Today is a big day for Foresight

We’re officially launching our first AI Nodes: physical hubs dedicated to AI for science and safety, one in San Francisco and one here in Berlin, where I’m writing this from. As Node Manager for the Berlin Node, the lead-up to this moment has substantially reshaped how I think about community, physical infrastructure, and what makes Berlin a unique place for projects like ours. As we put the final touches on launch day, I felt it an apt moment to share the first of my reflections from the last months.

I've been building communities in various forms for the better part of a decade now, across AI, web3, creative technology, and regenerative placemaking. That experience taught me something I carry into everything I do now: you can't build community for community's sake. It doesn't work. People don't show up because you curated a cosy space and sent enticing invitations. They show up because they're working on something hard and they need other people who understand the problem. What makes me more excited about Foresight than anything I've worked on before is that the hard problem came first. And now a community is forming around it. 

My academic background is in cognitive science, with a self-directed focus on interaction/experience design, and I think a lot these days about the concept of affordances: properties of an environment that invite certain behaviors without insisting on them. A park bench facing another park bench invites conversation. A well-placed whiteboard invites collaboration. I believe a designer's job isn't to control what happens, it's to make the desired behavior the easiest behavior. That's how I've come to think about our Nodes. Not as a community to be built, but as infrastructure around a set of hard problems that I'm watching a community crystallize around.

Physical proximity might just be one of the most powerful affordances we (humans) have. The bandwidth of a shared desk or room, the body language, the overheard watercooler conversations, the scribbles left on somebody else's whiteboard, the ambient awareness of what everyone around you is stuck on. All this is orders of magnitude higher than anything a video call can offer. I’m aware that's a cognitive science claim, but it's become my operating intuition as Node Manager over these last months.

I’ve noticed a new hard problem

AI development is outpacing our ability to secure, govern, distribute, and wisely apply it. Traditional security frameworks can't match the speed and complexity of novel AI-driven threats. Privacy-preserving technologies remain fragmented and hard to scale. And as AI systems proliferate into an ecology of coordinating agents, we don't yet have robust enough mechanisms that favor cooperation over zero-sum competition.

These are open technical questions that serious people (some of whom I’ve already met at our node) are actively working on. How do you build self-improving defense systems where AI autonomously identifies vulnerabilities and red-teams digital infrastructure? How do you construct confidential compute environments and scale encrypted mechanisms for handling data so that AI progress can happen openly without sacrificing privacy? How do you design decentralized intelligence ecosystems where AI systems cooperate, negotiate, and align, while still drawing meaningfully on tools from other domains like economics, game theory, and mechanism design?

These hard safety questions sit alongside equally ambitious scientific ones. Frontier models are opening new approaches to biological problems (and risks) that were largely intractable a few years ago: mapping and simulating biological intelligence to lay groundwork for brain-computer interfaces, accelerating longevity research from biostasis to molecular nanotechnology, and building better tools for AI-guided scientific discovery and epistemic preparedness. What I've seen since joining Foresight is that the safety work and the science work share infrastructure and methods, and progress in one often directly enables progress in the other. 

Our AI Nodes are to be built around exactly this kind of work. Each combines non-dilutive funding with office and event space, local compute, and programming, with a strong priority on open-source, local-first AI projects. The longer-term vision is a global network of nodes that share compute, talent, and progress while retaining their own character and governance.

This model excites me because its implications stretch far beyond Foresight.

I believe virtually every organization working on hard problems right now needs two things that are becoming increasingly difficult to come by independently: physical space where people can actually work (and be) together, and compute they can trust. Foresight can’t provide that for everyone. But our model is replicable. If you're running an organization centered on a truly hard problem and your people are scattered across Zoom calls, consider what changes when you give them a room, a whiteboard, and a GPUs they control. And consider what might happen if you don’t? 

I am often asked ‘why Berlin’?

Berlin's tech scene has a habit at the moment of benchmarking against other cities: "Boston an der Spree." "Siliconallee." I understand the impulse, I’ve even found myself doing it, but I think it sells Berlin a bit short.

Berlin has structural advantages that don't map to the Silicon Valley playbook. Universal healthcare (while painful for the pocketbook at times) means founding a company or joining an early-stage research project is less personally terrifying. Proximity to EU policy means the people building AI governance tools can realistically be in the room where regulation gets shaped. Just last week, I co-hosted a „Berlin Civic Lab“ where researchers from the Max Planck Institute sat at the same table as members of the Berlin district council and a swarm of independent builders. This kind of cross-pollination between research, policy, and building happens here with remarkably little friction. It's hard to imagine it happening as naturally anywhere else.

But there's something else that I believe gets often overlooked. Paul Graham wrote recently that in the AI age, taste becomes the differentiator. When anyone can make anything, what matters is what you choose to make. Berlin makes this apparent quickly. This is a city that has fought for its cultural identity through division, occupation, and reunification, and the result is a city with an unusual density of generalists with deep creative instincts. Artists, designers, musicians, researchers, and builders live, literally, on top of each other, and they've been sharpening each other's judgment for decades (not to mention they dance with each other, but that’s an article for another time). 

In a moment where AI is commoditizing execution, this kind of taste, or cultivated judgment — the ability to see around corners and ask the right questions — becomes not just aesthetic, but strategic. And it‘s not unrelated to why Foresight, an organization that has spent 40 years trying to see what's coming and funded multiple future Nobel laureates decades before their acclaim, chose Berlin as its first European Node location.

Berlin's tech and impact scene is moving toward what I've started recently calling radical collaboration. Organizations that would have seen each other as competitors two years ago are now sharing resources, platforming each other's work, and actively looking for ways to build together. Berlin has a specific quality that an NYU Berlin analysis described as "breadth and permeability" — people moving between creative industries, software, media, climate tech, and consumer products with relatively little friction (there’s that word again).

Not to mention this city carries a cultural memory that makes decentralization more than an abstraction. Surveillance, partition, reunification. The question of who controls information has always been existential here.

Berlin‘s unique social porousness is allowing a genuinely interesting ecosystem to emerge, without any one organisation at the center of planning it. Zuzalu, the popup city Vitalik Buterin convened in Montenegro in 2023, spawned over 20 derivative popup villages worldwide within 18 months, including ZuBerlin, which brought 80–120 technologists here for a two-week immersive residency in decentralized AI and cryptography. Edge City, which grew directly out of the Zuzalu experiment, now runs monthlong popup villages across four continents as a self-sustaining nonprofit. The Internet Archive's DWeb community, which has been convening builders and activists around decentralized web infrastructure since 2016, is holding its next camp this July in an ancient forest an hour southwest of Berlin.

Even closer to our new home, Æthos, a nonprofit hub for responsible AI startups, expanded from Cambridge to Berlin and is now our neighbor at CIC (we literally share a door). A responsible AI startup founder and a Foresight-funded safety researcher bumping into each other on the way to get coffee sounds trivial, but I've seen versions of this happen since I started, and those ambient collisions are often the most interesting part of my week. Affordances in action. 

None of these groups set out to build an ecosystem. They each formed around their own hard problem. Decentralized governance. Responsible AI. Effective resource allocation. Open web infrastructure. The ecosystem is what's emerging in the spaces between them. Paul Baran understood something about this in the 1960s when he designed communications networks at RAND: a distributed mesh, where each node connects to multiple others rather than routing everything through a central hub, is fundamentally more resilient than a centralized architecture. No single point of failure. Berlin's emerging AI and governance ecosystem works the same way. Its strength is not in its competition, but in its connections. And the flow of energy between them.

So what's to come?

Today‘s launch is a chance to showcase Foresight and give a platform to the community partners we've been building relationships with throughout these first months.

The event at CIC Berlin is a packed day: coworking in the morning, partner talks and lightning talks from our first residents in the afternoon, then dinner, drinks, and a live set from international collective Y Radio in the evening. Our ecosystem partners will share what they're working on and what we've been exploring together for 2026. Much of the year is still taking shape.

Starting this month, the Node's programming will include monthly salons and community partner meetups. Our grantees and visiting researchers will work together intensively, back to back. In July, we will host the Secure & Sovereign AI Workshop, a two-day gathering working across AI security, privacy-preserving compute, and decentralized cooperation, with presenters confirmed from Oxford, EPFL, UC Berkeley, Google DeepMind, Peking University, the Ethereum Foundation, plus others. Participants will be encouraged to stay afterward in Berlin to sprint at the node and keep building together.

I've spent the last several months learning that the most useful thing I can do in this role is stay porous, create the right affordances, and keep the hard problems at the center. I believe that here in Berlin the organizations already exist. The researchers already exist. What we're launching with our AI Node is simply the connective tissue: the shared physical infrastructure and the intentional scheduling of serendipity that lets us compound and exponentiate each other's work.

As I write this, Foresight is still accepting grant applications on a rolling basis for exceptional projects using AI to advance science and safety. Funding is flexible, and grantees get access to the Node's compute, community, and programming. If you're a builder or researcher who's been working on a hard problem alone, I'd love to talk. If you're an organization doing adjacent work and you've been looking for a coalition, let‘s put our heads together. And if you're in a position to provide compute or physical space to people working on hard problems, consider doing it. We‘d love to support you. Things are moving too fast to stay siloed. 

Bradley Clark Royes is the Node Manager of the Foresight Berlin AI Node. He holds a degree in Cognitive Science from Osnabrück University.