E99: Cheating scandals, Twitter updates, rapid AI advancements, Biden's pardon, Section 230 & more

Musk’s return to Twitter’s $54.20 offer was framed as a contract problem, not a fresh valuation decision. Chamath said the specific-performance clause left few exits, while the banks were roughly $1–2 billion underwater on committed debt and could not freely walk until the financing deadline.

Chamath’s case for making the Twitter price work depended heavily on radical cost reduction. He estimated Twitter’s stand-alone value at $32–35, argued roughly $10 a share could be recovered through operating cuts, and floated cloud migration as another multi-billion-dollar lever.

Chamath said Facebook’s growth team created the “data scientist” title almost accidentally while recruiting a Google PhD in 2007–08. The candidate rejected “data analyst” as beneath his self-image as a scientist; Facebook changed the offer, hired him, and the title then spread internally.

The strongest AI thesis was that computing progress had shifted from transistor density to brute-force scale in GPUs, power and capital. Chamath argued this keeps effective compute compounding despite Moore’s Law slowing, while multimodal systems push toward intelligence whose marginal cost approaches zero.

Friedberg’s most concrete prediction was that AI moves people from creators to “narrators.” Instead of manually producing blueprints, films or games, users would describe and iterate on what they want while software performs the creation—turning knowledge work into direction rather than execution.

Tesla AI Day was described less as a product launch than a recruiting demonstration of an integrated AI stack. Tesla showed early Optimus robots with a stated $20,000 target and emphasized Dojo, custom chips and simulated driving scenarios, placing robotics atop much of the same autonomy infrastructure.

The Section 230 argument became more interesting when the speakers separated content platforms from the infrastructure beneath them. Sacks softened his case for treating Twitter or YouTube as common carriers, but argued payment rails, banks, ISPs, cloud hosts and similar “pipes” should not discriminate by viewpoint.

Their most original compromise was “algorithmic choice”: platforms could expose recommendation systems like an app store rather than impose one feed. Users could choose family-safe, educational or other third-party filters, separating hosting from editorial ranking and creating a market for algorithms instead of one platform-controlled default.