Meta's scorched earth approach to AI, Tesla's future, TikTok bill, FTC bans noncompetes, wealth tax

Meta’s AI strategy is less about selling models than making models cheap enough to protect its existing app empire. The panel argued that open-sourcing Llama and Horizon OS, then embedding AI/search inside Facebook, Instagram and WhatsApp, could commoditize foundation models while giving Meta a new route into search.

Chamath’s criticism of Meta’s spending was not the AI investment itself, but what he saw as excessive reliance on Nvidia for inference. He argued inference will dwarf training and can use cheaper infrastructure, making Meta’s capital allocation—not its open-source strategy—the weakness behind the market’s negative reaction.

Chamath’s energy thesis is to turn American homes into competing mini-utilities, and he said Palmetto is his largest investment. He believes rooftop solar, batteries and net metering could erode roughly 1,700 incumbent utilities burdened by enormous future capital spending and debt.

The autonomy discussion’s most useful caveat was that “driverless” can still depend heavily on humans behind the scenes. Calacanis cited Cruise workers intervening every 2.5–5 miles and argued constrained robotaxis already work, while meaningful automation across perhaps 5–10% of rides could still take five to ten years.

Chamath said his only job interview after leaving Facebook was for a prospective “head of iPhone” role at Apple in 2011. He was willing to work for Steve Jobs, but said the process ended when the reporting line became Tim Cook; he subsequently founded Social Capital.

Before iTunes’ 99-cent store, Friedberg’s AOL team had built a 99-cent music-download beta using 25–30 million stored credit cards and Warner’s catalog. He said it performed well and was demonstrated to Steve Jobs and Eddy Cue, but AOL’s internal politics blocked expansion; iTunes launched nine months later.

The TikTok discussion identified its recommendation algorithm—not the video inventory—as the asset that makes the product exceptional. Chamath argued a U.S. version without that algorithm would resemble Shorts or Reels, while Friedberg cited reports of roughly $14 billion in revenue, 170 million U.S. users and 40%-plus growth.