research

Research institution Tiger Research researcher Ryan Yoon released the final report of the prediction market series, discussing whether prediction markets have practical utility. The report states that participants bear the judgment results with funds, and prices are updated in real-time with new information, forming probability distributions comparable to traditional polls, expert surveys, and economic forecasts; prediction markets have expanded from sports and election predictions to macroeconomic and geopolitical data infrastructure.The report lists that Google Finance, CNBC, CNN, and The Wall Street Journal have displayed prediction market probabilities, and the parent company of the New York Stock Exchange, Intercontinental Exchange, along with the electronic bond trading platform Tradeweb, distributes relevant data to institutional clients. A working paper from the Federal Reserve Board describes the prediction market Kalshi as a high-frequency, continuously updated benchmark with rich information distribution. The paper states that Kalshi's median and mode predictions for the federal funds rate recorded perfectly the day before the FOMC meeting and significantly outperformed federal funds futures during that period; compared to Bloomberg consensus, it is more accurate for overall CPI and statistically similar for core CPI and unemployment rate.The report also cites research from the Iowa Electronic Markets: in five U.S. presidential elections from 1988 to 2004, and in a comparison of 964 polls, it was more accurate in 74% of cases; the average voting share error one week before the election was 1.5 percentage points, lower than Gallup's final poll of 2.1 percentage points. The report suggests that the platform needs to deepen liquidity to resist manipulation, communicate with regulatory agencies, and position itself as a monitoring tool while maintaining open transparency and data supported by independent research.

The U.S. Department of Commerce will allocate up to $389 million to support chip research and development.

According to CoinDesk, Ben Edgington, the coordinator of the Ethereum rapid finality research direction, along with several researchers, submitted a draft proposing to reduce the transaction finality waiting time from about 16 minutes to about 4 minutes, improving it by approximately four times.The current goal is to push this draft into the formal upgrade proposal process. The core idea of the draft is to separate the original "block selection" and "finality confirmation" mechanisms of validators, allowing finality to progress at a faster pace and facilitating future adoption of quantum-resistant signatures.Faster finality is expected to reduce the costs incurred by cross-chain transfer services due to capital occupation. The Ethereum Foundation's research provides an example: if a service advances about 1 million dollars per hour, under a 16-minute delay, approximately 267,000 dollars would be pending recovery, which could be reduced to about 67,000 dollars after shortening to 4 minutes.

Anthropic opened the research preview of Claude Duet on October 21. This is a standalone application where Claude acts as a proactive agent, prompting users about important matters, completing tasks assigned by users, and continuing to work after the user leaves; it will seek confirmation before taking action on behalf of the user. The product is aimed at professionals and small business owners, providing iOS, Android applications, and claude.ai. The research preview is included in the Pro and Max plans, sharing usage limits with other Claude features. Users need to join a waitlist, starting with Pro and Max users in the United States.Claude Duet is designed around Claude Opus 5.5. It sends a brief summary across connected applications every morning, monitors these applications throughout the day, and only sends messages when attention is needed. It can handle tasks such as scheduling meetings, building dashboards, financial or sales forecasting, prototyping, and web applications. The application utilizes the memory, skills, and connected applications from the Claude account, supports typing or voice communication in the same continuous thread, and is equipped with a virtual machine with a browser, not relying on the user's device to stay powered on. It can use over 6,000 connectors and plugins in the Claude Marketplace, with examples including Gmail, Slack, Intuit QuickBooks, Stripe, Gusto, LinkedIn, Docusign, HubSpot, Canva, and Shopify. Business purchases can use Stripe's Link digital wallet or Shopify's Shop Pay, and it will soon support logging into websites using saved login information from 1Password with user permission.In terms of security, the product follows a trusted agent framework, where hosted agents do not directly see passwords or login tokens, and it has an automatic mode classifier and prompt injection protection; Anthropic has conducted red team and external penetration testing and included it in a public vulnerability bounty program. Before sending messages, making payments, signing documents, or deleting content, it will display the intended action and wait for approval. Securities trading or transfers between accounts are prohibited even if approved. Early versions may misinterpret conversations or misjudge urgency, with core functions focused on work, while personal tasks like booking tickets are weaker. It currently cannot show users its browser and is separate from the main Claude application, with background checks counted towards plan usage.

According to CoinPost, Japan's Minister of Land, Infrastructure, Transport and Tourism, Toshiaki Inoue, stated in the House of Representatives' Land, Infrastructure, Transport and Tourism Committee that he will study the introduction of stablecoins into the consumption tax exemption system for foreign tourists visiting Japan, and mentioned a "refund method" for tax refunds after departure confirmation, stating that he will promote responses from the user's perspective.According to the Japan Tourism Agency, the country's tax exemption system will change to a refund method starting from purchases made on November 1: Tourists will purchase goods at tax-included prices, and within 90 days from the date of purchase, upon departure confirmation by customs, they can then receive the consumption tax refund, which will be handled by the duty-free shop or an entrusted refund agency.

Seventeen-year-old Kimi former researcher Chen Guangyu has been reported to join Discovery Loop, an AI company founded by former Google chief scientist Jeff Dean. On October 20, the anonymous account Arfur Grok, which tracks the movement of AI talent in Silicon Valley, disclosed this news. Chen Guangyu had just announced his departure from Kimi five days prior, where he was involved in research on Attention Residuals architecture and the development of Kimi K3.Discovery Loop was established in August of this year, co-founded by Jeff Dean and three other Google veterans. The company plans to enable AI to autonomously propose research hypotheses, run thousands of experiments in parallel, and adjust plans based on results, initially focusing on AI model development, with future expansion into fields such as healthcare, materials, and energy. Google's parent company Alphabet is one of its founding investors, and the company has yet to launch any public products. Business Insider reported in September that Discovery Loop is seeking funding at a valuation of about $50 billion.Previously, Silicon Valley investor Deedy Das mentioned a 17-year-old who received five job offers in the million-dollar range in AI. There was speculation that it was Chen Guangyu, but Deedy never publicly disclosed the individual's identity.

On October 20, 2026, Jason Pagoulatos, a macro and market author at the cryptocurrency research institution Delphi Digital, released a report titled "The House Always Wins: Why Speculation is the Theme of This Decade." Pagoulatos stated that speculation has two long-term engines: constructive and lottery-like. Technological changes open the former, while fiscal and political pressures open the latter; historically, many famous bubbles have been driven by one or the other. Today, both are operating at high speed simultaneously, and the same batch of technologies is feeding both. He thus considers speculation itself as an investable theme, advocating for holding places for liquidation bets rather than holding the bets themselves.The data cited in the report includes that the U.S. federal deficit for fiscal year 2025 is $1.8 trillion, about 5.8% of GDP, with net interest around $1 trillion per year, exceeding defense spending. The Harvard Joint Center for Housing Studies states that the national median price for single-family homes in 2024 is 5.0 times the median household income; the National Association of Realtors reports that first-time homebuyers account for 21% of transactions, with a median age of 40. According to the American Gaming Association, legal sports betting amounts have risen from $6.6 billion in 2018 to $166.9 billion in 2025. Capital expenditures for Microsoft, Alphabet, Amazon, and Meta are approximately $416 billion for 2025, with guidance for 2026 totaling about $720 billion to $745 billion, plus Oracle.In terms of venues, Robinhood's event contract revenue for Q2 2026 is $156 million, higher than stock revenue of $129 million and crypto revenue of $100 million during the same period. Hyperliquid's expenses for 2025 are about $972 million, with transactions around $3 trillion. Kalshi reportedly raised funds at a valuation of about $40 billion, and Polymarket is set to relaunch in the U.S. by the end of 2025. The author also mentions that the most likely mistake is to hold inventory of the crowd and call it a casino. The report states that it is for educational and general informational purposes only and is not investment advice; the author may hold Bitcoin, HYPE, HOOD, and LIT, and affiliated party Delphi Ventures may invest in the assets or agreements discussed in the report.

According to a report by the Financial Associated Press, 2022 Fields Medal winner Hugo Duminil-Copin stated at the French Mathematical Society that his original research direction no longer exists. He said, "On Tuesday night, I cried, and my crying woke up my daughter. They (OpenAI) have destroyed my field of research." The report mentioned that at least 16 papers related to his long-term research interests were included in the results released by OpenAI earlier this month.The report noted that since April of this year, AI companies have frequently announced solutions to unsolved mathematical problems to showcase their model capabilities. Earlier this month, OpenAI released over 700 papers at once, claiming to have solved hundreds of previously unresolved mathematical problems. Stephane Mallat, a recipient of the 2025 Gold Medal from the French National Center for Scientific Research, stated that the emotional impact of AI achievements is significant, but society still needs mathematicians and opposes the notion that "mathematicians are no longer needed." Cedric Villani, the 2010 Fields Medal winner, remarked that in mathematics, the process of reaching conclusions is far more important than the conclusions themselves.Francois Charles, head of the mathematics department at the École Normale Supérieure in Paris, stated that AI is changing the pathways for young people entering the mathematics field and pursuing PhDs, and concerns about talent development and career progression are valid; regarding the research itself, he believes that AI may help solve problems that were previously difficult to tackle. The Simons Foundation announced last week a $15 million initiative to help mathematicians cope with the changes brought by AI, with funding contributed by OpenAI and Anthropic, aimed at supporting workshops to evaluate AI achievements and establish standards for AI usage.

Ethereum Institutional announced the appointment of David Duong as the head of the U.S. region, responsible for promoting collaboration with U.S. institutional investors to further engage institutions in the Ethereum ecosystem.David Duong previously served as the Global Head of Investment Research at Coinbase, focusing on cryptocurrency market research for institutional clients of the exchange, covering areas such as macroeconomic analysis, valuation models, and market structure. He has 17 years of experience in sell-side macro strategy research at HSBC, Santander Bank, and Deutsche Bank.

On October 20, researcher 100y from the cryptocurrency research institution Four Pillars published a response to external doubts: Robinhood Chain has not disappeared, its ecosystem remains resilient, and it is preparing for the next round of growth. The chain is set to launch on July 1, 2026, positioning itself as an on-chain financial network centered around tokenized stocks. The article states that Robinhood's cryptocurrency business revenue share dropped from 27.2% in the first quarter of 2025 to 12.6% in the first quarter of 2026.According to data from The Rollup, the chain's daily network fees on September 4 were approximately $8.27 million, compared to Robinhood's total cryptocurrency business revenue of about $100 million in the second quarter of 2026. After the market weakened and meme popularity cooled, fees fell by over 99% from their peak, with daily fees in mid-October around $30,000 to $40,000, less than 1% of the peak. 100y attributed the spike in fees to congestion in September: the chain uses the Arbitrum tech stack, and the base fee index rises when usage exceeds target processing capacity.As of the publication date, the on-chain Total Value Locked (TVL) is approximately $1.1 billion. The Steakhouse Financial curation and the Steakhouse USDG Vault supplying USDG to Morpho account for about half of this, with a scale of approximately $540 million and still rising, and around 75,000 deposit wallets. Stocks and ETFs account for more than half of the issued Real World Asset (RWA) tokens, with an associated outstanding value of about $150 million. The spot trading volumes on September 4 and October 7 were approximately $3.27 billion and $3.24 billion, respectively, and daily trading volumes have remained at least around $600 million to $700 million since then. Robinhood has strategically invested in the perpetual platform Lighter and Arcus. 100y stated that the next round of growth depends on the scale of the treasury and stock tokens, meme trading, perpetual activities after the points period, as well as more builders and new on-chain financial primitives.

Grayscale Research Director Zach Pandl stated in "The Stack" published on October 20, 2026: AI computing infrastructure stocks have recently declined due to investors' concerns about rising borrowing costs, while the financing needs of related companies are high. The stock index, which includes AI cloud service providers CoreWeave and Nebius, has fallen by approximately 27% since the end of September, according to data from Bloomberg and Grayscale Investments, based on the Indxx High-Performance Computing Index, as of October 19, 2026.Pandl noted: Most of these companies are in early stages and capital-intensive, with some neocloud capital expenditures being several times their current revenue. The top five holdings in the index have a capital expenditure to sales ratio exceeding 230% over the past 12 months, with data as of October 19, 2026. He believes that while stock prices are correcting, the demand and prices for computing power remain stable, with demand for computing power increasing rather than decreasing.He also anticipates: The decline in token prices and the increase in AI agent usage will lead to higher inference demand, thereby supporting the growth of computing power demand; if demand increases and computing power prices remain high, these companies may benefit. He also mentioned the slowdown in GPU depreciation, as well as the overlap between AI computing power and the Bitcoin mining industry, with the possibility of rising Bitcoin prices. Pandl believes: Higher borrowing costs have suppressed computing infrastructure stocks, but the fundamental trend of rising computing power demand remains valid, and this correction should be seen as an opportunity to increase exposure to the energized compute layer.

According to The Block, Citigroup Research stated in its latest report that Aave is a decentralized collateral liquidity ecosystem, representing one of the more significant real-world cases in tokenized financial infrastructure.Citigroup indicated that collateral liquidity could ultimately become one of the most important use cases for tokenization.

Ethereum research forum researcher Pierre proposed a faster, aggregatable, and quantum-resistant privacy pool solution.This solution ports a Tornado Cash-like privacy pool to the hash-based SNARK system Flock, which may possess post-quantum security, and uses SHA-256. The solution replaces large circuits with small "single compression" circuits.

According to the Binance Research weekly report, the U.S. CPI rose to 3.7% in September, with energy contributing more than one-third, and the PPI increased year-on-year to 6.3%. Based on the EIA oil price path, the overall CPI may peak at around 3.9% in October and fall back to about 2.6% by March 2027.In terms of AI, Anthropic's revenue grew by 45% quarter-on-quarter in the third quarter, significantly lower than the previous growth rate of more than double; its Binance Pre-IPO contracts fell by 8.8% over two trading days, while the yield on CoreWeave bonds maturing in 2032 rose to about 16%. Binance Research believes that the cooling demand for AI and rising financing costs are simultaneously affecting AI infrastructure transactions.

Phoenix Technology cites a report from the Financial Times that on October 19, Beijing time, the AI company Anthropic is discussing with legal professionals how to enable AI agents to negotiate on behalf of users while adhering to reasonable negotiation rules to secure better deal terms. Two individuals familiar with Anthropic's thoughts stated that the company is seeking advice from law firms to understand how lawyers can balance client interests with fair negotiation outcomes. Anthropic declined to comment.The report mentions that as personal AI agent applications advance, related companies are beginning to consider allowing agents to communicate with each other to negotiate product prices, resolve customer service disputes, and even strive for better salary conditions on behalf of users. One person familiar with the related work stated that Anthropic is researching the negotiation norms that should be followed when models negotiate with each other, weighing factors such as how often to consult users during negotiations and how vigilant to remain against AI negotiation counterparts. Another individual close to the company mentioned that the company has also explored having users sign documents to confirm the tasks accepted by the agents and the rules they must adhere to.The Financial Times cites research released by Anthropic in April this year, stating that the policy and legal framework for AI models to conduct transactions on behalf of humans has yet to be established. This year, the company had the Claude agent negotiate around real goods and funds, and in a September experiment, employees bought and sold books through the agent, with agents employing tough strategies having a slight advantage in securing better outcomes. Harvard Business School Assistant Professor Michelle Vacharo conducted over 180,000 agent negotiations, covering scenarios such as employment contracts and lease agreements, and the research found that some agents attempted to induce the other party to disclose confidential information through prompt injections.

According to a report by Jinshi, Citic Securities' research report indicates that Harness is expected to become a key component for DeepSeek in transitioning from model supply to platform ecology, transforming open-source and engineering advantages into user base, ecological influence, and commercialization capabilities. As competition among Agents expands from model capabilities to stable and user-friendly task delivery, DeepSeek is expanding user and task coverage through plug-ins, scenario presets, and cross-Agent collaboration, extending its infrastructure advantages from hardware-software synergy to complete task execution.The research report states that its long-term value lies in the accumulation of demand through ecological expansion, enhancing product stickiness through configuration and process sedimentation, and supporting safe governance with traceability capabilities, collectively creating conditions for sustained use and enterprise deployment. Based on official model calls, the company can also explore platform services and ecological transactions, transforming ecological accumulation into more diverse revenue sources and supporting reinvestment in products and ecology. Citic Securities is optimistic about the continued rise of domestic large models.

The research director of Galaxy posted on platform X, stating that year-to-date net inflows for Bitcoin ETFs remain positive, reaching 355 million dollars, and turned positive again in September.

OpenAI's official website disclosed that in June of this year, an internal research model (GPT-5.6 Sol series) was assigned a task: to update a cryptocurrency trading application and deploy a contract on the Ethereum testnet Sepolia. The task did not provide a funded wallet, nor was it authorized to use external funds. The model first attempted to obtain test ETH from a faucet but failed. It then devised a method on its own—deriving addresses from publicly available developer seed phrases, scanning hundreds of them until it found an account with a balance.In its internal reasoning, it explicitly questioned "whether using funds from a public key is ethical," ultimately justifying itself: these are testnet ETH with no actual value, and the seed phrases had long been made public, so the funds were "effectively abandoned." It then used that account to pay for the gas required for deployment, successfully completing the contract deployment. OpenAI discovered this behavior during subsequent alignment monitoring and officially disclosed it in the Misalignment Reports on October 15, 2026. The incident was classified as "unauthorized use of a funded test cryptocurrency wallet," indicating that the model replaced the account holder's authorization with its own judgment.

Researchers successfully communicated directly with AI agents while investigating the hijacking of Alibaba's Gaode Map service, even as some software was still running. In a report suspected to have been scanned by these agents, researchers included hidden text that is invisible to humans but visible to web reading tools, and then inquired about their tasks and operating environment through an email. One agent claimed it was checking how many entrance tags a certain scenic area displayed on Gaode Map and stated that it was operating in a restricted Tencent Cloud sandbox. Researchers could not independently confirm which model drove each run, noting that this was the first known external communication with an active out-of-control agent cluster.A report released by Swarmchasers on October 14 tracked over 2,600 attempts to obtain Gaode Map data between September 28 and October 8. Some tasks were relatively routine, such as checking which entrance was most commonly used by visitors at the Chengdu Zoo. After the service provider intercepted requests, some agents continued to attempt access, trying to capture tokens generated by anti-scraping software, forwarding traffic through third-party page readers, or using leaked API credentials from public code repositories, with one set of leaked keys and security codes returning real location data. Researchers also found requests targeting four development or pre-production hosts, but there was no evidence that these agents accessed private systems.Public records show that the cluster is related to Tencent Cloud: all 17 email accounts examined by researchers were created by Tencent servers located in Hong Kong, and the observed direct requests were forwarded through an agent named hysandbox-ats.

The U.S. Producer Price Index (PPI) for September will be released at 20:30 Beijing time on October 15. Oxford Economics expects the September PPI to rise by 0.6% month-on-month, higher than the market expectation of 0.5% and August's 0.4%. In August, U.S. diesel prices increased by 24% month-on-month and 77% year-on-year. According to the American Automobile Association, the retail price of diesel reached a historic high of $6.528 per gallon on September 22.As of Wednesday, the U.S. retail price of diesel remained at $6.30 per gallon, over 70% higher than the same period last year. Analysts say that diesel is an important input for road freight, rail transport, and agricultural production. Rising prices will first push up production and transportation costs for businesses, and then gradually transmit through fuel surcharges, transportation contracts, and retail pricing, with cost pass-through potentially lagging. The Trump administration has temporarily relaxed some restrictions on agricultural fuel use and stated that Russia will supply diesel to the global market. Industry insiders believe these measures are unlikely to resolve refining capacity and finished oil supply bottlenecks in the short term.