Daily TEA – Agents, Steel & SPV AI Debt

Browserbase, Ivan Zhao on AI, CCP-China rockets, CCP-China AI control, SPV-financed data centers

Dec 25, 2025

Hello, dear TEA-mates—here’s what you need to know today.

1.🛰️ Browserbase Taps Coinbase x402 for Agentic Browser Payments

Browserbase has launched support for Coinbase’s x402 protocol so AI agents can spin up real browser sessions and pay per use in USDC without API keys, manual billing, or card setup. Each x402 call deposits funds directly into Browserbase’s Coinbase Business account, which provides custody, yield on idle balances, real-time revenue tracking, and removes reconciliation and fraud risk. This setup lets agents discover Browserbase via the x402 Bazaar, understand pricing, and autonomously purchase fully managed browser sessions as part of larger workflows. Browserbase positions this as unlocking a new class of automation where agents navigate real websites, interact with live interfaces, and pay for compute as they go. (Read More)

🫖 TEA For Thought: Making it easier for users and agents to pay is crucial, and it will be fascinating to see how Browserbase scales this model.

2.✍️ Ivan Zhao: From Steel and Steam to “Infinite Minds”

Notion CEO Ivan Zhao compares AI to past “miracle materials” like steel and semiconductors, arguing that just as steel defined the Gilded Age and chips defined the Digital Age, “infinite minds” will define the AI era for knowledge work. He traces history from Andrew Carnegie’s rise in industrial Pittsburgh to today’s desk workers, noting that programmers already act as managers of multiple AI coding agents, turning a “10× engineer” into a 30–40× force by orchestrating agents that think and work while humans sleep. Zhao argues that most knowledge work is still “bicycles on the autobahn,” held back by fragmented context and weak verifiability compared with code, and that real progress will require consolidating workflows and moving humans from being “in the loop” to supervising leveraged loops. Using metaphors from steel frames and steam engines to modern megacities, he suggests AI will let organizations scale like skyscrapers and cities powered by “infinite minds,” transforming org charts, meeting rhythms, and the structure of economies once we move beyond merely bolting chatbots onto old workflows. (Read More)

🫖 TEA For Thought: Written by the beloved, romantic Notion CEO Ivan, this piece reads as artful as his photography—a reminder that while everyone is busy building, we still have to ask what future we are trying to create.

3.🚀 CCP-China’s Second Reusable Launch Test Underscores Gap With SpaceX

CCP-China has carried out its second reusable launch attempt in three weeks, using a Long March-series rocket in a bid to land and recover a booster and lay the groundwork for future reusable systems. State media reported that while the rocket’s upper stage reached its intended orbit, the reusable first stage was not successfully recovered, marking another setback as CCP-China tries to catch up to a decade of operational experience held by SpaceX. Reusable boosters are seen as key to lowering launch costs and enabling competitive satellite constellations, and multiple CCP-China state-owned and commercial teams are now racing to be the first in the country to nail a successful recovery. (Read More)

🫖 TEA For Thought: Building a reusable rocket is hard—Elon Musk’s team pulled it off in the U.S., while CCP-China’s state-backed giants are still struggling, underscoring how the U.S. environment for innovation, immigration, and empowerment remains uniquely powerful.

4.🧱 CCP-China Tightens Grip on AI as Party Fears Erosion of Control

CCP-China authorities are tightening oversight of artificial intelligence amid fears that powerful models and uncensored data flows could weaken Communist Party control over information and public opinion. New rules and enforcement campaigns target both domestic AI companies and online platforms, seeking to ensure systems align with “core socialist values” and do not generate politically sensitive or destabilizing content. Officials are particularly focused on the risks posed by rapidly improving open-source models and the proliferation of AI-powered apps that can be fine-tuned or deployed outside tightly managed CCP-China platforms. (Read More)

🫖 TEA For Thought: This is just the tip of the iceberg—think of fast, cheap open-source models like Qwen and the many companies built on top of them, while CCP-China tries to control the “ingredients” going into the bowl and still tastes every dish before it is served.

5.🏗️ SPV AI Data Center Debt Pushes $120bn Off Balance Sheets

A Financial Times analysis finds that tech groups including Meta, Elon Musk’s xAI, Oracle, and CoreWeave have shifted more than $120bn of AI data center spending into off-balance sheet special purpose vehicles (SPVs) funded by Wall Street investors. Private credit giants such as Pimco, BlackRock, Apollo, Blue Owl Capital, and banks like JPMorgan have supplied at least $120bn in debt and equity into SPVs that own data centers or Nvidia GPUs, letting hyperscalers preserve strong credit ratings while still financing massive AI infrastructure bets. Deals such as Meta’s $30bn Hyperion SPV in Louisiana and Oracle’s multibillion-dollar structures for OpenAI-related facilities mean lenders primarily have recourse to the underlying assets, raising questions about who ultimately bears the risk if AI demand or asset values fall. The report warns that as SPV financing and even securitisations of AI data center loans spread through opaque private credit markets, stress at a few major AI tenants could ripple across multiple funds and institutional investors, effectively hiding leverage while still adding “outstanding liabilities” to hyperscalers’ real economic risk. (Read More)

🫖 TEA For Thought: This reveals a different bubble—not in lofty AI company valuations, but in the layers of debt used to hyperscale data centers through SPVs that quietly shift default risk away from sponsors.

Prompt Tip of the Day: `“Merry Message Generator”`

The “Merry Message Generator” (cards/texts/emails) 

Write 12 different holiday messages I can send to: coworkers, close friends, extended family, neighbors. 
Mix tones: funny, heartfelt, minimal, and professional. 
No clichés. No religious language unless I ask. 
Keep each under 240 characters.