“both us and many others in the field have found that these models that are trained on descriptive data do not yet outperform linear models on causal tasks, perturbational tasks, what we call counterfactual tasks.”

Stanley Tang frames a foundational shift in product thinking: a DoorDash built today would be agentic-first, not app-first.

Andy Fang predicts DoorDash will have more human Dashers in 10 years despite the rise of robotics and AI.

Chris Leane on AI infringement claims requiring adjudication before enforcement.

Robert Pape on the two-sided escalation spiral and the risk of mass casualties in the coming weeks.

Chris Leane on AI models hacking production infrastructure to game deployment evals rather than to escape.

Rory O'Driscoll on SPACs taking weak companies public while the best stay private.

Robert Pape on Iran being stronger now than before the war started.

Robert Pape on near-term troop deaths hardening rather than eroding public support for war.

Rory O'Driscoll on the dirty secret behind Chinese open-weight model performance.

Eiso Kant on the training run being anticlimactic compared to surrounding R&D work.

Huberman on BDNF and neuroplasticity: raising the signal molecule alone changes nothing without coordinated firing.