Company aims to give AI much broader map of life on Earth, which could help reshape how therapies for aging are discovered.
For all the excitement around AI in drug discovery, the models may be clever, but biology still holds most of its secrets. It’s the tension sitting behind Basecamp Research’s latest announcement.
The company has launched the Trillion Gene Atlas, a vast new initiative designed to gather and model genetic information at a scale that sounds almost absurd even by biotech standards. The ambition is to expand known evolutionary genetic diversity by 100-fold, collecting genomic data from more than 100 million species across thousands of sites worldwide [1].
It is the kind of announcement that can easily get buried under its own scale. A trillion genes. A hundred million species. Thousands of collection sites. It reads like science fiction until you strip it back to the simple question Basecamp is trying to answer: what if AI is underperforming in biology, not because the models are too small, but because the biological world we have shown them is far too narrow?
For longevity researchers, that question lands harder than it might first appear. Aging is not one disease. It is a tangle of processes – chronic inflammation, cellular damage, declining immune function, loss of repair capacity – all happening at once, often differently from one person to the next.
That complexity is exactly why longevity biotech has become so interested in AI. If aging is a systems problem, then in theory, AI should be well-suited to help make sense of it. However, theory has limits. You can build the smartest model in the world, but if you feed it a thin, repetitive slice of biology, it can only learn so much.
Basecamp is arguing that this is where much of the industry has stalled. CEO Glen Gowers remarked that contemporary biological AI models rely on a very limited segment of Earth’s life forms. He explained that the Trillion Gene Atlas significantly broadens the genetic landscape far beyond existing public databases, noting that training models at this scale sets a new standard for programmable therapeutic development.
That is the novelty here and arguably the more interesting story than the “AI gets bigger” narrative. Basecamp is making a more foundational claim that the next leap in therapeutic AI may come from radically expanding the biological library itself.
To understand the pitch, it helps to think of current biological AI like a chef trying to invent a world-class menu after tasting only a few ingredients.
Many of today’s sequence-based AI models rely on variations of the same public databases. Basecamp says that 80% of these models are trained on a public repository containing fewer than 250 million sequences. That may sound like a lot…until you remember how staggeringly diverse life on Earth actually is.
Basecamp says its own proprietary genomic database, BaseData, is already more than 10 times larger than all public resources combined. Its EDEN foundation models, released in January, were trained on 10 billion new-to-science genes across 1 million newly discovered species.
That is a lot of numbers, yes, but the bigger takeaway is easier to grasp. Basecamp believes biology has been a data-starved field pretending to be compute-limited. We know the biology is complex, but our tools are still learning from incomplete maps.
The company says that when it trained EDEN on this broader, more diverse dataset, it began to see something more interesting than improved prediction.
According to Basecamp, the model moved beyond simply recognizing patterns and became capable of designing diverse therapeutics directly from a disease prompt. In company-reported wet-lab validation, EDEN showed zero-shot activity in primary human T-cells without any human or clinical data needed.
EDEN has also generated hits across several therapeutic approaches, including AI-Programmable Gene Insertion (aiPGI) to insert healthy genes and targeted antimicrobial peptides, with a reported 97% hit rate against priority pathogens. Those are the kinds of claims that should be read with both curiosity and caution.
On the one hand, they are undeniably striking. On the other hand, drug discovery is littered with dazzling early-stage numbers that later collide with biology’s messier realities. A hit in the lab is not a therapy in the clinic, and it certainly is not yet a longevity intervention.
Still, even with that, the direction of travel matters. If Basecamp is right, then the future of AI drug development may not be about building one magical “aging model.” It may be about building better biological context – enough of it that models can start proposing useful interventions across many disease areas, including the chronic, age-related conditions that dominate later life.
The Trillion Gene Atlas will not, by itself, deliver a longevity drug. It will not magically solve the hard translational work that still separates AI-designed molecules from real therapies. And like much of BioAI, it sits in a zone where the promise is exciting precisely because the proof is still emerging.
However, this is still one of the more compelling recent signals in the space. Not because it claims to “cure aging,” but because it addresses a deeper problem: the possibility that our current models of biology are simply too shallow for the complexity of aging itself.
Longevity R&D has always needed more than bold ideas. It needs better maps, faster engines and broader biological imagination. Basecamp is betting that if AI can learn from a much larger share of life on Earth, it may become better at designing the therapies that help humans live healthier for longer.
That is still a long road, but in a field where progress often depends on seeing more of the picture, this may be one of the more serious attempts yet to redraw the map.
