Benevolent Disruptors

In this episode of the Benevolent Disruptors Interview Series, we were joined by Oliver Cameron, co-founder and CEO of Odyssey. Oliver ran product and engineering at Udacity alongside Sebastian Thrun, then co-founded the self-driving company Voyage. After Cruise acquired Voyage, he led the launch of Cruise’s fully driverless service in San Francisco. In late 2023 he co-founded Odyssey with Jeff Hawke, formerly of Wayve, to build general-purpose world models, and the company has just released Odyssey-3. Throughout our conversation we cover: Physics is to world models what coding is to language models. Odyssey-3 is state of the art on nearly every world model metric, but the one Oliver cares about most is physics accuracy. He traces Odyssey’s origins back to a lesson from self-driving: by 2018, cars had near-perfect perception and still couldn’t drive themselves. What was missing was the human ability to predict what happens next. Why a billion robots need a general brain. Today drones learn from drone footage and driverless cars learn from driving data. Oliver believes this narrow approach will always be brittle. A world model learns from the whole world instead, so a car that has never seen an elephant on the road still understands how big and fast one can be. Odyssey-3 lets you embody almost anything, including a duckling in New York City, and Oliver explains why that playful demo points to something much more serious. Building a frontier lab across Palo Alto, London and Zurich. Oliver shares Odyssey’s two hardest problems. The first is data infrastructure that he expects will eventually exceed the scale of YouTube. The second is the research challenge of making generated video as stable as generated language. Oliver and Chris also discuss three related questions: why big tech’s data advantages haven’t turned into world model leadership, why startups can be more stable than the giants, and how the shift in tone from Dragons’ Den toward Shark Tank reflects a new confidence among European founders. 👉 Learn more at www.bnvtcapital.com

In this episode of the Benevolent Disruptors Interview Series, we were joined by Jeanne DeWitt Grosser, COO of Vercel. Jeanne worked on Gmail and helped shape Google Workspace at Google, then helped grow Stripe’s revenue from $100 million to several billion dollars. She now leads Vercel’s push into AI-native web development and sits on the boards of Shopify and Upside. Throughout our conversation we cover: The agentic era has arrived faster than anyone expected. In January, 3% of deployments on Vercel were written by AI agents. Less than a year later, it’s 60%. Jeanne explains why a new era of software needs new infrastructure: platforms that agents naturally reach for, new primitives like AI gateways and sandboxes, and infrastructure that essentially heals itself. Open source as a growth engine. Vercel was born open source with Next.js, and that ethos has become a strategic advantage. Companies talk about product-led growth, but Vercel has pioneered framework-led growth, and now agent-led growth, where well-documented, community-backed tools become the default choice for AI agents. Lessons from hyperscale at Google and Stripe. What Jeanne kept and what she left behind: Google taught her to think in orders of magnitude and to ask “what needs to be true?”, while Stripe’s narrative culture taught rigorous thinking. Jeanne and Chris also discuss the now widespread OKR mindset, and talk about needing to be mindful of a company’s greatest strengths becoming its biggest weakness. 👉 Learn more at www.bnvtcapital.com

In this episode of the Benevolent Disruptors Interview Series Benevolent we were joined by Mehdi Ghisassi pioneer of product at Google DeepMind, the first PM on Google Gemini and now spinning out AI71 where he helps governments and companies build their own AI stacks. Throughout our conversation we cover: An end of an era for protected research sanctuaries inside Big Tech - how DeepMind proved that if you give brilliant researchers and engineers an ambitious goal and real resources extraordinary progress can follow. The premium of the DeepMind Diaspora - how concentration of talent has dissipated into new ventures from Mistral, to Frontier AI labs, to Orbital Materials Unreaslist expectations around sovereign AI - Governments and enterprises talk about it constantly, but full sovereignty isn't realistic for anyone. The more useful question is which parts of the stack do you build vs buy? 👉 Learn more at https://www.benevolentdisruptors.com/

In this episode of the Benevolent Disruptors Interview Series we were very pleased to welcome Vincent Koc, Chief Architect from The Open Claw Foundation to explore what’s at the frontier of personal AI agents. Throughout our conversation we get into: Dark Factories where agents build software with minimal human involvement Why the human in the loop may be the wrong framing; the case for portable Open AI infrastructure What happens when 60 agents work simultaneously; Why the interface between humans and agents could become one of the most interesting design problems of the next decade 👉 Learn more at https://www.benevolentdisruptors.com/

In this episode of the Benevolent Disruptors Interview Series, we welcome Jack Zhang, CEO and Co-Founder of Airwallex, the global financial infrastructure company processing $60B+ in annualised transaction volume across 200,000 businesses in over 200 countries, built not from Silicon Valley, but from a Melbourne coffee shop. Jack turned down a billion-dollar acquisition offer from Stripe, survived COVID, and spent a decade quietly doing the hardest things in fintech, obtaining licenses in 70+ markets, integrating with 80+ banks, and building what is effectively the next-generation global bank for entrepreneurs. His philosophy? Take the path of maximum resistance. It's working. The conversation covers why starting outside the Valley was Airwallex's biggest advantage, how a single hiring philosophy shift changed everything, and why the AI era will produce two or three financial operating system winners, and Airwallex intends to be one of them. Key Takeaways: Why turning down Stripe's billion-dollar offer forced Airwallex to think at a completely different scale How ditching experience-based hiring for competency and first-principles thinking transformed the company's culture and trajectory Why Airwallex's Asia Pacific profit engine funded its rapid expansion into Europe and North America The realist case on stablecoins: even 10% of the $7 trillion daily FX market moving on-chain would be a trillion dollars a day, but 90% remains fiat, and that's the real problem to solve Why agentic payments will run through regulated wallets, with AI agents transacting via stablecoin as a universal currency Jack's 2030 vision: 1 million customers, $20K revenue per customer, $20 billion top line 👉 Learn more at https://www.benevolentdisruptors.com/