AI Cost Cuts and Market Risks: Key Takeaways

Polymarket prediction market reveals severe inequality: 0.1% of accounts capture 67% of profits, highlighting the illusion of easy money. — via 1
AI spending can be halved without curbing usage through model selection, caching, and context optimization, per @eladgil. — via [@eladgil]@x.com
US durable goods orders misleading; AI capex benefits Asian manufacturers, not domestic GDP, warns @LynAldenContact. — via 1
@pmarca contrasts US vs China AI ecosystems: China's Zhipu, MiniMax, ByteDance thriving; US must continue open-source models as cultural ambassadors. — via 1 2
Market bubble concerns: @MebFaber warns of extreme valuations (Japan bubble analog), while @awealthofcs and @MebFaber note long-term holding still yields ~8% even buying at highs. — via 1 2
@pmarca argues data centers are the steam engine of AI era, opposing Texas moratorium; @Jason calls for Microsoft to own frontier models and compete with OpenAI. — via 1 2

@eladgil reports that optimizing default model selection, intelligent routing, caching, context pruning, and usage transparency can cut AI costs by nearly half while token usage continues to grow. — via 1
@LynAldenContact notes that US durable goods orders are misleading: AI capex goods are largely imported, so GDP growth actually concentrates in Asian manufacturers. — via 1
@pmarca details China's AI landscape: Zhipu is akin to Palantir, MiniMax resembles Character AI with overseas revenue, and ByteDance's Doubao is China's most popular AI app. China is catching up in semiconductors via 3D IC and EUV R&D. — via 1
@pmarca firmly opposes Texas's proposal to pause new data centers, calling them the 'steam engine of the AI era' and essential for productivity. He argues the US must release open-source AI models as cultural ambassadors to prevent Chinese open-source dominance. — via 1 2
@Jason urges Microsoft to own a frontier model and compete directly with OpenAI, while raising p-doom from 7 to 11 and calling for Nvidia and Microsoft to amplify open-source efforts 100x to create balance. — via 1 2

@charliebilello reveals that on Polymarket, only 0.1% of accounts capture 67% of profits, calling out fake 'easy money' videos. — via 1
@awealthofcs and @MebFaber independently demonstrate that even buying at all-time highs and holding long-term yields ~8% annualized returns, supporting a buy-and-hold strategy. — via 1 2
@MebFaber warns of extreme market valuations: Japan's bubble had CAPE 100, earnings multiples near 100, and absurd real estate prices. He compares current market to a party where no one wants to leave but the clock has no hands. — via 1 2
@CliffordAsness criticizes Michael Saylor for using misleading 30-day Sharpe ratios and incorrect return calculations, accusing him of deceiving retail investors. — via 1
@charliebilello lists past decade asset returns: NVIDIA +17,076%, Bitcoin +9,600%, Tesla +2,849%, posing the question 'next decade?'. — via 1

@charliebilello highlights that from 2015 to 2025, US 30-year mortgage rates rose from 3.5% to 6.5%, median new home price from $350k to $540k, causing down payment to increase $38k and monthly payment to rise 117%. — via 1
@LynAldenContact reports that shipping through the Strait of Hormuz is gradually improving, a positive signal for global energy flows. — via 1
@pmarca notes that AI era incentives are shifting firms to replace human labor with tokens; department managers trade headcount for compute, creating an internal 'Hunger Games'. — via 1

Discover