AI Agent Rails, China's AI Data Access, and Soaring US Debt Lead the Day's Signals

Key Takeaways

1. US national debt has grown by $450B since July 1, averaging over $150B per day, intensifying fiscal-sustainability concerns. — via 1
2. Raoul Pal argues AI agents will need crypto-native settlement rails; Lyn Alden flags rising costs and security pressure on Bitcoin infrastructure. — via 1 2
3. Chinese AI labs reportedly buy the same training data as OpenAI and Anthropic from US providers, an unverified explanation for China's rapid AI catch-up. — via 1
4. AQR research says separating market and active exposure can produce better after-tax, cost-adjusted returns. — via 1
5. US gamblers have lost $3.9T since 1929 ($5.8T in 2025 dollars), and Shiller warns future bubbles will come with new narratives. — via 1 2
6. Port Alpha picks Brownsville for 4,000 acres and 10,000 jobs; China's 230:1 shipbuilding advantage underscores the strategic stakes. — via 1

1. AI, Crypto and the Agentic Economy

1. Raoul Pal says crypto's significance goes far beyond token prices: capital will soon move at “silicon speed” as billions of AI agents transact, negotiate, and settle with each other. Traditional political and regulatory systems cannot keep up, making on-chain rails the only infrastructure suited for that velocity; owning them is like owning the next economy's operating system. — via 1
2. Because AI agents cannot use banks, Pal argues they will build their own financial plumbing: wallets replace accounts, and final settlement happens in hundreds of milliseconds. Smart-contract platforms were designed for exactly this, and while regulation may slow adoption, it cannot stop it. — via 1
3. Lyn Alden flags a security research gap: US frontier models have not found any vulnerabilities in Bitcoin infrastructure, while $10,000 a day is being spent on open-weight models such as Kimi K3 and Qwen 3.8 in an attempt to find them. — via 1

2. Fiscal, Rates and Gold

1. US debt has risen by $450B since July 1, or more than $150B per day; Charlie Bilello ties the pace to John Adams' warning that debt enslaves nations. — via 1
2. Lyn Alden shares the view that rate hikes are far less effective against deficit-driven inflation than credit-driven inflation; doubts about whether a “Warsh Fed” would hike leave room for gold. — via 1
3. Raoul Pal notes that most investors fear consolidation, but smart money is using it to accumulate; his discussion covers cycle positioning, yield-curve signals, and the likely path forward. — via 1

3. Markets, Bubbles and Investment Strategy

1. Meb Faber points to data showing US gamblers have lost $3.9 trillion in nominal terms from 1929 to 2025, or $5.8 trillion in 2025 dollars, a striking measure of speculative losses. Robert Shiller adds that future bubbles will be built around new narratives and will need new names. — via 1 2
2. AQR's latest paper argues that separating market exposure from active exposure can improve after-tax and cost-adjusted performance. Clifford Asness compares the improvement to Watt's refinement of the steam engine. — via 1
3. Ben Carlson's trip to Comedy Cellar yields a sentiment data point: a comedian called Nvidia a TV brand, which he reads as a sign that the AI boom may still have a long way to run. — via 1

4. Geopolitics, Business and the AI Data Race

1. @jason highlights that Chinese AI labs buy the same training data as OpenAI and Anthropic from US data providers. He frames the practice as a possible explanation for China's rapid AI catch-up, but the claim remains unverified. — via 1
2. Port Alpha is landing in Brownsville with 4,000 acres and 10,000 jobs. The broader All-In discussion covers China's 230:1 shipbuilding advantage, the Navy's first autonomous rescue, a $3B destroyer vs. Marauder comparison, and autonomous weapons in the Taiwan context. — via 1
3. A startup once valued at $9B was almost sold for a token price, a reminder that private-market valuations can collapse quickly. — via 1