Every cell you have ever met came from another cell. That has been the rule for roughly four billion years, which is why "build one from scratch" lived in the same mental drawer as "build a time machine." A lab in Minnesota just opened the drawer. Elsewhere this week: cockroaches got dive certified, a Berkeley chip learned to smell your chicken going off, and a company in Berkeley grew the beginnings of human eggs out of blood. Let's go.
Table of Contents
🧫 A Cell With No Parents and a Full Ingredient List
RESEARCH
Kate Adamala's team at the University of Minnesota has built a thing called SpudCell, named after Sputnik, and it does something no engineered system has managed: it runs an entire cell cycle without ever having been alive in the first place.
The build is a fatty membrane bag containing a 90,000-base-pair genome split across seven separate plasmids, which is smaller than the 113 kbp that biologists had pencilled in as the theoretical floor for a living cell. Inside sits the PURE system, a defined mix of 36 purified enzymes that reads DNA and makes protein. It feeds by fusing with nutrient-carrying liposomes, and it divides without a cytoskeleton at all, just proteins crowding onto the membrane until mechanical stress splits it in two.
Which sets up the exciting bit. Cells making more of that crowding protein divide faster, so the researchers watched their synthetic bags grow, copy their genomes, and outcompete each other across multiple generations. Selection, in chemistry, with no ancestor anywhere in the lineage.
Adamala is refreshingly unwilling to oversell it. She calls SpudCells "obviously not living," and notes that without a cytoskeleton only about 30 percent of daughter cells inherit the full seven-piece genome after five generations. They also cannot build their own ribosomes, which she reckons is the hardest problem left. Also worth knowing: this is a preprint. Cell rejected it after a reviewer objected that SpudCells were "not real biology," which is either a serious methodological critique or the finest pull quote of 2026.
🪳 Cockroaches Have Been Issued Scuba Gear
RESEARCH
Cyborg insects are a real field with a real problem: the insect. Strap a controller to a cockroach, and you get a robot that costs almost no power to move, because the muscles are already there and already work. Then it hits a puddle and drowns.
Engineers at NTU Singapore and Waseda University fixed the puddle. Their diving suit is a 3D-printed tank, a flexible shell sealed to the abdomen, and four silicone tubes plumbed directly into the insect's breathing holes. Manganese dioxide decomposes dilute hydrogen peroxide inside the tank, releasing oxygen and nothing else. Lead author Hirotaka Sato describes it as working "like the oxygen tank used by human divers."
The numbers are stark. Suited roaches stayed active and steerable underwater for up to three hours; an unsuited control walked into the water and stopped moving within 45 seconds. They also ran the insects through plastic tunnels flooded and filled with carbon dioxide to simulate a collapsed, oxygen-poor space, and the suited ones walked the whole thing.
This is not a novelty. Sato's cyborg insects have already been deployed in real search-and-rescue work, including after the magnitude 7.7 Myanmar earthquake in March. Flooded rubble is exactly where human rescuers and conventional robots both fail, and a cockroach fits through a two-centimetre gap. The ethics of conscripting an animal into disaster response remain unresolved. The engineering, unfortunately for anyone hoping otherwise, works.
🥚 The First Human Eggs That Started Out as Blood
NEWS
Conception, a Berkeley company, says it has taken a blood draw, reprogrammed the cells into stem cells, grown those into miniature ovaries, and coaxed out early human egg cells. If that holds up, it is the biggest step yet toward making eggs on demand.
The specific claim is narrow and worth stating precisely: they report primordial follicles built entirely from iPSCs, with early egg cells progressing through meiosis, which they call a world first to their knowledge. Meiosis is the hard part, the chromosome-pairing step where errors turn into failed pregnancies, and seeing the machinery assemble correctly is a real result. Previous human attempts stalled at the oogonial stage, while the mouse version has been producing live pups since 2016.
Now the caveats, which are load-bearing. These are primary oocytes, not mature eggs. Nothing has been fertilized. This was announced in a company blog post and a CEO's social media, not a peer-reviewed paper, and no efficiency numbers, cell-quality data, or independent verification have surfaced.
Conception has raised $38 million and is run by Matt Krisiloff, a former Y Combinator Research director, while the American Society for Reproductive Medicine has said reliable primate safety data is a prerequisite before any human trial. So: extraordinary if true, and currently a press release. Both things at once.
👃 A Chip That Knows Your Chicken Is Done For
RESEARCH
The sniff test is a bad diagnostic that everyone uses anyway. Carla Bassil, a PhD student in Ali Javey's lab at Berkeley, built the replacement: a fingernail-sized chip carrying 16 gas sensors, each with a different sensing film, hitting 92.6% overall prediction accuracy at telling foods apart.
The engineering unlock is material. Electronic noses have existed since the 1980s and kept failing because you cannot easily get sixteen different sensing films onto one chip. Bassil swapped metal oxides for carbon nanotubes, which are sensitive at room temperature, so she could use fragile polymers that would cook on a heated sensor and deposit everything in a single step.
It was trained on seven foods plus chicken, milk, and eggs at fresh, 24 hours out, and 48 hours out. The allergen result is the one that matters: it detected 0.05 grams of walnut, about one hundredth of a shelled walnut, though it has not been tested when other gases are present, which is to say, in a salad. Or a cake. Or a fridge. Also, the portable iPhone-controlled version was not part of the study.
Berkeley claims that it beats human noses, and for a population where foodborne illness drives 128,000 hospitalizations and 3,000 deaths a year, that's enough. Nobody raced it against a dog. Dogs remain undefeated, and smug about it.
🤰 They Operated on Theo Before He Was Born
NEWS
Gastroschisis is a birth defect where the abdominal wall fails to close, and the intestines develop outside the body, floating in amniotic fluid. The complex form means months of neonatal intensive care, repeated surgeries, and sometimes no baby at the end.
Maisie Savage found out at a 20-week scan in London. Great Ormond Street referred her to Houston, and she flew out and stayed there because Texas Children's Hospital is running a first-of-its-kind trial to repair complex gastroschisis before birth, led by Michael Belfort, who previously pioneered in-womb spina bifida repair. Theo became the third baby in the world to go through it, and the first British one.
The procedure is not, despite some framing, non-invasive. In November 2025, at 26 weeks pregnant, surgeons partially exposed Maisie's uterus, drained and replaced the amniotic fluid, and eased Theo's organs back inside his abdomen using keyhole surgery, leaving him in place to finish gestating. He was born in Texas, and at five months he is thriving.
Worth keeping the frame honest, and the surgeons are doing it themselves. Sundeep Keswani, part of the Texas team, describes having the upfront conversation with a mother about something that is not proven. N equals three. A trial with three patients tells you something about feasibility and almost nothing about outcomes. But the direction of travel is clear. The uterus is becoming an operating theatre, and the patient is smaller than a mango.
Reading these together, the throughline is that biology's old membership rules are slowly lapsing. A cell no longer needs a parent. An egg no longer needs an ovary. A cockroach no longer needs to be entirely a cockroach. A patient no longer needs to be born.
The other throughline we see, although less flattering, is how much of this arrived as a press release rather than a paper. SpudCell was rejected and posted as a preprint. Conception's eggs exist in a blog post. Theo's trial has three participants. All of it might be exactly as good as it sounds, and none of it has been through the part where strangers try to break it.
Which of these gave you the strongest reaction? Reply and tell us. If you know somebody who would enjoy the phrase "cockroach diving suit" arriving unannounced in their inbox, forward this along.
Keep questioning everything (especially anything that describes itself as a world first),
P.S. If you are reading this from a lake somewhere, the cockroaches can now join you. They have the gear. They just need a lift.