Peter Diamandis asked whether biology will reach a point where "you can literally design an organism." In his version, a person could prompt a system with something like "give me an animal that does this." Ben Lamm, co-founder and chief executive of Colossal Biosciences, gave the idea a name: "it's the CAD software for biology." CAD is the computer-aided design software engineers use to draw machines and buildings before they are made.
Lamm said the technology is not ready. "DNA synthesis isn't quite there yet," he said. Still, he put that world "less than 10 years out." He then stated the forecast more precisely. Within a decade, he said, there will be technologies that can engineer key phenotypes (an organism's visible or measurable traits) in basic organisms across different clades, or branches of the tree of life. The organisms would be built through DNA synthesis or large-scale editing and then grown ex utero, meaning outside a mother's body.
The exchange took place at Moonshots Live 2026, recorded on 25 September 2026 for Diamandis's Moonshots podcast. Colossal uses genetic engineering and AI for species preservation and de-extinction, the effort to bring back lost species or their traits. Lamm's case for his timeline rested on how quickly his company's editing work has scaled.
From 20 edits to 1,000
Most genetic engineering changes one gene or a few. Multiplex editing makes many changes in the same cells at once. Lamm described this as the number of "edits delivered," meaning the changes introduced into cells in a single delivery.
He said that "we were taking victory laps at 20 edits delivered," and that nearly all of biopharma and academia would still treat that as a win today. Colossal, he said, now delivers "300 plus at 90 plus percent efficiency consistently." Then he added: "That was a year ago." The company is now testing 1,000 edits. Early results show "low efficiency, but it's working," he said, and he called it "highly likely" that the company will get it working. These are Lamm's descriptions of internal work. The episode did not include independent confirmation of them.
The other panelists saw a trend line in these numbers. Salim Ismail, founder of Open ExO, called it "an exponential growing curve." Alexander Wissner-Gross, a computer scientist and founder of Reified, asked whether it amounted to tripling year over year and began to extrapolate from there. Lamm stopped him more than once. "That doesn't mean it's going to continue," he said. When Wissner-Gross kept going, Lamm cut in with "No, no, no," and restated only what he would stand behind: "We are consistently north of 300. We're testing 1,000."
Why writing DNA may beat editing it
Ismail asked when the technology reaches the point where "you can do whatever you want." Lamm's answer was that editing may not be the path that gets there. "I think synthesis will replace multiplex editing faster," he said, and he gave synthesis "a higher likelihood of success faster."
The difference is between changing existing DNA and writing it from scratch. Editing makes changes at many separate places in a genome. DNA synthesis builds long stretches of new genetic code chemically, and those stretches can then be swapped in. Lamm said there is a point where "large cargo swaps with DNA synthesis is just better." The obstacle he described is commercial. In his account, the people who make synthesized DNA have not had a business driver to produce it beyond a certain scale, so Colossal "just started doing that internally." Diamandis summed it up as a machine that generates the code you want.
Predicting traits from genes
Dave Blundin, founder of Link Ventures, compared the situation to protein folding. He said that problem "snuck up on everybody" and was solved almost overnight. He asked whether the same could happen with the genotype-to-phenotype problem, which is predicting what a body will look like from its DNA sequence. Could Colossal's growing data set one day let it know the outcome without having to create the animal?
Lamm said the company currently works species by species but is trying to extrapolate to large clades. Body size was his main example. For its thylacine (Tasmanian tiger) project, Colossal's model species is the fat-tailed dunnart, a small marsupial. Lamm described the jump from that "marsupial mouse" to the "marsupial wolf" as a 1,500-fold size difference. Understanding it means looking at what regulates size, he said, "not just the genes, but how and when it regulates in development." Colossal's thylacine program page describes the dunnart as the source of cells, eggs and reproductive systems for the project and notes that thylacines were born at a very immature, roughly rice-grain-sized stage.
Colossal's earlier published work also looks at regulation. In a 2024 announcement, the company described finding regulatory regions, stretches of DNA that control when and where genes switch on, that may explain why thylacines and canids have similar skulls. It said it swapped three thylacine regulatory regions into mice and saw predicted changes in skull development. Its dire-wolf science page takes a comparative approach to size. It highlights LCORL, a gene regulator associated with body size in multiple mammals, and links a dire-wolf variant to changes in the same protein region in large domestic dogs.
Lamm said there are limits to how far size can be pushed. Asked whether you could make a killer whale the size of a pet goldfish, he said "probably not." Scaling could probably work "within certain levels of function," he said, and some groups, such as dogs and certain birds, already vary in size by more than 1,500-fold. He described two separate tracks. The "purist" de-extinction track looks for the specific genes behind an extinct species' traits. The other track generalizes across a clade to see whether size in other species can be shifted by a more modest "20, 50%."
Lamm said some traits will be easier than others. He expects coat color, size and the structures that form skin, scales and feathers to be "highly measurable" and quickly inducible. "I don't know about everything," he added. Colossal has "a whole AI team that's just working on" patterning and stripes, he said. "It's actually a really hard problem."
Eight edits and a woolly mouse
Diamandis mentioned that he had taken his sons to Colossal in Dallas and seen the company's woolly mice, lab mice engineered for mammoth-like hair. He asked how many edits they took. Lamm said the first generation, the one shown publicly, had eight: "eight base edits in one delivery." Base editing changes single letters of DNA without cutting both strands. Colossal's research index lists the woolly-mouse work first as a March 2025 preprint and then as a peer-reviewed paper in Cell Reports Methods on 31 August 2026. The paper is titled "A multiplex genome editing pipeline for rapid combinatorial trait engineering."
Lamm hinted there could be "another version at some point that's more interesting." He also joked that the company missed a chance when William Shatner visited the other day. If the mice had been made tailless, he said, Shatner could have been "very excited about a Tribble," the fluffy creatures from Star Trek. When Diamandis said Colossal could bring Tribbles back, Lamm said "back is the wrong word... We could bring them forward."