Pվ��Ƶ

Takeaways from SciFoo 2025

As my team will attest, I remain fascinated with how artificial intelligence is reshaping the way we conduct science, make discoveries, and translate them into progress. The last year has offered striking proof of this transformation. The 2024 Nobel Prize in Chemistry recognised the breakthrough contributions of Demis Hassabis, John Jumper, and David Baker for —ending a decades-old grand challenge in biology that will accelerate drug discovery (and published in Nature, no less). In Physics, John Hopfield and Geoffrey Hinton were honored for —foundations that underpin the AI systems we now use every day. Beyond these milestones, we also see AI accelerating discovery in additional areas such as materials science with DeepMind’s , in climate forecasting via , and even in astronomy, where NASA applies .

These examples underscore a pivotal truth: AI is not an adjunct to science—it is the accelerator of scientific progress. With this perspective, I travelled to the Bay Area and joined SciFoo 2025, an extraordinary unconference co-hosted by Digital Science, Google, O’Reilly, and Nature. Across conversations with scientists, technologists, and thought leaders, five themes stood out to me on how AI is reshaping discovery.

Publishing for Machines

The traditional unit of science has been paper. But as AI agents increasingly become both readers and authors of research, we must ask: Is a paper enough? Perhaps the true unit of science is the experiment, shared in modular, reproducible formats. Tools like already enable scientists to present methods, code, and data in executable form, driving reproducibility and open science. Projects like , led by Andrew White, are pushing this vision further—designing AI agents that automate aspects of discovery in biology, from hypothesis generation to experiment planning. Is rethinking publishing for machines, not just humans, the next great inflection point in science? And if it is, what are the implications for publishers?

Decentralised Intelligence

A highlight of SciFoo was hearing from Prof. Ramesh Raskar at MIT about . NANDA explores a future of decentralised intelligence—an “Internet of AI agents” that collaborate across trustless networks. Imagine scientific experiments, peer review, and replication shared not through central servers, but across distributed systems where provenance, transparency, and speed are built into the architecture itself. This could democratize access to scientific infrastructure, reduce bottlenecks, and accelerate progress by orders of magnitude. It’s a continuation of a lineage that runs deep: Tim Berners-Lee created the World Wide Web (WWW) at CERN in the early 1990s to help physicists share research papers more easily. Later, Google’s PageRank algorithm was inspired by the way citations in scientific literature measure influence, turning scholarly referencing into the foundation for web search. In the same spirit, initiatives like project NANDA could drive the next great leap in science, innovation, and discovery through decentralised intelligence.

Democratising Science Communication

Science matters only when it is understood. Communicating directly to taxpayers—the ultimate funders of much scientific research—on issues like climate science, the energy transition, or drug discovery is vital. Today’s AI tools can make this communication both accessible and compelling. Google’s Veo 3 enables high-quality video generation, while Adobe Firefly allows creative, on-demand visuals. Together, they give researchers new ways to turn complex findings into infographics, explainers, and short-form content.

Just as importantly, we must meet younger audiences on platforms of their choice. Gen Z learns through TikTok, Instagram Reels, and YouTube Shorts. If we want to educate and inspire the next generation, we publishers must be ready to translate science into their formats and languages, without losing nuance. AI can help scale this democratisation of science communication—bringing discoveries to millions in ways that inform, inspire, and build trust.

The Future of Books

Books have always been humanity’s vessels of knowledge. They will continue to educate, entertain, and inspire us—whether in print, digital, audio, video, or even AR formats. Personally, I still enjoy the quiet, distraction-free experience of a print book. Yet consumption habits are changing. Pվ��Ƶ’s Books + Courses are one example of how content is becoming multimodal.

The next frontier is interactive books with AI. Imagine being able to query a book directly, ask clarifying questions, generate tailored summaries, or test your understanding. Tools like NotebookLM hint at this future. In a memorable conversation with Peter Norvig, we explored how books could become living, conversational companions—combining the timeless depth of long-form knowledge with the responsiveness of AI.

Pop-Up Journals for Policy Impact

Finally, I was intrigued by the potential of . These are agile, rapid-response publishing platforms created to address urgent policy questions. Instead of waiting months or years for traditional journals, policymakers could access rigorously reviewed, thematic collections of research at the moment decisions need to be made. Whether it’s climate adaptation, pandemic preparedness, or AI governance, pop-up journals could serve as a bridge between the speed of science and the urgency of policy.

Looking Ahead

SciFoo 2025 reaffirmed that AI is not only accelerating the mechanics of science but transforming its communication, accessibility, and societal relevance. At Pվ��Ƶ, our mission is to is to be part of progress by enabling millions of researchers, educators, learners and health professionals to access, trust and make sense of the latest insights and discoveries – every day. In this new era, progress will be driven not just by human creativity but by the collaboration of humans, machines, and communities across the globe.

AI is the accelerating force of progress—and the journey has only just begun.