Sourcery: AI Safety, Energy Demand, and the Summit’s Wildest Hot Takes

Molly O'Shea (0:01)

When the biggest names in tech, energy, food, and media gather in one room, the result is less like a conference and more like a live referendum on the future. The All-In Summit 2026 brought together people like Brad Gerstner, Naveen Rao, Jonathan Neman, and even Jake Paul, with major figures like Elon Musk and Jensen Huang shaping the backdrop. And the central question running through everything was how power, intelligence, and risk reshape the world from here.

What struck me most was the AI safety conversation. It felt like the room was moving toward pragmatic middle ground rather than panic or denial. Exactly. Brad Gerstner said the summit showed how consensus forms in real time. People are converging on responsible action, company-led control, and some form of peer review. Even Elon Musk said, 'I tend to agree with Dario,' while Jensen Huang insisted the labs themselves need to control the technology and do a better job. That's interesting because Naveen Rao of Unconventional AI added a hardware angle to all this. He says his company can demonstrate 'many orders of magnitude of efficiency' with chips using 'nanojoules per image generated.' He's arguing that power efficiency is becoming the bottleneck for AI scaling. And he thinks cheaper intelligence could eventually reach 'your toaster, to everything.' His larger view is pretty blunt: 'AI is the next level of evolution for our species,' and fears of extinction are mostly 'coordinated vibes.' That optimism about abundance really shows up in the energy conversation too.

Toby Rice of EQT says the AI revolution will drive huge demand for natural gas, both for data centers and the grid more broadly. His message is simple: energy security, AI growth, and manufacturing all point toward more natural gas, not less. Building on that acceleration theme, Wen Sang of GenSpark made a different case. He says, 'We won't slow down' because the real problem isn't capability, but access. Most people haven't learned how to use AI yet. GenSpark is trying to make AI as easy as Google so ordinary knowledge workers can command agents without needing technical expertise. What do you think about the food and health takes? Jonathan Neman of Sweetgreen had a sharp one: 'AI is not going to kill us, but processed food will.' That was memorable. He pushed back on summer fears about cyclospora, argued that fresh food was unfairly stigmatized, and says Sweetgreen is built on knowing its farmers and sourcing responsibly. He also shared a small secret for customers: mix two dressings. Then there's Jeff Woo of Antifund going full philosophical, saying, 'I don't believe in free will.' He claims human beings mostly narrate decisions already shaped by genetics and environment. He also says the singularity is already here, just unevenly distributed.

Some of the other takes were more political and practical. Nick Shirley says California should stop building high-speed rail after spending more than $15 billion without connecting a single city.

Jake Paul argues we need an amended constitution, saying modern weapons may make the old Second Amendment framework obsolete.

And Kent Draper emphasized execution in AI infrastructure, while Michael Brandt frames the human body as the next major innovation platform, built around ketones, wearables, and performance optimization. Together, the episode captures a summit obsessed with one theme: abundance is coming, but only if the world can build fast enough.