D'Arcy Doran

.@JackJohnson’s third act huckmag.com/article/jack-j…

Today at #ICLR, we’re unveiling new scaling laws for DNA foundation models - showing 23% faster scaling on BaseData vs. public datasets. The takeaways: performance follows predictable power laws, efficiency keeps improving with scale, and the emergence of new capabilities.

If nature evolved solutions to survive the impossible… what could it do for us? 🧊 @Basecamp_Res has signed an expanded partnership with the British Antarctic Survey (@BAS_News). For decades, BAS scientists have explored some of the most extreme environments on Earth — from mountain summits to the deep Southern Ocean — building one of the most ambitious biological archives ever assembled. Now we’re bringing that data into the Trillion Gene Atlas: our effort to map life at planetary scale. Every sample is a chapter of nature we haven’t read. Huge thanks to the BAS team — and to everyone at Basecamp helping make this possible. Welcome to Bethany King & Jacob Harmison 🚀 #Antarctica #Genomics #TrillionGeneAtlas

@vitrupo Here's the pre-print in case anyone would like to delve deeper into the work and we'll also be presenting at @icml biorxiv.org/content/10.648…

Medicine is becoming programmable. Basecamp Research CEO Glen Gowers says we can now identify the gene behind a disease, design a protein, and insert a healthy copy. Personalized medicine has been talked about for decades. Now we can program it.

DeepMind stayed in London because it is better for talent than Silicon Valley. "I saw London and the UK as having incredible talent from top universities like Cambridge, Oxford, Imperial and UCL. There is a deep heritage of scientific breakthroughs and world-class thinkers. There was less competition for that talent, which made it a huge structural advantage for building DeepMind." @demishassabis What is the single biggest advantage of building in Europe for you @torsten @antonosika @MaxJunestrand @matiii @ChrisParsonson @cjpedregal @matthewclifford @torstenreil @alanchanguk

📈Surreal to see @Basecamp_Res on the @NYSE after being named one of @FastCompany’s Most Innovative Companies. Most biology hasn’t been discovered — but we act like it has. We’re working to map the unknown 99% with the Trillion Gene Atlas. We're grateful for our incredible team and collaborators. 🚀

AI mastered Go. Then language. Now it’s learning to write DNA. At SXSW, @Basecamp_Res CEO @glen_gowers joined Microsoft’s @AliciaFrame1 and UPenn’s @delafuentelab to unpack what that actually means in practice. The conversation moderated by @sharongoldman explores how models trained on one of the largest evolutionary datasets ever assembled (~1M new species) are enabling a potential “beyond-CRISPR” approach to precise gene insertion—one of the hardest problems in genetic medicine. Watch the full "Decoding Nature: How AI is Learning to Program Biology" panel: youtu.be/TJkRhNJaBMk

🧬@Basecamp_Res has been named by @FastCompany one of the world's Most Innovative Companies — for a second year running. 🏆 The past year has been defining for our team. NVIDIA CEO Jensen Huang announced onstage direct investment in Basecamp Research. We expanded the “tree of life” with 10 billion unique protein sequences and launched the Trillion Gene Atlas — doubling down on a core belief: the future of biological intelligence will be built on better data. Fast Company highlighted EDEN, our foundation model trained on a dataset 10× larger than all public databases combined. By leveraging NVIDIA’s compute, we’ve moved beyond simple DNA edits to AI-programmable gene insertion. In lab tests, this approach designed enzymes to target specific diseases, succeeding on the first attempt across all tested sites. As Fast Company editor-in-chief Brendan Vaughan put it, we are “turning breakthrough innovation into real-world value.” A huge thank you to our partners at NVIDIA, César de la Fuente and the University of Pennsylvania, and our global team for pushing the boundaries of what’s possible. fastcompany.com/91497087/biote…

A new Eye on AI episode asks: could data from 100 million species help cure disease? A startup bets on it as AI advances intersect biotech. Catch the latest insights, including OpenAI’s next moves and more in this briefing by Sharon Goldman. … ift.tt/oTBt1EO

Trillion Gene Atlas Expands Evolutionary Datasets for Next-Generation AI Therapeutics @Basecamp_Res launched the Trillion Gene Atlas to collect #genomicdata from 100 million species to power #AI models for faster, scalable #drugdiscovery #datascience hubs.li/Q047yj1T0

🧬@Basecamp_Res is launching the Trillion Gene Atlas with @AnthropicAI, @nvidia , @PacBio and @UltimaGenomics. Our goal: expand known genetic diversity 100x and collect genomic data from 100M+ species — building foundation models grounded in how evolution actually works. Read the announcement here:

Excited to join the @sxsw panel “Decoding Nature: How AI is Learning to Program Biology” with Glen Gowers, CEO of @Basecamp_Res, @sharon_goldman @FortuneMagazine, & @AliciaFrame1 @Microsoft. As the first academic lab to work with the EDEN model, we helped demonstrate its potential by designing antibiotics from evolutionary data. Excited to see what comes next! #SXSW schedule.sxsw.com/events/PP11625…

@gdb 🧬@basecamp_res will be talking about the science behind this story at #SXSW this afternoon at 2:30, join us for "Decoding Nature: How AI is Learning to Program Biology" schedule.sxsw.com/events/PP11625…

How AI empowered Paul Conyngham to create a custom mRNA vaccine to cure his dog’s cancer when she had only months to live. The first personalized cancer vaccine designed for a dog:

“Over one million new species.” That’s the dataset behind Basecamp Research’s EDEN AI models—mentioned by our Co-founder Oliver Vince on the @BBCr4today. A listener asked the BBC’s More or Less to investigate. Listen to the podcast here: open.spotify.com/show/3sVMOI29n… podcasts.apple.com/gb/podcast/has…

In an incredible collaboration with @Basecamp_Res and @nvidia, today we announce EDEN: an evolution-scale DNA foundation model built on a simple idea—if we dramatically expand the biology we learn from, AI can stop overfitting to a handful of model organisms and start learning principles that truly generalize. Our lab at @Penn helped validate EDEN, which is trained to learn the underlying grammar of biology from vast parts of life we’ve barely sampled. EDEN’s largest model (28B params) was trained on 9.7T nucleotide tokens from BaseData, a dataset enriched for environmental and host-associated metagenomes, phage, and mobile genetic elements — including 10B+ novel genes from 1M+ newly discovered species. One example: using EDEN, we designed new antibiotic peptides with 97% experimental success (32/33 active) — including potent candidates against critical-priority pathogens. More broadly, I’m excited by what this represents: scaling biodiversity changes what models can learn, and it moves us closer to foundation models that don’t just read biology — they help us design within it. Huge thanks to this incredible team: @GeraldeneM49733, Gavin Ayres, @carlagrec, Keith Kam, Gus Minto-Cowcher, John St John, Tanggis Bohnuud, @bakalar_h, @wchowb, Robert Pecoraro, @mdt_torres, Aaron Kollasch, @leungmarcus, @qsirelkhatim, @francescofarin, Connor McGinnis, Srijani Sridhar, Daniel Anderson, @FrancescoOTERI3, Ali Takhabakshi, Jeremie Dona, @TylerShimko, Cedric Stenbeeke, Alexandros Papadopoulos, Malcolm Krolick @JohnsHopkins, @fspoen @OPIGlets @UniofOxford, Purba Gupta, Sandeep Kumar, Anne Bara, Jared Wilbur, @ferruz_noelia @CRGenomica, Timur Rvachov, Fangping Wang, @Hanqun_CAO, Hyun-Su Lee, Japan Mehta, Raphael Chaleil, Valerio Pereno, @potti_siddharth @Stanford, Chris Emerson, Roy Tal Dew, @KevinKaichuang @MSFTResearch @MSRNE, @exnx, @TadimetiNeha, @BanfieldJill @UCBerkeley, Alicia Frame @Azure, @bolton_emma_, @druau, Rory Kelleher, @anthonycosta, @kpowgerade, @glen_gowers, Oliver Vince, Jonathan Finn, & Philipp Lorenz! Link to paper: basecamp-research.com/wp-content/upl…

We are witnessing the start of the transformer moment for biology. @Basecamp_Res has unveiled EDEN, the first foundation model to unlock AI-Programmable Gene Insertion (aiPGI). Trained on 1,008 NVIDIA H200 GPUs using the world's largest ethically sourced evolutionary dataset and accelerated by #BioNeMo, EDEN designs novel Large Serine Recombinases (LSRs) capable of inserting large genetic payloads with high specificity across all tested disease targets. In validation, these AI-designed enzymes also achieved therapeutically relevant levels of CAR integration in primary human T cells and a 97% functional hit rate in antimicrobial peptide design, paving the way for programmable cures in cancer and genetic disease. AI is becoming an engineering discipline for biology. Preprint 👇

Former Tome Bio leader John Finn is continuing his search for the ideal gene insertion tool at Basecamp Research. Today, the startup revealed its first AI-generated enzymes (large serine recombinases) that can insert genes without CRISPR. More in @endpts: endpoints.news/nvidia-partner…

Many people contributed to this massive preprint, and I don't have enough characters in a tweet to give a 📣 to all. But 👏 to @Basecamp_Res and the biodiversity partners, collabs at @nvidia, @MSFTResearch , @delafuentelab, @ferruz_noelia, @KevinKaichuang, @BanfieldJill. ➕👇

A UK-based company has announced the first AI model capable of replacing faulty genes and reprogramming cells to create new medical treatments. Dr Oliver Vince of Basecamp Research says AI is ‘beginning to understand life’ and using that to design medicine.