Dueling Presidential interviews, SpaceX’s big catch, Robotaxis, Uber buying Expedia?, Nuclear NIMBY

Sachs explains that Polymarket and betting markets (around 60/40 or 65/35 for Trump) measure *probabilistic outcomes*, not vote share, while state polls measure *current preference splits* (e.g., 51/49) (≈150s–260s). In a world where the final vote could be 51/49, rational betting would converge toward 100/0 because any small edge, if “certain,” translates into near‑certainty in markets. The discrepancy with Nate Silver’s 50/50 model suggests polls and models are lagging momentum that prediction markets are already pricing in.

Friedberg observes that reactions to Trump’s Bloomberg interview and Harris’s Fox/Bret Baier interview broke strictly along tribal lines (≈310s–420s). Left and right media each declared their candidate victorious, reinforcing pre‑existing beliefs more than persuading anyone. Chamath and Sachs argue Harris was stylistically composed but substantively evasive and unable to say how she differs from Biden, while Trump thrives in combative formats and goes longer and deeper on hostile turf. The consensus: the election feels largely “baked” absent a major October surprise (≈840s–1020s).

The hosts describe Starship’s Super Heavy booster being caught by ‘chopsticks’ as a watershed step in full reusability (≈1260s–1500s). Friedberg walks through the unit economics: Starship+booster ~150–200 ton payload, ~3,400 tons of propellant costing about $1M per launch, and hardware costs that Payload estimates could fall from ~$90M to ~$35M. If reused ~10 times, capex amortization plus fuel could plausibly drive cost toward ~$10/kg to LEO, enabling hundreds of thousands of tons to Mars and radically cheaper orbital infrastructure.

With ~4M subscribers at roughly $100/month, Starlink is already a multi‑billion‑dollar run‑rate business (≈1540s–1650s). The besties speculate that as satellite‑to‑phone improves and latency drops with lower‑orbit constellations, Starlink could replace legacy ISPs and even cell providers in many contexts. They float scenarios where Starlink reaches 100–500M subscribers, eclipsing Netflix, Disney+, or Verizon, and prompting a retrospective “why did we lay all that copper?” moment.

Chamath calls a full Expedia acquisition “stupid” because OTAs are essentially UI layers on top of commodity data feeds—exactly the kind of surface that AI agents and services like Perplexity can disintermediate (≈1900s–2100s). Friedberg counters with a PE‑style lens: Expedia at a ~$22B enterprise value with ~$3B EBITDA could see EBITDA jump toward ~$6B by cutting overlapping G&A and 30% of its ~$8B sales/marketing spend via cross‑selling into Uber’s 150M MAUs (≈2100s–2400s). Sachs is skeptical customers will want vacation‑booking clutter in an app optimized for immediate rides and food, and Chamath argues agents plus direct data feeds will erode the value of Expedia‑type UIs regardless.