AI+X Summit 2026 Journey Song
On October 1st, I spent the day at the AI+X Summit at StageOne in Zürich-Oerlikon, the flagship gathering of the Zurich AI Festival, hosted by the ETH AI Center together with ZHAW and UZH.ai. Around two thousand researchers, founders, students, and industry leaders came together under one roof to ask what AI means for every field it touches. The X is the part that matters.
Specifically, I was there:
🔹 To explore the Swiss AI ecosystem, and the place where ETH, academia, and industry meet.
🔹 To meet Swiss startups and university students working on AI, and listen for where we might learn and build together.
A People Without Knowing It
Arrival · Being Human
I arrived in Zürich during the week of the Zurich Film Festival. Walking near the Opera House, in the heart of the festival, I came across Greta Thunberg. Her words have shaped how I think about the future, so I approached her, told her they had changed my life, and thanked her. She smiled and thanked me back. I was glad not to come across as a stalker.
More often than not, I feel that many of us are working toward the same goal without ever having met. That quiet, shared understanding makes us a people. I find it one of the most valuable parts of being human.
I carried that feeling with me into the summit, a day devoted to exploring what AI is and how our relationship with it can grow.
Europe Has the Talent. Does It Have the Patience?
10:35 · Europe's Moment
The day opened with Lukas Leitner walking us through the 2026 Deep Tech Report. His definition was clean: deep tech is the work of carrying a scientific breakthrough out of the lab and turning it into a product for the first time. Every era has its own. The combustion engine was deep tech once.
The numbers told a story of contrast. Europe produces more deep tech startups than the US at the earliest stages, and the gap in seed funding is narrower than most assume. But at growth stage the gap widens dramatically, and European startups fail more often between rounds. Meanwhile, nineteen of the twenty critical minerals the energy transition needs are refined in China.
What stayed with me was not the gap itself but what it asks of us. Europe has the research, the talent, and the problems worth solving. What it lacks is long-term capital willing to stay through the hard middle. That is a question for families as much as for funds.
Give Them the Problem, Not the Instructions
10:55 · Agency & Ownership
In the Next Gen Foundation Models session, Robert Jakob and Pascal Bertrand told the story of how the Agentic Systems Lab came to be at ETH. It did not start in a lab or an incubator. It started among students at the Student Project House, building with coding agents faster than any curriculum could follow, with almost no support and a great deal of peer learning.
So the lab turned the usual model around. Instead of handing students small pieces of a professor's research, it invites them to bring their own ideas, and mentors rather than supervises. Admission is earned by building a proof of concept, not by grades. Eight hundred applied this year. One master's student published nine papers in six months. Students on the industry track built prototypes that have already been deployed.
Pascal named the quality they look for: ownership. When people work on their own ideas, they go the extra mile when things break. In a world where anything can be built in a week, the real question becomes what you choose to build. I left thinking about how much this applies beyond universities, to every team and every family trying to grow the next generation of leaders.
Listening Before Building
11:25 · Agency & Ownership
The Pavoot team started in Zürich with a simple problem: event photos that arrive weeks late or never. That idea took them into Y Combinator and to San Francisco. Before deciding what to build next, they sat in the offices of more than a hundred event marketers, not to interview them but to watch how they actually work. They stopped only when they could predict the answers before hearing them.
Afterwards I sat down with Ana Yoon Faria de Lima. Pavoot now helps companies see what their events actually produce, connecting each gathering to the relationships and outcomes that follow months later. I shared how much of our own work happens through convenings: at TÜSİAD, FBN, the Turquoise Coast Environment Fund, and soon at COP31 in Türkiye. Our events are rarely about leads. They are about bringing systemic players together. We explored how that kind of gathering could be held across many rooms and remembered as one thread.
We agreed to stay in touch, and explore meaningful use cases. A gathering is only the beginning of a relationship. The tools that honor that are the ones worth building.
Trust Is Built Slowly and Lost Quickly
11:45 · Trust & Sovereignty
Stefan Klauser of Aisot asked the question every investor using AI eventually faces: can we trust agents where the stakes are high? Technically, he said, agents can already run the whole investment chain, from gathering data to building and monitoring portfolios. The harder question is what we should let them do. Finance is a trust business. Lose it once and you are gone.
His answer rested on three things: precision, reliability, and control. The right model for each task. Data from verified sources, processed where the client chooses. Every output traceable back to the factors that produced it. And always a human who sets the objectives, holds the accountability, and makes the final call. A live demo showed how a general-purpose model swung wildly when a single number in a news article changed, while a purpose-built one held steady.
Afterwards I met Stefan at their booth and walked through a demo. I shared where my curiosity sits: with the capacity to perceive and process information, AI should eventually help us make better investment decisions than we do alone. We are not there yet. But the path runs through exactly what he described, machines that analyze and humans who remain answerable.
Home Ground
Afternoon · Trust & Sovereignty
One word kept returning through the afternoon: sovereignty. At the Unique booth, I learned how a Swiss team brings generative AI into private banks. One of their largest clients runs the platform for six thousand employees, entirely on its own servers, on its own models, with nothing leaving Switzerland. Many of their data scientists come from ETH. A small security team does nothing but follow regulations that shift month to month.
The main stage carried the same question at a different scale. Open Models for Real-World Impact traced the next chapter of Apertus, the open model built in Switzerland. Later, a workshop asked plainly: Hyperscalers vs Home Ground, how much sovereignty can your AI strategy really handle?
For a bank, sovereignty is a compliance requirement. For us, it is becoming something more. When the knowledge you hold is made of other people's trust, where it lives and who controls it is not a technical detail. It is the foundation.
You Can't Ship What You Can't Specify
14:30 · Agency & Ownership
Jordis Herrmann brought us from software into the physical world, where robots walk real stairs in real buildings. Before any training begins, engineers walk the exact path the robot will take. They scan, photograph, and note every step height, gap, and surface. Then they imagine the same place in a Norwegian winter: ice, darkness, a different world.
Those specifications become millions of simulated trials, then lab tests, then a gradual rollout. Every failure in the field is studied and fed back into the loop. Progress is not linear. A capability gained in one version can quietly disappear in the next, so every requirement is tested again, every time. Done well, one airline terminal took a single week to deploy, and the robot has worked on its own since.
The title stayed with me all day. Whether it is a robot or an AI colleague, the hardest work comes before the building. You have to know, precisely, the world you are building for.
Knowledge That Belongs to Those Who Built It
Afternoon · Knowledge as Commons
The conversation that met my question most directly was with João, co-founder of Uthereal, an ETH spin-off. Their work starts from a conviction: organizations that spent thirty, forty, seventy years building expertise should not have to hand it over to a handful of large platforms to stay relevant. Uthereal helps them turn what they know into living knowledge they own, and can offer to others on their own terms.
I shared where we are. Weya takes notes from our conversations with investors, founders, and partners, because much of our work is connecting people who should know each other. That web has grown beyond what any of us can hold in mind. I described three layers we are reaching for. A personal layer that remembers. An organizational layer that tells me when a colleague's conversation connects to mine. And a commons layer, where organizations working on the same challenge, decarbonization for instance, could find one another through what they already know.
João's answers on privacy were concrete: models they host themselves, zero data retention agreements wherever data must travel, servers in Europe. We set a time to meet again in October. If we get the foundation right, knowledge stops being something we store and becomes something we share with care.
Bringing It Home
Evening · Building Weya
At the end of the day, I connected with our AI team to evaluate where Weya stands. We carried the day's conversations into our own work, looking honestly at what is working, what is not yet, and where to focus next.
The summit did not hand us answers. It gave us clearer questions about knowledge, ownership, and trust. The work of answering them is ours to do, together.
The Meaning We Give Our Tools
Night · Being Human
After the summit, I asked Claude what the festival offered that evening and how it would rank the options for me. One of its suggestions took me to the gala premiere of Danny Boyle's Ink.
The film tells the story of Rupert Murdoch buying The Sun and, with editor Larry Lamb, turning it into the best-selling newspaper in the country. Along the way, they redefined the relationship between media and the people it serves. It felt strikingly close to today, to how we live with social media and how algorithms now shape what we see.
It made me think that a tool alone does not define reality. The meaning we give it, the relationship we build with it, and the choices we make around it do. That is exactly where we stand with AI. It felt fitting that an AI pointed me to the film.
Conclusion
I came to explore "What is AI going to do next?" yet I ended up looking deeper into "Who are we going to become alongside AI?"
Somewhere, people I will never meet are asking the same thing.
That, too, makes us a people.