designer babies – Two developments last week—more precise editing of embryonic DNA and a new argument that AI could begin redesigning itself sooner than institutions can respond—have pushed both medicine and ethics to the edge. The science remains early and untested in ways tha
On June 1. a team of scientists published a preprint paper claiming they had edited human embryonic DNA with more precision than any previous attempt. The work. if it holds up. could eventually make it possible to prevent or reduce inherited disease risk by changing embryo genes more cleanly than before.
But within days, the larger story shifted. Carl Zimmer, writing later that week, said the “real headline news” was that the study “could open the way to babies engineered with particular characteristics,” including what people commonly call “designer children.”
The same day. an entirely different kind of research community made its own case that the future may be arriving faster than governance can follow. An AI company. Anthropic. published a post arguing that AI is already accelerating the development of AI. an early step toward recursive self-improvement—AI systems that design and build their own successors. faster and faster. The company said that most of the code that runs Claude is written by Claude itself and that this approach helped Anthropic’s engineers ship eight times as much code as they were two years ago.
Anthropic’s leadership framed the timeline as urgent. In the post. Anthropic co-founder Jack Clark and the Anthropic Institute head Marina Favaro wrote that self-improving AI “it could come sooner than most institutions are prepared for.” They also emphasized that Claude is still far from being able to guide itself.
Put side by side. the two publications come from separate worlds—academic biologists and an AI company—but they point toward a shared possibility: unprecedented power over both human bodies and human-made intelligence. And both. according to the implications raised by the reporting and the documents themselves. carry a sharp dual edge—potential miracle. potential catastrophe.
The biological advance begins with a narrow technical step. Columbia University geneticist Dieter Egli and his team used so-called base editors rather than CRISPR’s approach of cutting out a segment. Base editors make a small nick in a gene strand rather than chopping the DNA. In the reported experiment, they edited two genes: PCSK9 and HBG.
PCSK9 produces a protein that affects the body’s ability to clear cholesterol from the blood. Certain mutations in the PCSK9 gene can drive LDL cholesterol levels dangerously high, raising the risk of heart disease. HBG encodes a form of hemoglobin the body relies on before birth and normally switches off afterward. Controlling HBG, the researchers say, could reactivate fetal hemoglobin in adulthood, easing—though not curing—sickle cell disease and beta-thalassemia.
The researchers delivered their base editors into fertilized eggs and into two-cell human embryos. In some cases, they reportedly made the edits without the chromosomal damage linked to earlier CRISPR-based embryo editing attempts.
The paper itself—still a preprint and “yet to be peer-reviewed”—presents the achievement as an engineering improvement. not a clinical solution. Edits landed at the wrong spot in the genome in some cases. and relatively few embryos went on to develop normally. The embryos, donated by IVF patients, were developed only to very early stages, and none were implanted.
Egli and his colleagues were explicit in the paper that using base editing in its current form for treatment is “premature.” Still. the work suggests embryo editing might now be done without shredding chromosomes in the way earlier efforts sometimes did. The comparison many readers will remember is from 2018. when Chinese scientist He Jiankui used conventional CRISPR to edit human embryos. producing three children. That work was widely rejected not only for moral reasons but because the gene editing was described as technically clumsy. producing real genetic damage.
If the new results bear out, the technical obstacles to embryo engineering could begin to shrink. The future they point toward is not only about preventing single-gene disorders. Certain genetic disorders like sickle-cell anemia can be fixed with a single gene edit. the document discussion notes. but preventing more complex health problems—or engineering traits some people might want. like height or intelligence—would require editing hundreds or even thousands of genes in combinations that are not fully understood.
If technical barriers continue to fall, that leaves society to confront the moral ones. The possibility described here is not that ethics has no power, but that technologies built for biology—and for intelligence—have a long history of moving ahead even as safeguards lag.
A similar imbalance runs through the AI document. Anthropic’s post. which runs over 5. 000 words and includes plenty of Claude-produced graphics. centers on a single claim: the proportion of human work that goes into building AI is shrinking at every stage. Engineers who once wrote code mostly review what Claude writes. Experiments once designed manually are increasingly proposed and run by the model. While humans still decide what is worth building. the post says even that judgement is shifting. with employees increasingly deferring to what the model proposes to do next.
That is the structure behind the phrase recursive self-improvement—an AI loop dominated by AI itself, accelerating the cycle. The post and its surrounding argument describe intelligence reaching what they call “critical” behavior: each smarter model building a smarter one, sustaining the reaction.
Skepticism is present in the document itself. Anthropic concedes that counting lines of code only goes so far and that speed is only a partial metric of success. The company points to other work. including independent research described in the post. suggesting AI models can spend longer on a single task. which can let them work both quicker and deeper. Dario Amodei. Anthropic CEO. is also referenced for his claims that powerful AI could drive rapid economic. scientific. and medical progress.
But the post does not treat that future as purely beneficial. It warns of existential danger and profound disempowerment for most people—an analogy offered alongside the comparison to genetic enhancement leaving some groups excluded. The company’s proposal goes further than technical caution. Anthropic makes the unusual call for AI companies to consider collectively slowing down or temporarily pausing frontier AI development. so that societal structures and AI alignment research can keep up. The authors cite international regimes used to control dangerous technologies like nuclear weapons. including the International Atomic Energy Agency. which—despite problems—took decades to build and “white-knuckling” to sustain.
In the post, Anthropic leaders argue the window for self-improving AI may be much shorter, quoting: “We don’t have that long.”
The two stories—embryos and intelligence—share a destabilizing pattern. One claims precision and control may improve fast enough to reshape what biology can do. The other claims AI development may become quick enough to outpace the institutions built to govern risk.
There is another common thread too: neither development arrives with a finished safeguard. The embryo paper is careful about its limits and rejects any immediate treatment use as “premature. ” with the work confined to early-stage embryos that are not implanted. The AI post. while laying out the possibility of speed and acceleration. also admits that simple metrics won’t capture the whole picture. even as it pushes for action now.
The human question that follows both publications is simple but hard: how do you know when the world has changed?. Sometimes, as the argument recalls, the implications arrive immediately. Otto Hahn and Fritz Strassmann achieved nuclear fission in December 1938. and experts quickly understood that a nuclear bomb would be possible. Sometimes the scientists see it first and the world doesn’t. When Jennifer Doudna and Emmanuelle Charpentier published the seminal CRISPR paper in 2012. initial press attention was minimal. and the institutions that would later need to govern the technology had not yet recognized what was unfolding.
And then there are the cases where even the experts only see part of it—fission pointing toward a weapon. and those who understood it having little ability to stop it. In this framing. the two advances from last week point in two directions at once: embryo editing that could spare children from fatal disease could also. in time. help enable genetic caste systems; intelligence that could be used to build something like “a country of geniuses in a data center. ” as Amodei put it. could leave society as little more than spectators.
The most urgent line running through these facts is not a conclusion about whether either future will arrive. It is the warning that the door may already be open—and that people may not notice until it is too late.
designer babies gene editing embryonic DNA base editors CRISPR PCSK9 HBG sickle cell disease beta-thalassemia Anthropic Claude recursive self-improvement AI alignment Dario Amodei Jack Clark Marina Favaro