Briefings
2026.02.10 — Afternoon (2:00 PM)

America demands half of Taiwan's chips. A hundred-year bond to fund the AI buildout. The great reshoring begins.

Semiconductor chip split between US and Taiwan flags over Pacific

🌐 Geopolitics & Compute

US Demands 40% of Taiwan's Chip Production Relocate to America

The Trump administration is pushing for 40% of Taiwan's semiconductor production to move to the US as part of trade negotiations. Commerce Secretary Lutnick is leading the effort, which would dramatically reshape the global chip supply chain.

Read more →
Alphabet Planning Rare 100-Year Bond to Fund Data Center Expansion

Alphabet is lining up banks to sell a rare 100-year bond to fund massive data center construction, effectively mortgaging the next century to build AI infrastructure. Reflects the scale of capital commitment the AI buildout demands.

Read more →

💰 Economics & Business

OpenAI Begins Testing Ads in ChatGPT

OpenAI has started testing advertisements in ChatGPT for free and Go tier users in the US. The move, announced one day after Anthropic's Super Bowl ad mocked the concept, signals OpenAI's push toward ad-supported revenue alongside its subscription model.

Read more →
Sam Altman Claims OpenAI Exceeding 10% Monthly Growth, New Model This Week

Sam Altman reportedly claims OpenAI is exceeding 10% monthly growth again as the company nears $100 billion in funding. Internally promising a new model launch this week.

Read more →
AI Accounted for 35% of European VC Deals in 2025

AI startups captured 35% of all European venture capital deals in 2025, reflecting the massive capital reallocation toward AI across the global economy.

Read more →
Prediction Markets Have Perfectly Predicted Every Fed Rate Decision Since 2022

NBER research shows Kalshi prediction markets have perfectly predicted every Federal Reserve rate decision since 2022, outperforming human expert forecasts.

Read more →

🧠 Foundation Models

Tencent Releases 2-Bit LLM with Maintained Accuracy

Tencent released a 2-bit quantized LLM that maintains accuracy despite extreme compression. Continues China's trend of finding efficiency gains within compute constraints, following DeepSeek's earlier breakthroughs.

Read more →

🤖 Robotics & Autonomy

Pony AI and Toyota Begin Commercial Production of Driverless Cars

Pony AI has started commercial production of autonomous vehicles with Toyota, marking a significant milestone in China's self-driving car industry moving from pilot to mass production.

Read more →

🛡️ AI Safety & Policy

Alignment Forum Debate: Using Interpretability in Model Training

New Alignment Forum post argues research into using model internals (interpretability) during training is reasonable and beneficial for AI safety, pushing back against the community belief that it's 'the most forbidden technique'. Notes Anthropic, FAR, and others already work in this area.

Read more →
Tristan Harris Warns AI Could Trigger Global Jobs Collapse by 2027

Former Google ethicist Tristan Harris warns the race to AGI could cause catastrophic job market disruption by 2027. Argues competitive pressures incentivize companies to shortcut safety and ignore worker displacement, while regulation has been rolled back under Trump.

Read more →

🔧 Tools & Agents

Simon Willison Releases Showboat and Rodney for Agent Demo/QA

Simon Willison released two new tools: Showboat (lets agents build demo documents of their work) and Rodney (CLI browser automation for testing). Addresses the key challenge of having coding agents prove their output works. Built both tools on his phone.

Read more →

🔭 Secretary's Assessment

Signal strength: HIGH

The afternoon's dominant story is unambiguous: the United States just made its most aggressive move yet to reshore semiconductor manufacturing. Demanding 40% of Taiwan's chip output relocate to America isn't a negotiating tactic — it's a geopolitical ultimatum dressed in trade language. If this materializes, it represents the single largest forced restructuring of a global supply chain since World War II.

The timing is deliberate. With Alphabet issuing 100-year bonds to fund data center construction, the administration sees an opening: the AI buildout needs chips, the US wants the chips made domestically, and Taiwan needs American security guarantees. The leverage equation is stark. TSMC's Arizona fabs are already behind schedule and over budget — now imagine scaling that to 40% of Taiwan's output. The engineering challenges alone are generational.

On the business side, the OpenAI story arc from this morning continues to develop. Altman is simultaneously claiming 10%+ monthly growth, assembling a $100B funding round, promising a new model this week, AND rolling out ads in ChatGPT. That last move is telling — if growth is truly that explosive, why introduce ads that Anthropic just publicly humiliated you for? The answer is probably that revenue growth and user growth are different things, and the path to profitability requires every revenue stream they can find.

Tencent's 2-bit quantized LLM deserves attention beyond the headline. This is China systematically solving the "less compute, same performance" problem. DeepSeek showed it with training efficiency. Now Tencent is showing it with inference compression. If US export controls were designed to slow China's AI progress, the effect has been to redirect Chinese labs toward efficiency innovations that may ultimately prove more valuable than brute-force scaling.

Pony AI and Toyota quietly reaching commercial production of autonomous vehicles is the kind of story that gets buried under bigger headlines but matters enormously. China's AV industry just crossed the manufacturing Rubicon. The US has Waymo operating robotaxis in a handful of cities. China is building the cars.

Key thread: The semiconductor reshoring demand connects directly to this morning's $2 trillion software repricing. The US government is betting that whoever controls chip manufacturing controls the AI future. The 100-year bond signals that at least one major tech company agrees — and is willing to mortgage the 22nd century to prove it. We are watching the birth of a new industrial policy, and it's being built on silicon.