Cradle

Future ready? In a world of accelerating digital evolution? Let us talk about it... I had the pleasure of joining the 15th Digital Roundtable in Life Sciences, hosted at the Boehringer Ingelheim headquarters and organized by McKinsey & Company As the 15th roundtable of its kind, it yielded an open, critical, but hopeful discussion around the premise: Life Sciences 2030: Becoming future ready in a world of accelerating digital evolution. What I valued most was the density of different experiences - Whether discussing how gender gaps manifest in data and then in models, reviewing how quantum computing might impact what we are building, or exploring how to digital transformation is influenced by where AI is required, exploratory, or simply nice to have. Boehringer Ingelheim ensured that the human perspective came together in one room. AI is not just about speed or efficiency. The goal is still to help scientists make better decisions, design better drugs, and ultimately support patients more effectively. Events like this show that progress in life sciences will come from combining scientific depth, technological ambition, and human judgement. Thank you to Moritz Wolf, Dr. Sophia Ehlers and colleagues for the invitation as well as Nikhil Reddy Podduturi, Daniel Kuhn, Abhishek Pratap, and Jessica J. Federer for the engaging discussions

Hiring more data scientists won't fix your AI problem in R&D. Because the problem usually isn't headcount. It's that the computational team and the biology team have separate definitions of a good outcome, and nobody ever forced the conversation. The ideal team looks something like this: -> BIOLOGISTS who understand enough about what models need to change how they design experiments. -> COMPUTATIONAL SCIENTISTS who understand enough about wet lab reality to know why the data looks the way it does. It really just needs enough mutual literacy to ask each other better questions. One computational lead I spoke to recently said something I keep coming back to: "I spent the first six months trying to improve the model. Then I spent a week in the lab and realised I should have spent those six months improving the data." You can't get there by throwing more bodies at the problem; you might actually achieve the opposite. You get there by building the habit of explaining your world to the person next to you. Two questions worth asking your team this week: → When did a biologist last explain a protocol decision to a data scientist, and why it mattered for the output? → When did a computational scientist last explain a model decision to a biologist, and what it needed from them? If neither has happened in the last month — that's the gap. Not the lack of people.

Case study alert: Scientists using Cradle optimized a CRISPR system for on-target editing activity across three rounds. After a prior screening campaign of >3,000 variants achieved best-in-class performance of <25%, Cradle delivered candidates reaching 50%, 56%, and 68% on-target activity in successive rounds. Full case study: https://lnkd.in/erKRCvdR

71% of biotech AI leaders use AI for protein structure prediction. For generative design, biomarker analysis, and ADME, adoption drops off a cliff. Benchling just published a survey of ~100 biopharma orgs actively using AI. The pattern in their data is the same one I keep running into in customer conversations. The tools that work are the ones with clean, verifiable data behind them. Structure prediction has decades of crystallography to learn from. Literature review (76%) has indexed papers. Scientific reporting (66%) has the document itself. Generative design, biomarker discovery, ADME all run on internal assay data that lives across spreadsheets, ELNs, and scientists' laptops. 90% of the pharma teams I talk to don't collect data with ML in mind. They often can't reliably link a measurement back to the DNA sequence that produced it. The report flags the #1 reason AI pilots fail: data quality. That's an infrastructure and process problem, and most pharma R&D systems weren't built for it. Which is also why the buy-vs-build question keeps coming up. Production AI for molecular engineering needs 40-80 people across ML, software, and lab science to get right. Hiring two ML researchers and spinning up GPUs doesn't get you there. Pharma outsources email to Microsoft and CRM to Salesforce because those are non-core. The 71% is what you get when the data is ready and the right team is building on it. Great report by Benchling - check out their 2026 Biotech AI report.

🧬 Cradle is heading to PEGS Boston next week (May 11–15)! The team and I will be at the Protein & Antibody Engineering Summit in Boston — one of the biggest gatherings of biologics researchers and engineers in the world. If you're attending and want to connect, chat about protein design, or just say hi — reach out! We'd love to meet you there. See you in Boston 👋 #PEGSummit #PEGS2026 #ProteinEngineering #Biologics #Cradle

Excited about joining this panel at SynBioBeta - what questions do YOU want to ask us? "The programmable protein era: how AI rewrites the rules of biomolecules" Wednesday, May 6, 11:30 AM, main stage. Biopharma and industrial biotech rarely sit on the same panel. But the shift we're all seeing is the same one: proteins used to be things you discovered through trial-and-error. Now they're things you design against multiple constraints at once. A few things I want to dig into: → Where AI predictions actually hold up in the wet lab, and where they don't yet → How R&D timelines change when generation stops being the bottleneck → What "programmable" really means when you still need to make and test the molecule → Whether the same approach scales from a clinical antibody to an industrial enzyme If you're at the conference, come find me. If you're working on protein engineering and not on the panel, I'd love to hear what question you'd want us to tackle.

After 25 years in journalism, the last decade as a freelance features writer, I'm excited to announce that on Friday I joined Cradle as content lead. Cradle is an Amsterdam-Zürich startup that leverages machine learning to accelerate protein engineering, enabling faster and more efficient design of proteins with real-world applications in therapeutics, agriculture, and industrial biotechnology. It's work that underpins new vaccines, therapies, sustainable materials and food alternatives that don't require a farm or a factory. It won the MT/Sprout Startup of the Year award in 2024 and was a Bloomberg European Startup To Watch last year. It has raised more than $100M from investors including IVP and Index Ventures. The science is real, the applications are significant and the solutions it offers are to problems worth solving. My job is to build Cradle's content operation: I'll be working on bioengineering content, ML content, executive thought leadership, customer case studies, employer branding, and developing our content engine across text, video, social--you name it. I'm bringing a journalistic approach to what the industry calls "content marketing" (which should really be called marketing content). The distinction matters. One starts with what's true and interesting; the other starts with what serves the brief. A few things I'm looking forward to: colleagues who return emails. Finding reasons to say yes, not enduring that one reason to say no. And—after more than a decade of rates that haven't kept pace with inflation (the one client that gave me a raise since 2013 got the majority of my work)— compensation that reflects my skill level. What I'll miss: Getting paid to travel to some of the world's most remote, dynamic and dramatic ecosystems (I reported from 21 countries on 5 continents). Talking to people at the top of their field about potentially world-changing stuff no one else knows about yet—oh, wait: I'll still be doing that. Thanks to the friends, colleagues, colleagues who became friends and friends who became colleagues who provided support, inspiration and assignments along the way (in no particular order): Brad Wieners Eric Noe Jeremy Keehn Vera Titunik Maer Roshan Susan Murcko David Hochman Hillary Rosner Larry Smith Eddie Alterman Ana Cox Ty A. Tom Clynes Alissa Quart Mike Kessler Nikhil Swaminathan Vince Beiser Hugo Lindgren Sylvia Tan Danielle Stein Chizzik Nichol Nelson Jeff Muskus Brendan Maher John Mecklin Stephan Faris Adam Grant Lisa Kearns Alan Burdick Brendan Borrell Don Hoyt Gorman 💻 Aaron Gell David Rocks David E. Rovella Christopher Mims Sandra Upson Adriane Ohanesian Jim Aley And blessed are the factcheckers! I don't look at this as quitting journalism, but joining two of the fields currently with the greatest potential for impact that improves people's lives - the reason I got into journalism in the first place.

A client asked me recently to be honest with them. "Are we doing something wrong? Because we feel like we're not getting the most out of this." I told them I'd think about it seriously. Then I started writing down what I see in the teams that are getting the most out of AI — not just with us, but in the field more broadly. Six patterns came up. None of them are about which platform you chose. All of them are about how you're set up before, during, and around the evaluation. I'm going to share them one by one over the next few weeks. Starting with the one that surprises most people. The teams that win aren't the ones with the best data. They're the ones where the research department and the AI/Computational team share the same goal. In most organisations I walk into, research is trying to advance a programme. The AI or computational team is trying to demonstrate the value of a new capability. Those are different objectives. They produce different behaviours at every decision point. The research lead wants a hit. The data scientist wants a model that trained. Those can be the same thing, or they can be pulling in opposite directions. The difference is whether someone at the top defined success the same way for both teams — before the pilot started. If they didn't, you already have your answer to "what are we doing wrong."