Inspired with Alexa von Tobel

The episode has a solid vein of genuinely technical content - freeway reliability budgets, empty-truck inefficiency, construction labor dynamics - but is padded significantly by an immigration backstory, a Civilization video game tangent, a quick-fire book-recommendation segment, and speculative city-transformation musings that add little operator value.

“the first hundred miles of driverless with the Jaguars, our ah, reliability that we had to hit was 1,000 times higher than what was already a pretty sizable scale already driverless in San Francisco. And so three orders of magnitude.”

“you have by design like 11 hour driving limits, federally regulated to where um, you cannot exceed that because of fatigue... 55, 60% of your time is spent on overnights”

The humanoids-as-dot-com-bubble framing is a genuinely contrarian take delivered with conviction, and the freeway reliability decomposition (software churn vs. hardware redundancy) is non-obvious. However, the autonomous-vehicles-will-change-cities section and the physical-AI-is-the-next-decade thesis are now widely circulated arguments.

“I'm pessimistic about this wave of humanoids... this could be almost like a dot com style situation where uh, it's not wrong, it's just off by a big period of time”

“when you compound all of them, you can't go driverless on a freeway without a solution... you have like catchalls with software that's constantly changing every single release and thrashing, the reliability rates. And you have to have like a space shuttle level of reliability with a system that's evolving all the time”

Boris Sofman is an exceptionally strong practitioner guest: co-founder of Anki (4M consumer robots shipped), five-year Waymo executive who actually led perception teams and the trucking program, and now founding CEO of a ~$2B robotics company - he is speaking from firsthand operational experience, not commentary.

“I was at Waymo for about five years. So I was an executive, uh, leading autonomous trucking. But then I was also leading various technology teams that supported kind um, of all platforms, cars and trucks. And so, uh, I was leading a perception team”

“we helped launch the Jaguars into San Francisco... we unified the tech stack to where cars and trucks were as unified as possible in terms of the overall architecture”

The episode is impressively data-rich in places - kernel panic frequency, reliability orders of magnitude, trucking market size comparisons, Waymo safety stats, construction labor figures, and excavator costs are all named and concrete. Some sections (city transformation, future manufacturing) drift into unsupported speculation.

“a computer, ah, will have a kernel panic that just crashes completely out of your control every 500,000 miles. But that'll eat up. That's way more than your entire budget”

“Waymo has uh, over 200 million driverless miles. 10 times safer than a human... over 40% of ride hailing in the city is waymos”

The host lands a few genuinely useful follow-ups - asking for a concrete long-tail reliability example and probing the retrofit-vs-build decision - but consistently validates rather than challenges, never pushes on speculative timelines or valuation claims, and wastes meaningful airtime on a childhood origin story, a video game anecdote, and a quick-fire mantra round.

“Can you give us an example of one of those sort of real long tail challenges that you had to think through?”

“We share that opinion”