No Priors Ep. 144 | The 2026 AI Forecast with Sarah & Elad

The hosts argue that reports downplaying AI’s productivity gains miss the long diffusion curve; adoption in areas like coding, medical documentation, and legal work is already substantial and will compound over the decade regardless of short-term sentiment swings.

Doctors, lawyers, compliance, and accounting—historically slow to adopt tech—are embracing AI for documentation and decision support, suggesting that reasoning over unstructured data maps particularly well to high-value expert workflows.

Humanoid and semi-humanoid robots are expected to see small-scale deployments that won’t fully match the hype, creating a sentiment pullback similar to early self-driving; long term, self-driving and industrial robotics should still become highly consequential markets, likely with strong roles for incumbents like Tesla, Waymo, and Chinese firms.

While scale and capital advantage push the field toward a few dominant labs, the hosts see room for alternative architectures (diffusion, SSMs, continual learning) and evolutionary or self-improvement approaches that more closely resemble specialized biological systems and may unlock new performance leaps.

Apart from chat-style interfaces, there are few truly novel consumer AI experiences; the hosts expect new agentic, context-aware products and hardware experiments (many of which may fail) and note that only a relatively small pool of truly elite consumer PMs may be capable of inventing enduring new interfaces before incumbents copy them.