The Nanosecond Scientist

A BolesBlogs.com reader forwarded me a transcript and asked whether any of it holds up. A man is talking about human origins. He says Homo sapiens appeared out of nowhere. He says a missing link sits at the base of our species, and that the link is easy to find: hand any person’s DNA to a scientist who works with CRISPR, ask whether the sample has been edited, and the answer arrives in a nanosecond. Yes. Edited. From there he moves to the identity of the editor, God or a visiting intelligence with advanced technology, and he closes by asking why Britain colonized America. Here is the short answer before the long one. Three of the speaker’s sentences are accurate. One sentence is invented, and that invented sentence carries the entire argument. Everything downstream of it, the aliens, Dr. Moreau, the colony, the closing questions, rests on a laboratory result that has never happened and cannot happen. The combination deserves slow handling, because the arrangement recurs everywhere now, and the arrangement does more damage than the content.

The parts that survive contact with the literature

Neanderthals are not our ancestors, and he states that correctly. He also correctly says that mainstream science holds this position. The two lineages separated from a shared ancestral population roughly five hundred thousand years ago, after which one line became Neanderthals across Eurasia and the other became modern humans in Africa.

CRISPR is real, and his one-sentence description of it lands close enough for a layman. A guide RNA carries the Cas9 enzyme to a matching stretch of sequence, Cas9 cuts both strands, and the cell repairs the break. Edits happen inside that repair. His account of taking a gene out and putting a gene in compresses several distinct techniques into one gesture, though the gesture points in the right direction. Third, popular writing about human origins is littered with missing-link talk, and a reader who absorbed his picture of paleoanthropology from magazine headlines would find the framing familiar. He is repeating something he has heard many times, and as far as I can tell he is repeating it in good faith.

Three accuracies function as ballast. A claim with no verifiable content attached gets rejected on contact; wrapped in three checkable statements, the same claim borrows their credibility, and the listener’s guard drops at the moment it should rise.

The Neanderthal sentence costs him the argument

His opening buys him something specific. He needs Homo sapiens to arrive without antecedents, so he wants Neanderthals off the family tree. He gets that. Neanderthals occupy a branch beside ours.

The genetics that removed them from our ancestry also demolished his conclusion in the same stroke. Sequencing archaic bone established that the two populations met and produced children. Most people with ancestry outside Africa carry somewhere between one and two percent Neanderthal DNA today. A 2024 analysis in Science traced nearly all of that gene flow to a single extended episode running from about fifty thousand five hundred to forty-three thousand five hundred years ago, with an average date near forty-seven thousand years. Denisovan ancestry runs higher in some populations, reaching four to six percent among Melanesians.

So we can locate, inside a living person’s genome, identifiable segments inherited from a different hominin species, dated to a specific window, distinguished from shared ancestral sequence by segment length and divergence patterns. That is the level of resolution the field now works at. A man who tells you that human DNA arrived from nowhere is describing a genome that geneticists have already parsed into layers, each layer tagged with a source population and a date range. The rhetorical move here is effective because it recruits real science for the first half of a two-part claim and then abandons that science before the second half arrives. The move is not effective as argument because the same body of work that supports his premise refutes his conclusion, and any reader who follows the citation for one sentence walks straight into the refutation of the next.

A phrase from 1851

The missing link does no work in professional paleoanthropology, and it has not for roughly a century. The phrase predates evolutionary theory. Charles Lyell used the plural form in the third edition of his Elements of Geology in 1851, eight years before Darwin published, describing gaps in the geological record. Darwin himself never used it for his theory. The metaphor descends from the Great Chain of Being, the pre-Darwinian ladder that ranked minerals below plants below animals below man, and a ladder is the only structure where one missing rung breaks the whole climb.

Ian Tattersall of the American Museum of Natural History has described the notion as a holdover from a period when human ancestry was pictured as a single chain running back into deep time, and he points out that the record instead shows many extinct species crowded together, competing for ecological space. John Hawks at Wisconsin calls the term outmoded and says most of his colleagues would like it forgotten. The University of Maryland teaches the concept’s own self-refutation to undergraduates: once you find the specimen, nothing is missing, and the odds against any particular individual entering the fossil record at all are severe.

The drawers are not empty. Fossils from Jebel Irhoud in Morocco, excavated by a team under Jean-Jacques Hublin and Abdelouahed Ben-Ncer, were dated by thermoluminescence to 315,000 years plus or minus 34,000, and they show a mosaic: a face and jaw within the modern range, a braincase still elongated and archaic. Omo I from the Kibish Formation in Ethiopia, found in 1967, was redated in 2022 by Céline Vidal’s team at Cambridge, who matched the geochemical fingerprint of the ash layer above the fossil to an eruption of the Shala volcano and pushed the minimum age to 233,000 years plus or minus 22,000. Herto, also Ethiopian, dates to roughly 160,000. Behind those sit Homo heidelbergensis, Homo erectus, Homo habilis, the australopithecines, a fossil record running back millions of years with the anatomical transitions distributed across it. That is what “out of nowhere” has to mean: five individuals from a Moroccan cave, a skull under volcanic ash from a datable eruption, and a century of collected material, all of it nowhere.

The claim that carries all the weight

Here is the sentence everything rests on. When DNA is altered with CRISPR, the speaker says, a distinctive mark is left behind, so a technician can look at a sample and identify it as edited. No such mark exists. The claim attributes to DNA a property that DNA does not possess. What CRISPR leaves behind follows from its chemistry. Cas9 cuts. The cell repairs the cut, most often through non-homologous end joining, which typically adds or removes a small number of bases at the break. With a donor template supplied, the repair can copy the template’s sequence into the site. Either way, the product is four bases in a row, chemically indistinguishable from every other stretch of the same genome. Adenine deposited by a repair enzyme after an engineered cut looks like adenine that has occupied that position since the Pleistocene, because that is what it is.

Which is why detecting an edit requires a comparison. Every validation method in the working literature is comparative. Enzyme mismatch cleavage using T7 endonuclease I depends on mixing edited and unedited strands so the mismatched duplex can be cut, and the technique throws false positives when it meets ordinary polymorphism. Sequencing works by reading the target region and holding it against a reference or against parental DNA. A 2023 paper in Nucleic Acids Research states the problem with unusual clarity: the authors had to build synonymous diagnostic substitutions into their donor templates specifically because an edit matching the wild-type sequence cannot be distinguished from an unedited one. Researchers install markers on purpose, in advance, because nature supplies none.

The strongest confirmation of this comes from the other direction. When Craig Venter’s institute assembled the synthetic bacterial genome JCVI-syn1.0 in 2010, the team encoded four watermark sequences into the chromosome, each over a thousand base pairs long, spelling out the names of forty-six authors and contributors, a web address, and three quotations, using a custom code built from DNA triplets. The institute’s own language explains why: the watermarks were an essential means of proving that the genome was synthetic and not native, and of identifying the laboratory of origin. Engineers who had just written 1.08 million base pairs from scratch understood that their creation carried no inherent signature, so they wrote one in by hand. The transcript’s whole argument requires the opposite of the fact that motivated the most famous watermarking project in the history of molecular biology.

The nearest approach to a real trace is residual delivery hardware: plasmid backbone, vector sequence, a lingering Cas9 transgene, a selection marker. Those get looked for, and they get found in laboratory organisms. They also have nothing to do with the transcript’s scenario. Ancient germline machinery would have been diluted out of a lineage within a generation or two, and the human genome is already about eight percent endogenous retrovirus, which makes foreign-looking insertions part of the background.

A defender of the transcript has one retreat available here, and closing it off takes a moment. He could say the ancient editors used a technology that leaves no trace at all, so the absence of a mark proves nothing. That retreat costs him everything he came for. His argument is affirmative: he claims a detectable signature exists, claims a technician would find it in a nanosecond, and treats that finding as the evidence. Swap in a traceless method and the evidence evaporates while the conclusion survives, which is the signature of a belief held for reasons other than the ones being offered. An unfalsifiable claim is available for entertainment and useless as a citation. Even the PAM sequence offers nothing. Cas9 needs a short adjacent motif to cut at all, so an edit does sit beside one. That motif is native sequence, present before any intervention, recurring constantly through every genome on the planet. A cotton study using whole-genome sequencing on fourteen edited plants found thousands of variants relative to wild type and attributed most of them to inherent or tissue-culture variation, specifically because they lacked an adjacent PAM. Absence of the motif rules an edit out. Presence rules nothing in.

How editing actually gets caught

The one real case makes the point better than any argument. In November 2018, He Jiankui announced twin girls whose embryos he had edited at CCR5, intending to reproduce the natural CCR5-delta-32 deletion that confers HIV resistance. His edits missed. One twin carried a fifteen-base in-frame deletion on one chromosome and an unedited allele on the other. The second carried a four-base deletion and a single-base insertion. Neither girl received delta-32. Both appear to be mosaics, with edited and unedited cells mixed through their tissues.

None of that was determined by inspecting the DNA for a signature. He and his critics knew the target gene, knew the guide sequence, held the results against the reference CCR5 sequence and against the well-characterized delta-32 variant, and read the differences. Strip that context away and the twins’ CCR5 alleles are three novel indels in an immune gene. Novel indels in immune genes occur constantly. Every person reading this carries dozens of de novo mutations their parents did not have.

Blue eyes, and what the request reveals

The illustration offered in the transcript is a parent ordering blue eyes: pull the brown gene out, drop the blue gene in, get a blue-eyed baby. Here his grasp of the technology becomes measurable. Someone will defend this as harmless simplification for a general audience, and simplification for a general audience is legitimate work that I do myself every week. The test is whether the simplified version points at the truth or away from it. This one points away, and it does so in the direction the argument needs. A single swappable gene per trait makes species-level redesign sound like changing a fuse, and that easy picture is what allows the leap to an ancient engineer casually assembling a new hominin. Accuracy would have cost him the leap.

There is no blue-eye gene to install. The variant doing most of the work, rs12913832, sits in an enhancer inside intron 86 of HERC2, and it does not code for pigment. It regulates how strongly the neighboring OCA2 gene gets transcribed, by controlling a long-range chromatin loop. The ancestral allele permits the loop, OCA2 expression rises, melanin accumulates, eyes read brown. The derived allele weakens the loop, expression falls, melanin drops. Blue irises contain no blue pigment at all. Underpigmented tissue scatters short wavelengths, which puts iris color closer to the physics of sky than to the chemistry of dye. A genome-wide association study of nearly 195,000 people identified fifty additional loci beyond those already known, implicating TYR, TYRP1, SLC45A2, SLC24A5, IRF4, and others. A significant share of individuals self-report an eye color that contradicts their rs12913832 genotype.

His confidence about consequences fails just as badly, and “you can change anything” is effective as rhetoric because it flatters the listener’s sense of a technological present tense that has already arrived. The line is not effective as description because the people running the experiments publish papers saying the opposite. Three research groups reported in 2020 what CRISPR does to human embryos at the target site. Kathy Niakan’s team at the Francis Crick Institute edited POU5F1 in eighteen embryos and found large rearrangements and deletions of thousands of bases in roughly twenty-two percent of them. Dieter Egli’s group at Columbia, attempting to correct a blindness mutation in EYS, watched about half the tested embryos lose large chromosome segments, sometimes the whole chromosome. Shoukhrat Mitalipov’s team at Oregon found comparable damage. Egli’s assessment was that a 2017 result claiming successful correction of a heart-disease mutation may instead have involved the loss of the chromosome carrying it, and his stated conclusion was that the method has no business in a clinic at this stage of human development. The gap between the promise in the transcript and the chromosomal wreckage in the journals is the entire live ethical controversy in the field.

Millions of years

The speaker asks who could have altered human DNA millions of years ago. That timeline destroys his own premise. Homo sapiens is roughly three hundred thousand years old on current evidence. Editing the genome of a hominin two million years back means editing something several species removed from us, which reintroduces the whole ancestral sequence he opened by denying. The physical chemistry finishes the job. DNA degrades. The oldest sequenced genome came from mammoth molars pulled out of Siberian permafrost, dated near 1.2 million years, and the researchers described that as approaching the theoretical limit for the method. Environmental DNA from clay sediments at Kap København in northern Greenland reached about two million years, and it survived because free molecules bound to mineral surfaces, not because they sat inside bone. Against that, the previous record for hominin DNA recovered from sediment stood at roughly 250,000 years, from Denisova Cave.

So the technician in the imagined scene has no ancient sample to compare anything against. He holds a modern cheek swab. Whatever was done to a hominin two million years ago, granting for argument that anything was, would today be ordinary inherited sequence, carrying no residue of the instrument that produced it. The scenario requires forensic access to a crime scene where the evidence decayed a million years before the first sequencer was built.

God, an astronaut, or a novelist

Having established the edit, the speaker offers two candidates for editor: God, or an extraterrestrial with advanced technology. Two options, presented as exhaustive, for an event that has not been shown to have occurred. Logicians call this a false dilemma, and it arrives here in an unusually clean specimen, since the choice is offered before the thing being explained has been established. The intellectual lineage deserves naming, since the transcript presents its account as fresh reasoning. Erich von Däniken sold tens of millions of copies of Chariots of the Gods? from 1968 onward, arguing that ancient monuments imply visiting engineers. Zecharia Sitchin built a longer edifice on misread Sumerian tablets, complete with a species that came here for gold and manufactured workers from local primates. The transcript reproduces Sitchin’s structure with the vocabulary updated: mining becomes colonization, divine engineering becomes CRISPR.

There is a serious version of this family of ideas, and its absence from the transcript is telling. Francis Crick, who co-authored the structure of DNA, published a paper with Leslie Orgel in Icarus in July 1973 proposing directed panspermia, the possibility that organisms were deliberately dispatched to Earth by intelligent beings on another planet. Their abstract concedes that the scientific evidence is inadequate to say anything about the probability, and it points toward the kinds of evidence that might bear on the question. Scientific speculation looks like that. It names what would change its mind.

Then comes the island of Dr. Moreau. H.G. Wells published that novel in 1896, four years before Mendel’s work was rediscovered and fifty-seven years before Watson, Crick, Franklin, and Wilkins gave us the double helix. Moreau does his work with a scalpel and no anesthetic; the book argues about vivisection and the ethics of inflicted pain, written by a man who could not have known what a gene was. Citing it as precedent for genetic engineering means citing a novel about surgery, published before genetics existed, as evidence for a genetic event two million years ago.

The colony question he does not answer

The transcript ends on a series of questions about Britain in America, Spain in California, France and Germany in Vietnam and Africa. To what end did they colonize? Why did countries colonize Earth? Assessed as persuasion, the colonial turn is effective because it moves the listener from a contested claim to an uncontested one and lets the certainty flow backward. Every educated person knows empires extracted wealth. Agreement with that produces the feeling of having agreed with the whole chain. The turn is not effective as evidence because analogy transfers structure and never truth, and the structure being transferred was assembled around a laboratory result that does not exist.

He also never answers his own question, which is the most skillful thing in the passage. A question closes the argument, the listener fills in the answer from general knowledge about extraction and empire, and the listener’s own reasoning becomes the conclusion. Ownership transfers. Nothing has been asserted, so nothing can be corrected. The analogy collapses on its own terms as well. European colonizers went to places already populated and conscripted the people they found; nobody engineered a workforce, because populations existed and ships were expensive. An intelligence capable of crossing interstellar distance has solved energy storage and closed-loop life support at scales we cannot model. Such a civilization needing to breed primates into laborers for gold is a shipping conglomerate crossing an ocean to teach beavers carpentry.

The method, which is the actual subject

The remaining question is what a conversation of this kind accomplishes, sounding serious while it misleads, and the answer concerns me more than any factual error above. Count the techniques in a passage of under four hundred words. “That’s not me saying it. That’s mainstream science” transfers authority to a consensus that holds the opposite view, banking on the listener never checking. “We know this” deploys a plural that includes nobody. CRISPR, splice, gene, altered: technical vocabulary functioning as credential in place of description. Two options offered where none has been earned. A closing question that makes the listener the author of the conclusion.

Then the centerpiece, and as a dramatist I recognize the craft in it, because it is craft. He writes a scene. The setup goes to himself, and the technician gets a line: they look at it and in virtually a nanosecond, they go, yeah, it’s been altered by CRISPR. Then he plays the objection, no, it hasn’t, it’s my DNA, awards the technician the capper, that’s impossible, and takes the tag for himself: exactly. A two-hander with an expert witness who exists in order to agree, timed for the pause and the laugh. In the theatre we call that a scene partner written to lose. The fabricated witness is the whole engine. A false statement can be checked and retired. A staged conversation teaches a method, and the method survives the correction of any particular claim, because it tells the listener that confident delivery plus technical vocabulary plus imagined professional agreement adds up to evidence. Anyone who accepts that equation has been armed against every future correction, since the next voice using the same three ingredients will sound just as authoritative as this one. The specific falsehood is the smaller casualty. What gets damaged is the listener’s instrument for judging the next claim, and that damage outlasts any correction I can publish.

The trade is what galls me. Set against an invented lab result: the geochemical fingerprint of a volcanic eruption, matched across a quarter million years to date the ash lying above a woman’s skull. A pinch of powder drilled from a mammoth’s molar, yielding a genome from 1.2 million years ago. Faces 315,000 years old from a Moroccan cave, modern jaws mounted on archaic braincases. One to two percent of Eurasian ancestry traceable to a population that vanished around forty thousand years ago. Forty-six scientists who spelled their own names into a bacterium because they knew a genome carries no signature unless somebody writes one in. All of that was on the table, and he chose a technician in an imaginary room, giving an imaginary answer, in a nanosecond that never elapsed. The correction stands above, and anyone is welcome to it. Somebody gave away the mammoth’s tooth to keep a made-up lab result, and that is the most revealing fact in the transcript.