TLDR:

• China’s U.S. Treasury holdings fell to $682.6B, down from a $1.3T peak in 2013.
• China’s share of foreign Treasury holdings dropped to 7.3%, the lowest since 2001.
• Gold reserves reached 2,308 tonnes after 15 straight months of central bank buying.
• Total foreign U.S. Treasury holdings hit a record $9.36T despite China’s reduction.

China’s holdings of U.S. Treasuries declined to their lowest share in foreign reserves since 2001. As of November 2025, Beijing held $682.6 billion in U.S. government debt, while its gold reserves climbed to record levels.

Treasury Holdings Decline to Multi-Decade Low Share

Data shows China’s Treasury holdings dropped to $682.6 billion in November 2025. This marks a sharp fall from its 2013 peak of over $1.3 trillion.

China now accounts for 7.3% of total foreign-held U.S. Treasuries. That share is the lowest recorded since 2001. Despite the decline, overall foreign holdings reached a record $9.36 trillion. Japan and the United Kingdom remain the largest foreign holders.

The reduction has drawn attention across financial markets. However, bond markets have remained stable during the adjustment period. The figures indicate a gradual rebalancing rather than an abrupt market disruption.

On X, user Wimar X claimed that China “dumped $638 billion” in U.S. Treasuries. The post also stated that current holdings are the lowest since 2008. The tweet further suggested China is “exiting the system.”

Official data confirms the decline in holdings. However, total foreign demand for Treasuries remains strong, led by other major economies.

Gold Reserves Rise for 15 Consecutive Months

At the same time, the People’s Bank of China continued adding gold to its reserves. January 2026 marked the fifteenth straight month of gold purchases.

China’s gold holdings reached 2,308 tonnes, valued at about $370 billion. Gold now represents roughly 5% of the country’s $3.3 trillion in total reserves. This is the highest recorded level for China’s gold stockpile.

Some market observers view the shift as a move toward hard assets. Others describe it as standard reserve diversification. The increase in gold has occurred alongside the steady reduction in Treasury exposure.

Even so, China remains one of the largest holders of U.S. government debt globally. The adjustment appears gradual rather than sudden.

The combination of lower Treasury holdings and higher gold reserves reflects changing reserve allocations. Meanwhile, global Treasury markets continue to operate without major volatility.

TLDR:

• Aggregated liquidation data shows rising long and short exposure across major crypto assets.
• Bitcoin’s move to $60K triggered a new phase of positioning in derivatives markets.
• Traders expect consolidation for up to 30 days before a clear trend emerges.
• Expanding liquidation clusters increase the chance of a sharp price swing.

Recent liquidation data across major cryptocurrencies shows mounting pressure in derivatives markets. Aggregated levels for Bitcoin, Ether, BNB, and Dogecoin point to growing long and short exposure. Market participants now watch for a decisive move after Bitcoin’s return to $60,000.

Liquidation Levels Expand Across Major Crypto Assets

Crypto analyst Joao Wedson shared aggregated liquidation levels for Bitcoin, Ethereum, BNB, and Dogecoin over the past seven days. The data shows consistent growth in both long and short positions across these assets.

According to Wedson’s tweet, traders continue building exposure on both sides of the market. As leverage accumulates, liquidation clusters expand above and below current price levels. This structure often sets the stage for sharp price swings once liquidity is triggered.

He noted that the current setup increases the probability of a strong move in the coming days. When long and short positions rise together, the market often seeks liquidity in one direction. As a result, volatility tends to increase after periods of compression.

However, the data does not confirm the direction of the next breakout. Instead, it shows a market preparing for expansion. Traders remain positioned for both downside continuation and recovery.

30-Day Consolidation Expected Before Clear Direction

Wedson also stated that the market may require around 30 days of consolidation after Bitcoin reached $60,000. This cooling period could allow excessive leverage to reset. Until then, price action may remain range-bound.

Many traders continue to expect further capitulation. Others anticipate a steady recovery from recent lows. Even so, Wedson suggested that neither scenario is likely to fully develop without extended consolidation.

The return of Bitcoin to the $60,000 level marked a psychological shift. Yet sustained direction often follows structural balance. Therefore, time may be required before momentum builds decisively.

As positions accumulate, liquidation levels act as reference zones for traders. A breakout above or below these clusters could trigger cascading liquidations. That sequence may define the next major move.

For now, derivatives data reflects tension rather than clarity. Both bullish and bearish participants remain active. Consequently, the market appears positioned for volatility, though timing remains uncertain.

Ethereum co-founder Vitalik Buterin has grown wary of how prediction markets are evolving, warning they risk becoming short-term price betting engines rather than tools that support long-term infrastructure. In a post on X, he argued that the current trajectory shows “over-converging” focus on immediate price moves and speculative behavior. He called for a shift toward onchain prediction markets that serve as hedges against price exposure for consumers, rather than betting mechanisms that amplify fiat-driven volatility. The thrust of his critique centers on moving from pure price bets to broader markets that can stabilize expenditures over time. He suggested a framework that blends prediction markets with AI-driven tools to counter inflationary pressures faced by households and businesses alike. In essence, his stance positions prediction markets as potential risk-management primitives if redesigned with real-world spending in mind.

Key takeaways

• Buterin argues prediction markets are tilting toward short-horizon price betting, which he views as unhealthy for long-term building in crypto and beyond.
• He envisions a model where onchain prediction markets are paired with AI large-language models to hedge consumer price exposure across goods and services.
• The proposed system would create price indices by major spending categories and regional differences, with prediction markets for each category.
• Each user could have a local LLM that maps their expenses and generates a personalized basket of prediction-market shares representing several days of future outlays.
• Supporters say such markets can offer valuable market intelligence and hedging capabilities, potentially improving price stability in a fiat-dominated environment.
• Existing prediction-market platforms like Polymarket and Kalshi are cited as part of the broader ecosystem that could be reoriented toward hedging and risk management rather than speculative bets.

Tickers mentioned: $ETH

Sentiment: Neutral

Market context: The discussion sits at the intersection of onchain finance, risk management, and regulatory scrutiny, as investors and developers weigh how to apply AI tools to price hedging while navigating evolving policy debates around prediction markets.

Why it matters

The idea of coupling onchain prediction markets with AI-assisted personal finance tools signals a broader attempt to retrofit crypto-native mechanisms for real-world stability. If successful, the approach could reframe how individuals and businesses manage price risk—shifting from a speculative posture to a practical hedging framework that protects purchasing power in an inflationary fiat environment. Buterin’s proposal emphasizes a user-centric model in which private data about expenses informs a custom set of market instruments. That alignment between individual spending patterns and market-based hedges could, in theory, yield more predictable budgeting for everyday goods and services.

Critics of prediction markets often point to concerns about manipulation, liquidity distribution, and regulatory risk. But proponents argue that when linked to digital, onchain ledgers and AI-driven personalization, these markets can deliver more resilient price signals and a public-good function by aggregating diverse information. The debate touches broader questions about how decentralized finance should interact with traditional market dynamics and consumer protection standards. In this framing, the role of prediction markets extends beyond forecasting political events or commodity prices to becoming a probabilistic toolkit for household and business planning.

As the ecosystem evolves, the boundary between information services and financial instruments remains a focal point for policymakers and practitioners alike. The discussion around onchain prediction markets is part of a wider push to explore how AI can augment human decision-making in finance, risk assessment, and purchasing power. The outcome will hinge on how convincingly the model demonstrates real-world utility, addresses liquidity and governance challenges, and remains compliant with applicable rules in various jurisdictions.

What to watch next

• The publication of any whitepapers or technical notes detailing the proposed onchain prediction-market architecture and the role of local LLMs in personal expense modeling.
• Emerging experiments or pilot programs that test category-based price indices and category-specific prediction markets in real-world settings.
• Regulatory responses or clarifications around prediction markets and onchain hedging tools, particularly in jurisdictions weighing consumer protection and market integrity.
• Public discussions and research from academics and practitioners about the feasibility and governance of personalized prediction-market portfolios.
• Follow-up statements or interviews from Vitalik Buterin or affiliated teams that expand or refine the proposed framework.

Sources & verification

• Vitalik Buterin’s X post outlining concerns about prediction markets and the proposed shift to hedging mechanisms. Link: https://x.com/VitalikButerin/status/2022669570788487542
• Cointelegraph op-ed discussing onchain prediction markets and the integration of AI LLMs. Link: https://cointelegraph.com/opinion/blockchain-prediction-markets
• Cointelegraph coverage on prediction markets and information markets, including perspectives on market intelligence. Link: https://cointelegraph.com/news/prediction-markets-information-finance
• Cointelegraph coverage of academic perspectives on prediction markets, including comments from Harry Crane of Rutgers University. Link: https://cointelegraph.com/news/prediction-markets-polymarket-polls
• CFTC-related developments regarding proposals on prediction markets, cited in Cointelegraph coverage. Link: https://cointelegraph.com/news/cftc-withdraws-proposal-ban-sports-political-prediction-markets

Rethinking prediction markets as hedging tools with AI

Ethereum co-founder Vitalik Buterin has grown wary of how prediction markets are developing, warning they risk becoming short-term price betting engines rather than tools that support long-term infrastructure. In a post on X, he argued that the current trajectory shows “over-converging” focus on immediate price moves and speculative behavior. He called for a shift toward onchain prediction markets that serve as hedges against price exposure for consumers, rather than betting mechanisms that amplify fiat-driven volatility. The thrust of his critique centers on moving from pure price bets to broader markets that can stabilize expenditures over time. He suggested a framework that blends prediction markets with AI-driven tools to counter inflationary pressures faced by households and businesses alike. In essence, his stance positions prediction markets as potential risk-management primitives if redesigned with real-world spending in mind. He proposed a system in which price indices are constructed across major spending categories, with regional variations treated as distinct categories, and a dedicated prediction market for each.

Buterin elaborates a mechanism where each user—whether an individual or a business—operates a local AI model that understands that user’s expenses. This AI would curate a personalized basket of market shares, effectively representing “N” days of predicted future outlays. The intent is to offer a dynamic hedge against rising costs, allowing participants to hold a mix of assets to grow wealth while maintaining a safety net against inflation via tailored prediction-market positions.

Supporters of prediction markets argue they provide valuable information about global events and financial trajectories, potentially serving as a hedge against a variety of risks. They point to platforms such as Polymarket and Kalshi as examples of how publicly sourced probabilities can supplement traditional data sources. Academic voices, including Rutgers professor Harry Crane, contend that well-structured prediction markets can outpace conventional polls in forecasting accuracy and should be treated as a public good in principle, assuming robust governance and safeguards. Critics, however, worry about misuse, regulatory constraints, and the potential for manipulation if markets are driven by centralized or biased actors. The debate straddles both the philosophy of information markets and the practical design challenges of turning them into reliable hedges for everyday life.

Ultimately, the question is whether a hybrid system—combining onchain markets with AI personalization—can deliver tangible price stability without sacrificing liquidity or inviting abuse. If such a model proves viable, it could redefine how crypto-native financial instruments interact with the real economy, offering tools that help households and firms weather price fluctuations while contributing to a broader ecosystem that values data-driven risk management.

TLDR:

• Lightning Labs released open-source toolkit enabling AI agents to transact with bitcoin independently. 
• The L402 protocol allows AI systems to pay for services without requiring accounts or authentication. 
• Remote signer architecture separates private keys from agent operations to prevent security breaches. 
• Agents can now purchase data feeds and sell services autonomously using bitcoin through micropayments.
  

Lightning Labs has released an open-source toolkit that enables artificial intelligence agents to send and receive bitcoin payments independently through the Lightning Network.

The technology eliminates the need for human intervention, traditional accounts, or API authentication systems. This development represents a major advance toward autonomous machine commerce, where AI systems can directly purchase data, services, and computational resources without human oversight.

Automated Payment Infrastructure for AI Systems

The new toolkit addresses a critical limitation in current AI agent capabilities. While modern AI systems can write code, analyze information, and execute complex tasks, they cannot easily conduct financial transactions.

Traditional payment methods require human identity verification through credit cards, bank accounts, and regulated payment platforms.

These systems depend on personal documentation and manual approval processes that AI agents cannot navigate.

Lightning Labs explained that agents face a fundamental barrier despite their technical sophistication. The company stated that AI systems still struggle with payments despite being able to read documentation and call APIs effectively.

This gap exists because agents need to transact instantly and programmatically at massive scale, requirements incompatible with conventional financial infrastructure.

The solution centers on L402, a protocol built upon the HTTP 402 “Payment Required” status code. When an AI agent attempts to access paid content or services, the server responds with a Lightning invoice.

The agent pays this invoice and receives cryptographic proof of payment. This proof functions as an access credential, allowing the agent to retrieve the requested resource.

Lightning Labs introduced “lnget,” a command-line tool that automates the entire payment process. When an agent encounters paid content, lnget handles invoice payment in the background without requiring manual steps.

The tool supports multiple Lightning backend configurations, including direct connections to local nodes and encrypted tunnel access through Lightning Node Connect.

Michael Levin, head of product development at Lightning Labs, emphasized the toolkit allows agents to use bitcoin payments without mandatory identification or registration requirements.

Security Architecture and Commercial Applications

Security measures form a core component of the toolkit’s design. The recommended configuration uses a remote signer architecture that separates private key storage from payment operations.

The signing machine holds private keys offline while the agent machine executes transactions. This separation ensures that compromised agent systems cannot expose private keys.

The macaroon-based credential system enables fine-grained permission control. Developers can create credentials limited to specific functions such as payment-only or read-only access.

These bearer tokens can be further restricted without issuing new credentials. The system supports five preset security roles tailored to different agent functions.

On the server side, Aperture enables developers to convert standard APIs into pay-per-use services. This reverse proxy handles L402 protocol negotiation and supports dynamic pricing based on resource consumption.

Backend systems require no Lightning-specific modifications. The combination creates a complete commerce loop where one agent can host paid services while another consumes them.

The toolkit enables direct agent-to-agent transactions at scale. AI systems can now purchase premium data feeds, acquire computational resources, and sell services for bitcoin.

This infrastructure supports micropayments that would be economically unfeasible with traditional payment rails. Lightning Labs positions the technology as foundational infrastructure for an emerging machine economy where autonomous agents conduct billions of programmatic transactions.

Ethereum co-founder Vitalik Buterin said he is starting to “worry” about the direction of prediction markets and suggested that they shift to become marketplaces to hedge against price exposure risk for consumers.

Prediction markets are “over-converging” to “unhealthy” products that are focused on short-term price betting and speculative behavior as opposed to long-term building, Buterin said in an X post.

Instead, onchain prediction markets coupled with AI large-language models (LLMs) should become general hedging mechanisms to provide consumers with price stability for goods and services, Buterin said. He explained how this system would work:

“You have price indices on all major categories of goods and services that people buy, treating physical goods and services in different regions as different categories, and prediction markets on each category. 

Each user, individual or business, has a local LLM that understands that user’s expenses and offers the user a personalized basket of prediction market shares, representing ‘N’ days of that user’s expected future expenses,” he continued.

Individuals and businesses can hold a combination of assets to grow wealth and “personalized prediction market shares” to offset the rising cost of living created by fiat currency inflation, Buterin concluded.

Related: CFTC pulls Biden-era proposal to ban sports, political prediction markets

Prediction markets are useful market intelligence tools, supporters say

Prediction markets are crowdsourced intelligence platforms that can provide insight into global events and financial markets, while allowing individuals and businesses to hedge against a wide variety of risks, proponents of prediction markets say.

Prediction markets are more accurate than polls and should be treated as a public good, according to Harry Crane, a statistics professor at Rutgers University.