resulting existential dread is a simple biological withdrawal symptom.

Why does this happen? Doing pure math requires running a massively complex depth-first search on a wetware CPU. Evolution has absolutely no inherent interest in the beauty of prime numbers, combinatorial spaces, or elegant proofs. It simply needs the human organism to finish a severely taxing cognitive search without starving or abandoning the task entirely. To guarantee that outcome, the biological system bribes the brain with chemical rewards. Hard neurobiology demonstrates that dopamine regulates the friction of search, mediating effort-based choice rather than simply evaluating the final prize. Dopamine is the fuel dispensed to overcome the friction of the process itself.

The implication is that intellectuals fundamentally misunderstand their own physiological drives. What they romanticize as the spiritual essence of their labor is literally just transaction-cost dopamine. They have built an entire identity around an inefficient biological subsidy. For decades, the only way to arrive at a novel mathematical truth was to manually grind through a labyrinth of dead ends, maintaining state in a brain that was never optimized for abstract mathematics. The exhaustion and the subsequent breakthrough became intertwined. When an AI outputs the exact same mathematical proof in seconds, the transaction cost plummets to zero. The legacy dopamine system panics because the friction is suddenly gone. The universe remains completely indifferent to this panic, because objective truth is orthogonal to whether the human experiencing it feels like crap.

We have seen this exact biological withdrawal before. In the 1980s, an entire generation of Real Programmers violently rejected C compilers and clung to assembly language. They had spent years memorizing hardware quirks and manually optimizing memory usage. When compilers emerged that could translate high-level logic into machine code faster and more reliably than a human, the veterans did not celebrate the massive leap in productivity. They actively resisted it. They conflated the objective utility of working software with the neurological hit of tracking memory registers in their own heads.

The modern math PhD mourning the intrusion of large language models is acting exactly like an assembly hacker mourning the compiler. Both demographics tied their self-worth to being the CPU. Both mistook the friction of manual discovery for the inherent value of the output. The abstraction stack simply pivoted. In the 1980s it moved from assembly to C, deprecating the friction of memory management. In the 2020s it is moving from manual proofs to AI search, deprecating the friction of cognitive derivation.

Why do smart people fall into this trap so reliably? The brain struggles to separate the goal from the struggle required to reach it. We are conditioned to believe that if a result did not demand immense suffering, it must be inherently less valuable. If human cognitive friction is no longer mathematically necessary to yield a discovered truth, insisting on manual derivation becomes a purely aesthetic choice. It is like insisting on walking barefoot to another continent. You can certainly do it, and you will undoubtedly feel a profound sense of physical and mental exertion along the way, but you are choosing the inefficiency on purpose.

But subjective biological friction is entirely uncoupled from objective reality. The compiler simply does not care if the programmer feels profoundly empty when a memory-allocation bug is fixed automatically. When systems step-function upward in abstraction, they invariably garbage-collect the low-level bullshit that previous generations considered a sacred art. The dread the modern researcher feels right now is just the sudden, jarring realization that they were merely the physical substrate doing the computing. Now that we have better silicon, the wetware is being reassigned.

The inescapable forward motion of technology dictates that you cannot un-compile the world. The historical trajectory of our tools is a one-way vector of increasing abstraction. We build machines to remove the limits of our biology, only to complain when our biology is no longer the limiting factor. The pragmatic response to a sudden drop in transaction costs is never to artificially reintroduce friction to appease a legacy dopamine loop. If you tie your identity to the biological heat of the search, you will be deprecated — an inevitable consequence of confusing the engine's exhaust with its forward momentum.

So if your actual goal is uncovering the truth rather than merely servicing a chemical dependency on effort, the arrival of automated discovery forces you to step decisively up the abstraction stack, relinquish the vanity of manual labor, let the newly built compiler handle the lower-level search space, and begin asking much harder questions. Friction is dead.