$80 Billion Hedge Fund Warns: Bear Market Risks Rising, AI Boom Could Trigger Chain Sell-Off

According to Bitget market data, Asian stock markets are generally lower. Specifically: South Korea’s KOSPI index extended its decline to 4%, standing at 5,624.02 points; Taiwan Stock Exchange Weighted Index fell 3% to 43,570.41 points; Indonesia’s benchmark index widened its intraday drop to 2%; the Nikkei 225 index is down 0.33%; and the Shanghai Composite Index is down 0.71%.

Latest data from prediction market platform Predict.fun shows the probability of the U.S. midterm elections resulting in the Democratic Party winning both the Senate and the House of Representatives is 77%. Mainstream polls also favor the Democrats: the average of Silver Bulletin’s national congressional vote polls as of November 2 shows the Democrats leading the Republicans by 8.9 percentage points; the Seldon forecast model gives the Democrats a 79% chance of winning the Senate and a 99% chance of winning the House. The probability of "the Republicans keeping the Senate while the Democrats take the House" is 19%, the second most likely outcome in the market. The Polling Forecast model notes that polls have overestimated the Democrats in 5 of the past 6 elections; if polls overestimate the Democrats by 3.5 percentage points this time, the Senate will still be controlled by the Republicans. The probability of "the Republicans keeping both the Senate and the House" is 5%; the probability of "the Democrats taking the Senate while the Republicans keep the House" is 1%. Current mainstream analysis projects the Democrats are most likely to gain a net 12 to 22 seats in the House, far exceeding the 3-seat threshold needed to retake the majority. Additionally, in the market related to "the Democrats retaking the House", Predict.fun puts the probability at 96%.

Beating AI News: Israeli AI startup 42AI exits stealth mode, launches "Behavioral World Model" for predicting individual decisions. The company said it has partnered with 10 major brokerages to help them anticipate which assets clients will trade, whether to buy or sell, and transaction size. The model analyzes not only historical transactions but also leverages client onboarding data, fund transfers, customer service call records, even call tone and contact frequency. It deduces clients' risk preferences and behavioral tendencies from this data, then combines market changes to predict their next potential actions. 42AI noted that 98% of actual transaction volumes fall within the model's predicted range. For trade direction and volume predictions, the F1 score reaches 0.8 to 0.85 (out of a maximum of 1). Such predictions do not need to cover every transaction; as long as they can anticipate clients' overall buy/sell direction and size in advance, brokerages can arrange liquidity, adjust margins, and pre-hedge. 42AI added that one partner brokerage generated around $30 million in additional value via dynamic hedging. The $62.5 million seed round was invested incrementally over three years by the founders and Infinity Group. Infinity Group was founded by 42AI's CEO's father, who also serves as the company's chairman.

Beating AI Express: In its updated IPO prospectus filed on November 2, Anthropic warned that artificial intelligence (AI) could soon cross a key threshold related to Recursive Self-Improvement (RSI), which would drastically accelerate AI development. The firm noted past cases where significant R&D acceleration was not detected until after it had already occurred. Currently, Claude is already involved in writing code, running experiments, and processing training data to help build stronger next-generation models, which in turn speed up subsequent R&D—creating a self-reinforcing cycle of accelerating progress. Under Anthropic’s safety policy, crossing this key threshold would mean one of two scenarios: 1) AI can complete the work of an entire AI research team at comparable cost; or 2) AI technological progress speeds up by at least double, cutting progress that would take two years to achieve down to one year. As AI becomes more deeply integrated into its own R&D, model capabilities may grow faster than updates to safety evaluations, alignment research, and protective measures. Anthropic has not yet identified any existing model that has reached this level; the latest warning is based primarily on internal estimates and projections for future models. The company is now considering revising its safety policies to ensure critical R&D processes stay under human supervision and control.

Decentralized exchange platform Extended has unveiled the tokenomics model for its native token XTEN, with a fixed total supply cap of 1 billion tokens, all minted at genesis and no inflation mechanism. Of the total, 30% (300 million tokens) is allocated to eligible participants of the Extended Points program, with full unlocking at the Token Generation Event (TGE) and no fees required. XTEN is primarily used for trading fee discounts, staking rewards, and Vault yield boosts. The staking function will launch concurrently with TGE: users who stake at least 500 XTEN will enter corresponding tiers, offering trading fee discounts up to 30% and Vault yield boosts up to 5x, with a maximum individual Vault boost deposit cap of $2 million. In token allocation, the ecosystem fund accounts for 20% (200 million tokens), including 150 million for staking rewards and 50 million for ecosystem development. The team, founders, and investors collectively hold 50% (500 million tokens), which will remain locked for the first 12 months post-TGE before being released in tranches per respective vesting schedules. Additionally, XTEN’s total supply will not increase: trading fee revenue may be used for token burning, and Extended’s affiliated entities may also repurchase XTEN on the open market. The token will also natively support Circle’s Cross-Chain Transfer Protocol (CCTP), enabling cross-chain transfers between Arc and other blockchains that support the protocol. Extended has not yet announced the TGE date or the airdrop claim window. Notably, users from the European Economic Area (EEA) will not be able to conduct XTEN spot trading during the platform’s initial launch phase, though they will still be eligible to claim, hold, stake, and transfer the token.

Morgan Stanley’s latest research report points out that global power demand growth is expanding beyond just AI data centers to three key drivers: AI, industrial relocation, and air conditioning and refrigeration. The power sector is entering a new "super cycle" trend that the market has not yet fully priced in. The firm raised its global power consumption forecast by roughly 10%, projecting global electricity usage will reach 36 trillion kilowatt-hours by 2030, with a compound annual growth rate (CAGR) of around 4.2%. Asia will contribute about 64% of global new electricity demand added between 2025 and 2030, with the largest upward revisions to power demand growth forecasts seen for India, the U.S., and Australia. On the supply side, Morgan Stanley expects fossil fuels will still account for roughly 50% of the global power mix by 2030, with reliable coal and gas-fired power generation set to command higher scarcity premiums. Meanwhile, wholesale power prices in major markets are projected to rise around 20% year-on-year in 2026, as pricing power in the power industry shifts from equipment manufacturers to power generation operators. The firm estimates that the expansion in power demand and rising electricity prices could create around $500 billion in value, and drive a roughly 14% upward revision to 2028 earnings per share (EPS) forecasts for related companies, a trend not yet fully reflected in stock markets.