🧠 Autism: Musk and I Share Something, and Why RFK Jr. Is Clueless
There is a particular flavor of public commentary that treats an entire neurological category as a single ruined potato. In April 2025, Health and Human Services Secretary Robert F. Kennedy Jr. stood at a briefing on CDC autism numbers and delivered the version that has been quoted ever since: these are kids who will never pay taxes, never hold a job, never play baseball, never write a poem, never go on a date, and many of whom will never use a toilet unassisted. Autism, he added, destroys families and destroys children. He called the rise an epidemic that must be environmental because “genes do not cause epidemics.” After the predictable backlash he clarified that he had meant the profound, low-functioning end of the spectrum, roughly a quarter of diagnoses. The clarification arrived the way most clarifications do — after the blanket sentence had already left the building.
The spectrum is real, and the severe end is not a social-media aesthetic. Some people with autism need lifelong support, do not speak, and will not file a 1040. Pretending otherwise helps no one. What the potato speech does is collapse the rest of the distribution into the same bin and then declare the bin an environmental crime scene. That is the part that fails basic contact with the evidence.
It also fails the calendar. The trait cluster is older than Tylenol and older than the modern vaccine schedule. Villages did not have a code for it. They had ruder words: odd, eccentric, difficult, brilliant-but-impossible, the one who forgets to eat when a problem is open. Leo Kanner’s 1943 paper and Hans Asperger’s writing from the same years were describing something already in the population, not something a twentieth-century molecule had just invented. For decades after that, the same children were filed under childhood schizophrenia or blamed on cold mothers. DSM-III in 1980 split autism off and kept the net tight. DSM-IV in 1994 added Asperger’s and the “not otherwise specified” bucket. DSM-5 in 2013 folded the labels into one spectrum. Each revision caught people who, a generation earlier, would simply have been called odd. Prevalence moved from a few per ten thousand to something closer to 1 in 30-odd children. That is what a wider net does. It is not proof that acetaminophen or a syringe rewired the species. Sibling-controlled studies have repeatedly watched the Tylenol association shrink or vanish once family genetics are held constant. The vaccine hypothesis has been tested for decades and rejected at population scale. Kennedy has kept both stories anyway.
Twin and family studies have been saying the genetic part since the 1970s. Concordance for autism is far higher in identical twins (commonly estimated in the 60–90 percent range across studies) than in fraternal twins. Narrow-sense heritability estimates from large Nordic family data sit around 80 percent. Hundreds of genes are implicated: a mix of rare, often de novo mutations that carry large individual risk, and a much larger cloud of common inherited variants that each nudge probability a little. Environmental contributions exist — prenatal factors, some infections, extreme prematurity — but they do not erase the genetic architecture. Genes do not have to “cause epidemics” in the infectious-disease sense to explain why the same families keep producing the same traits. An epidemic of recognition is not the same object as an epidemic of new toxins.
Some researchers push that one step further and read the milder end as an evolutionary specialization that sometimes overshoots. Baron-Cohen’s systemizing account is the best known version: a strong interest in lawful systems — tools, tracks, mechanisms, rules — and a weaker interest in the noisy social channel. Set the dial higher and you get people who will stare at a register map or a turbopump until it closes. Set it too high, or stack it with other hits, and language may not arrive and daily life needs full-time support. A related line of work notes that many autism-linked alleles overlap with alleles tied to higher intelligence, which looks paradoxical until you read it as imbalance rather than a simple defect. The pop translation is X-Men: a specialization that is useful in small doses and catastrophic when the dose is wrong. It is a hypothesis, not a consensus, and it is not a comic-book origin story. The catastrophic pole is not theoretical. Profound autism exists. Acknowledging the severe cases is not the same as describing the whole spectrum as ruined children.
Which brings us to the inconvenient counter-examples. Elon Musk said on Saturday Night Live in 2021 that he has Asperger’s (the old label now folded into autism spectrum disorder) and joked that he runs human in emulation mode. He has since built or scaled companies that launch orbital rockets, mass-produce electric cars, and operate a global satellite constellation. None of that requires him to be a model of social grace. The public record of awkward press conferences, impulsive posts, and literal-minded replies is long. The point is narrower: the same cognitive style that produces social friction also produces relentless focus on engineering constraints. Rockets are not assembled by people who find small talk effortless and then wander off. They are assembled by people who will stare at a turbopump until the numbers close.
Historical retrospection is weaker evidence and should be labeled as such. Isaac Newton left contemporary descriptions of extreme social withdrawal, forgotten meals, lectures delivered to empty rooms, and an inability to sustain ordinary friendship. Simon Baron-Cohen and others have noted the overlap with what used to be called Asperger syndrome. Newton was never evaluated; mercury poisoning from alchemy and ordinary 17th-century eccentricity remain competing explanations. The same pattern-recognition exercise has been run on other obsessive systematizers — the ones who treated the world as a machine that either worked or failed cleanly enough to iterate. Diagnosis after the fact is not diagnosis. It is pattern recognition, and the pattern keeps showing up in people who moved the technical frontier under the old name, which was simply “odd.”
I sit somewhere on the same map, lower altitude, smaller rockets. At fourteen I wrote my first game in assembly on a TI-99/4A. Not a tutorial port — actual assembly, memory maps, sprite routines that fit because they had to. The machine had no patience for vague intentions. Either the registers held the right values or the screen showed garbage. That constraint felt natural. Social ones did not. Dark humor and sarcasm arrived early and stayed; they are efficient compression algorithms for people who find unfiltered small talk expensive. Odd timing, missed cues, and the occasional remark that lands two standard deviations left of the room are the ongoing maintenance cost. Musk has the larger balance sheet and the orbital hardware. I am funnier. Credit where it is due: the shared circuitry did not prevent either outcome, and it did not turn either of us into a non-taxpaying potato. Neither of us was manufactured by a bottle of acetaminophen.
Kennedy’s list fails in both directions. It erases the people who hold jobs, pay taxes, write code, design hardware, publish papers, or simply manage ordinary adult life with uneven social firmware. It also flattens the genuinely severe cases into a morality play about preventable toxins, which steers research money and parental anxiety toward culprits the better-controlled data have not supported. The cost figures he cites are real for high-support needs; the causal story he attaches to them is not. Declaring genetic inquiry a “dead end” because prevalence rose is an argument that would have discarded every heritable condition whose diagnostic net widened. Diabetes, asthma, and several cancers would like a word.
None of this is a claim that autism is a superpower sticker. The same traits that make assembly language feel like a native dialect make office politics feel like a badly documented API. Sensory load, executive-function taxes, and relationship friction are not optional DLC. Population-level success stories do not cancel individual hardship, and individual hardship does not license a cabinet secretary to describe the category as children who will never do anything an accountant or a baseball coach would recognize. The accurate statement is narrower and less cinematic: autism is a highly heritable neurodevelopmental spectrum; it was present, under ruder names, before the modern pharmacopeia; broader criteria explain most of the counted rise; some scientists read the milder pole as systemizing that overshoots into disability; its severe pole requires serious support; its other poles include people who ship rockets, write low-level code at fourteen, and make dark jokes about both. RFK Jr.’s version skips the data, skips the history, skips the variance, and keeps the potato. That is the clueless part.