8 things I learned when the biggest strategy report of 2026 turned to AI x bio

For almost two decades, Amy Webb's annual tech trend report has been something close to a bible for corporate strategists. Fortune 500 boards read it. Governments cite it. SXSW builds a keynote around it.

This year Webb killed the format and replaced it with something bigger: the Convergence Outlook 2026 - a 318-page analysis of what happens when AI, biology, capital, geopolitics, and human behavior collide simultaneously. The report identifies ten convergences reshaping the world.

Two of them - Living Intelligence and Programmable Biology - land squarely on the beat BAIO has been covering since we launched in February.

Yes, a strategy report read by Fortune 500 leadership now treats biological AI not as a research curiosity but as an operating environment that executives need to prepare for. Right here, and right now.

I read the 40-plus pages covering biology and AI. Here are 8 things that stood out to me - and where BAIO's own reporting connects.

The Convergence Outlook identifies ten forces reshaping the global economy. Two - Living Intelligence and Programmable Biology - are about the fusion of AI and biology. Two more - Human Augmentation and Autonomous Care - are adjacent. That's nearly half the report touching the things BAIO covers.

This isn't a niche section buried in an appendix. It's front and center, with the same strategic weight as compute infrastructure and the future of labor.

For BAIO readers: this is validation that the space you've been following is entering the mainstream strategic conversation. The report treats AI × bio the same way it treats AI × energy or AI × robotics: as an inevitable transformation that demands corporate action now.

The report's pharma section is blunt. Under “industries: winners and losers,” it states that winners will be those who build repeatable design-build-test-learn engines fusing computational models with automated labs - and that drug developers stuck on slow, human-paced cycles will lag.

The framing is important: the competitive moat isn't the best AI model. It's the automation, the feedback loop, the throughput.

BAIO readers have watched this shift take shape across multiple issues. The company closest to the report's vision may be Insilico Medicine, which has nominated 28-plus drugs using its AI platform - typically in 12-18 months versus 3-6 years in traditional workflows - and just signed a $2.75 billion deal with Eli Lilly (we covered it in Issue 12). The Prompt-to-Drug framework Insilico published with Lilly (Issue 3) is essentially a blueprint for the design-build-test-learn pipeline the report describes.

Lilly itself is building the infrastructure to run these loops at scale: a $1 billion NVIDIA co-innovation lab with agentic wet labs tightly linked to computational dry labs (Issue 3).

OpenAI connected GPT-5 to Ginkgo's autonomous lab and ran 36,000 closed-loop experiments (Issue 1). Ginkgo then opened that same infrastructure to anyone with a browser at $39 per experiment (Issue 5).

JURA Bio turned manufacturing itself into the generative model, producing roughly 10 quadrillion designed sequences through its variational synthesis process (Issue 10).

Penn Engineering built LIBRIS, a robotic platform producing around 1,000 nanoparticle formulations per hour specifically to feed AI models (Issue 7). And Adaptyv Bio launched a wet-lab API giving AI agents programmatic access to automated protein synthesis and screening (Issue 13).

Not all of these are the fully industrialized flywheel the report envisions - some are early infrastructure, others are closer to production - but the direction is unmistakable.

The report dedicates a full use case to digital twins of real cells, citing Geneformer, TranscriptFormer, Xaira Therapeutics, and NVIDIA's Virtual Cell Challenge by name. It frames virtual cells as a potential breakthrough that could save pharma companies years and billions in failed trials.

BAIO is tracking this race closely. We’ve covered:

☑️ Xaira's X-Cell and its 4.9 billion parameter virtual cell model (Issue 9).

☑️ PerturbGen and Tripso from Mo Lotfollahi's lab - one predicting how perturbations reshape cell trajectories over time, the other decomposing cells into separate biological programs to surface targets that conventional analysis would miss (Issues 7 and 13).

☑️ The Nature Biotechnology review by Bo Wang, James Zou, and Patrick Hsu (among others) mapping where the field stands and where it falls short (Issue 10).

The report's honest caveat matches what we've reported: these models still rely heavily on gene activity data, lack richer inputs like cell imaging, and have limitations. But scientists across the field are convinced the technology will mature rapidly.

Under the heading “What leaders are underestimating,” the report states directly: “Agentic biology is now real. Biology is being wired into closed loops that can sense, decide, and execute.” It calls treating this as “just biotech R&D” a category error.

This maps to a thread BAIO has been pulling across many issues:

☑️ Stanford's Virtual Biotech staffed 37,000 AI agents to analyze 56,000 clinical trials (Issue 4).

☑️ Eubiota designed a gut bacteria therapy from scratch and validated it in mice (Issue 5).

☑️ PantheonOS connected agents to 22 biology foundation models and let them evolve their own algorithms (Issue 5).

☑️ LabOS put AI behind smart glasses to catch scientists' mistakes in real time (Issue 5).

☑️ Markus Buehler's MIT lab released ScienceClaw × Infinite, where autonomous agents conduct scientific investigations without central coordination, selecting from 300-plus composable tools (Issue 9).

Caveat: Some of these have produced real results in real labs. But the gap between a validated proof-of-concept and the continuous industrial-scale agentic biology the report envisions remains significant.

The report gestures at data as a bottleneck across its biology sections - virtual cell models rely too heavily on gene activity data, protein design remains “mostly artisanal” - but doesn't sharpen it into a strategic principle. Covering the field in real time - as BAIO does - makes the pattern plainer. It's not just that data is scarce. It's also that data diversity, not model size, is, perhaps, the bigger binding constraint.

Here’s some of what we reported on:

☑️ Xaira's X-Cell found that adding parameters beyond 1.6 billion didn't help without corresponding increases in data diversity - the gains came from pairing the larger model with 152,000 unique experimental conditions (Issue 9).

☑️ And Xaira’s Bo Wang said it directly at NVIDIA GTC: “the architecture of the data matters as much as the architecture of the model” (Issue 9).

☑️ Basecamp Research launched the Trillion Gene Atlas to expand known genetic diversity 100-fold, arguing that biological AI follows steeper scaling laws with diverse data than with more compute alone (Issue 10).

☑️ And a Renaissance Philanthropy initiative just funded ten projects to fill specific dataset gaps holding the field back (Issue 15).

The report's Autonomous Care section describes a fundamental power shift: patients moving from passive recipients waiting for permission at every step to active consumers assembling their own health systems. The winners, it argues, are delivery networks that act on real-time signals. The losers are providers dependent on episodic encounters and manual coordination.

What makes this shift possible is AI collapsing the cost and access barriers that have kept diagnostics institutional, slow and expensive. BAIO has tracked several examples:

☑️ A deep learning model that estimates a $3,500 genomic test guiding chemotherapy decisions from a routine pathology slide costing under a dollar - published in The Lancet Oncology (Issue 11).

☑️ ProtAIDe-Dx screening for five neurodegenerative diseases simultaneously from a single blood draw (Issue 13).

☑️ Perimeter's Claire giving surgeons real-time margin assessment during breast cancer surgery rather than waiting days for pathology results (Issue 6).

The pattern: AI is turning expensive, slow, specialist-dependent diagnostics into something cheaper, faster, and closer to the patient. That's the infrastructure the report's autonomous care vision requires.

The report's Programmable Biology section includes something BAIO hasn't covered that much: dual-use risk. The same tools that let you engineer bacteria to produce insulin can be retooled to produce toxins, it warns. Gene synthesis equipment fits in a garage and costs less than a car. Knowledge that once required institutional access now circulates in online forums and AI training datasets.

The report calls this out explicitly as a risk that's scaling as fast as the capability. It frames biosecurity not as a theoretical concern but as an operational reality that companies need to bake into their systems now - the same way they treat cybersecurity.

As a dedicated topic, AI + biosecurity probably deserves more space in our coverage. Let us know where you stand on that below.

Buried in the Programmable Biology section you’ll find the phrase “aging as a maintenance schedule.” The report envisions periodic tune-ups where damaged tissue is replaced and cellular age markers are reset. It makes this sound way more straightforward than it is - but the direction is worth paying attention to.

What's genuinely new is that AI and aging biology are starting to converge. This wasn't really a field a year ago. Now we're seeing early indications. Eli Lilly stated at the ARDD 2025 aging conference that longevity is part of its corporate strategy, then signed a $2.75 billion deal with Insilico Medicine - whose founder, Alex Zhavoronkov, has spoken publicly about choosing drug targets that score across multiple hallmarks of aging (Issue 12). Gladstone and NVIDIA built MaxToki, an AI model that predicts how cells age across a full human lifetime and identified pro-aging gene targets validated in mouse hearts (Issue 14). And Insilico's Longevity-LLM aims to replace a decade's worth of specialized aging clocks with a single model (Issue 14).

The strategy world isn't catching up here - it's arriving at roughly the same time as the science. That the Convergence Outlook includes aging as a serious commercial category, not a fringe aspiration, is itself a signal worth noting.

The Convergence Outlook doesn't tell BAIO readers much they haven't already encountered in our 15 issues. That's not a criticism of the report - it's mostly excellent, and its strategic framing is genuinely useful. It's more of an observation: the signals we've been tracking week by week are now being synthesized into boardroom strategy for the world's largest companies.

What the report does add is perspective. It places AI × biology alongside compute infrastructure, geopolitical competition, and labor market transformation as one of the defining forces of the next decade. It maps winners and losers in specific industries.

And it underscores a transition that many in this space have felt for some time: biology is becoming an engineering discipline, and the organizations that treat it like one will define what comes next.

BAIO is free and covers the AI × biology field twice a week. Give it five minutes of your time and you’ll be up to speed.