I Am Often Wrong
I Am Often Wrong
In January I wrote The Last Levee. In that essay, I said that with prediction solved and delivery nearing solutions, the last barrier to gene editing would be validation. Somebody has to edit a real genome and watch the real organism to learn whether the model was right. The glib and euphemistic way of saying that is “The residuals are the curriculum.” Meaning you have to learn from your mistakes and those “mistakes” turn out to be real living things. Humans. And I said the someone who’d do it first would be a motivated state actor with different ethical constraints, running the sensitivity analysis in vivo, on embryos that can’t consent.
I was wrong.
I was naming the right street but the wrong door.
Wednesday’s (6/4/2026) New York Times: scientists at Columbia base-edited human embryos with unprecedented accuracy. They flipped PCSK9 and HBG at once in the same embryo, and they did it without the chromosomal catastrophe that the old CRISPR used to cause. The work is privately funded, because the federal government won’t touch embryo research. The funder is Nucleus Genomics, the outfit that papered the subway with ads telling New Yorkers to “have your best baby” and describes itself, out loud, in an email to a reporter, as building a “Genetic Optimization stack.”
Imma say that one again, cuz… just wow. “Have your best baby.”
This timeline just gets more and more I dunno… YA dystopian fiction. Amiright?
So my prediction call held. (Sadly.) Base editing was the precision effector I flagged. Mosaicism is what happens when you don’t get it right (ie., the “residual”), and the next round of studies is explicitly about grinding down the errors. A Berkeley geneticist warned that the method hands the “baby improvers” a how-to manual for going past the ethical pale. That is my thesis, read back to me by a man with tenure and more credential acronyms than I’ll ever have.
Owning it, though: I pointed at the wrong vector. I assumed the West would convene its IRBs and self-restrain while somebody abroad did the wet work, and I built a whole geopolitics on top of that assumption. But I’m wrong. The breach is happening here in the West, in English, with fucking venture money and a marketing department. Nobody stormed the levee, they’re just dismantling it one carefully worded press email at a time, and they’d be insulted if you called it anything but inevitable and beautiful “progress.”
(This whole situation is worse, incidentally. A state actor you can sanction. A wellness brand with a Series A and a subway billboard campaign, you cannot.)
This is the thing about being a cross-domain dilettante who points at doors for fun. You get the street right and the address wrong, over and over. The shape-of-things is the part I’m decent at. The exact address requires knowing the actual humans and their actual incentives, and I am guessing while I write a blog in New Orleans. (Who blogs anymore, anyway?)
So I am often wrong, and I will keep being wrong in this way, because the method that lets me see the structure five months early is the same method that blinds me to which specific asshole walks through the door first.
The levee is still ethical, not technical. It still holds only as long as everyone agrees not to test it.
I just didn’t think the test would show up with a series A, a CMO and an ad budget.
Colin Steele is not a geneticist, which is exactly why this essay exists. He endeavors to spot collisions between domains at colinsteele.org.