marsbit Published on 2026-03-28 Last updated on 2026-03-28 The Wall Street Journal reveals the decade-long personal rift between Anthropic and OpenAI founders, Dario and Daniela Amodei, and OpenAI's Sam Altman and Greg Brockman. The conflict, rooted in philosophical differences over AI development and governance, began in 2016. Tensions escalated over leadership disputes, credit attribution, and management styles, culminating in the Amodeis and nearly a dozen employees leaving OpenAI in late 2020 to form Anthropic. Today, both companies are valued over $300 billion, competing fiercely while their founders' unresolved personal animosity continues to shape the global AI landscape. Wall Street Journal reporter Keach Hagey published a long-form investigative report, systematically disclosing for the first time, through extensive interviews with current and former employees and people close to executives at both companies, the decade-long personal feud between the founders of Anthropic and OpenAI. What shaped the global AI landscape was not just a battle over technological roadmaps, but also a never-healed personal wound. Dario Amodei's rhetoric internally in recent months has been far more intense than in public. He compared Sam Altman's legal dispute with Elon Musk to "Hitler vs. Stalin," called OpenAI President Greg Brockman's $2.5 million donation to a pro-Trump super PAC "evil," and likened OpenAI and other competitors to "tobacco companies selling products they know are harmful." After the Pentagon dispute escalated, he wrote on Slack calling OpenAI "mendacious," stating, "These facts indicate a pattern of behavior I have seen repeatedly in Sam Altman." Internally, Anthropic refers to this branding strategy as creating a "healthy alternative" to its competitor. An ad during this year's Super Bowl, which implicitly mocked OpenAI for embedding ads in its chatbot, is a public manifestation of this strategy. The story begins in the living room of a shared house on Delano Street in San Francisco in 2016. Dario and his sister Daniela Amodei lived there, and OpenAI co-founder Brockman often visited due to his personal friendship with Daniela. One day, Brockman, Dario, and Daniela's then-fiancé, effective altruism philanthropist Holden Karnofsky, sat together arguing about the right path for AI development: Brockman believed all Americans should be informed about what was happening at the AI frontier, while Dario and Karnofsky believed sensitive information should be reported to the government first, not broadcast to the public. This disagreement later became the philosophical dividing line between the two companies. Impressed by OpenAI's talent roster, Dario joined in mid-2016, staying up late with Brockman training AI agents to play video games. But over four years of working together, conflicts deepened around power and a sense of belonging. In 2017, Musk, OpenAI's main funder at the time, demanded a list of each employee's contributions and conducted layoffs based on it. About 10% to 20% of the roughly 60-person team were fired one by one. Dario saw this as cruel; one of those laid off later became an Anthropic co-founder. That same year, an ethics advisor hired by Dario proposed that OpenAI act as a coordinating entity between AI companies and the government. Brockman extrapolated from this the idea of "selling AGI to the nuclear powers on the UN Security Council." Dario considered this近乎叛国 (close to treasonous) and once considered resigning. After Musk exited in 2018, Altman took over leadership. He and Dario agreed that employees lacked confidence in the leadership of Brockman and Chief Scientist Ilya Sutskever. Dario stayed on the condition that the two would no longer be his supervisors, but soon discovered that Altman had simultaneously promised the latter two the authority to fire him—two contradictory promises. After the development of the GPT series began, the most intense conflict among executives erupted over who could work on the language model project. Dario, then research director, barred Brockman from involvement. Daniela, who co-led the project with Alec Radford, threatened to resign as lead. Radford's personal wishes were caught in a proxy war among the executives. Dario's seniority grew with the success of GPT-2 and GPT-3, but he felt Altman downplayed his contributions. He was angry when Brockman went on a podcast to discuss the OpenAI charter, feeling his greater contribution to the charter warranted an invitation; he was similarly displeased to learn that Brockman and Altman were meeting former President Obama but excluded him. The conflict came to a head in a confrontational meeting. Altman called the Amodei siblings into a conference room, accusing them of encouraging colleagues to submit negative feedback about him to the board. They denied it. Altman said the information came from another executive. Daniela immediately called that executive in to confront them, and the person said they knew nothing about it. Altman then denied having said that. A fierce argument ensued. In early 2020, Altman asked executives to write peer reviews for each other. Brockman wrote a strongly worded review accusing Daniela of abusing power and using bureaucratic processes to exclude dissenters; Altman previewed it and called it "tough but fair." Daniela rebutted it point by point. The argument escalated to the point where Brockman once proposed withdrawing the review. At the end of 2020, the team centered around Dario decided to leave, with Daniela leading negotiations with lawyers regarding their departure. Altman personally went to Dario's home to persuade him to stay. Dario proposed reporting directly only to the board and explicitly stated he could not work with Brockman. Before leaving, he wrote a long memo dividing AI companies into "market-oriented" and "public benefit-oriented" types, suggesting the ideal mix was 75% public benefit, 25% market. Weeks later, Dario, Daniela, and nearly a dozen employees left OpenAI to found Anthropic. Five years later, both companies are valued at over $300 billion and are racing to be the first to IPO. During the group photo at the closing of the AI summit in New Delhi this February, Indian Prime Minister Modi and the tech leaders present raised their hands high. Amodei and Altman chose not to participate, only awkwardly bumping elbows. Related Questions Q What was the core philosophical disagreement about AI development between Dario Amodei and Greg Brockman in 2016? A The core disagreement was about how to disseminate information about AI advancements. Greg Brockman believed the information should be broadcast to the entire American public, while Dario Amodei and Holden Karnofsky argued that sensitive information should be reported to the government first, not broadcast publicly. Q What internal brand strategy does Anthropic use to position itself against OpenAI? A Anthropic's internal brand strategy is to position itself as a 'healthy alternative' to its competitors, specifically OpenAI. This was exemplified by a Super Bowl ad that implicitly criticized OpenAI for placing ads in its chatbot. Q What major event involving Elon Musk at OpenAI did Dario Amodei view as 'cruel'? A Dario Amodei viewed Elon Musk's directive in 2017 to rank every OpenAI employee by contribution and subsequently lay off 10% to 20% of the roughly 60-person team as a 'cruel' event. Q What was the final ultimatum Dario Amodei gave to Sam Altman in an attempt to stay at OpenAI before leaving? A Dario Amodei's final ultimatum was that he would only stay at OpenAI if he reported directly to the board and made it clear that he could not work with Greg Brockman. Q How did Dario Amodei categorize AI companies in a memo written just before leaving OpenAI, and what was his proposed ideal ratio? A In his memo, Dario Amodei categorized AI companies into 'market-driven' and 'public benefit' types. He proposed that the ideal mix for a company should be 75% public benefit and 25% market-driven.

JPMorgan Research Report Interpretation: Global Rate Hikes Restart, Stock Markets Still Anchored by Corporate Earnings JPMorgan Research Report Interpretation: Global Rate Hikes Restart, Stock Markets Still Anchored by Corporate Earnings JPMorgan's research report, dated September 18, 2026, analyzes the restart of global interest rate hikes and its implications for equity markets. The report notes a shift towards synchronized monetary tightening among major developed market central banks, including the Fed, ECB, and Bank of Japan. The Fed's "insurance cuts" from late 2025 are being reversed, with JPMorgan forecasting a 25bps hike in December 2026 and a risk of a third hike in early 2027. The core thesis is that as long as the hiking cycle remains shallow, stock markets will be anchored by corporate earnings, with limited direct impact from rising rates. Historically, the S&P 500's P/E ratio can remain supported with 10-year yields near 6% if robust earnings growth (consensus forecasts over 20%) materializes. The immediate financial impact on corporates is gradual due to fixed-rate, long-term debt. Sector analysis shows Communication Services, Information Technology, and Energy outperform when rates rise, while Industrials and Real Estate are highly rate-sensitive and underperform. Large-cap stocks are relatively insulated compared to small-caps. JPMorgan reiterates a preference for large-cap, tech, and communication services stocks. Other key points include: oil (Brent) is unlikely to sustain above $100/barrel despite Middle East tensions; an upcoming US-China summit carries high symbolism but limited substantive progress is expected. JPMorgan maintains an overweight stance on global equities and emerging markets, viewing credit as the most resilient asset class to Fed hikes. The main risk is if inflation forces a shift from a shallow to a broad tightening cycle, potentially testing the earnings anchor for stocks. marsbit 55m ago marsbit 55m ago Galaxy Research Report: U.S. Stocks Moving Towards 24/7 Trading, SEC Begins Overhaul of the Foundational 'Ownership Ledger' Galaxy Research Report: U.S. Stocks Moving Towards 24/7 Trading, SEC Begins Overhaul of the Foundational 'Ownership Ledger' The SEC has proposed significant updates to transfer agent rules, explicitly allowing blockchain technology to maintain ownership records. This move addresses foundational infrastructure needed to support 24/7 trading in US markets. While rules from the 1970s/80s assumed physical certificates, the new proposal modernizes the system by permitting distributed ledger technology for official stock ledgers. A key question raised is whether digital wallet addresses could replace the current requirement to collect a holder's name and physical address. This shift could enhance privacy but complicates issues like beneficial ownership reporting when wallet clusters cross regulatory thresholds (e.g., 5% or 10% ownership). The proposal does not make securities law inapplicable to tokenized assets but seeks to reduce unnecessary personal data in centralized databases. These changes align with other steps toward 24/7 trading, including extended clearing hours and overnight volatility limits. The goal is to improve market efficiency and reduce reliance on privileged OTC access after hours, moving toward a more unified, competitive, and accessible market structure. OpenAI in the Robotics Field, Could It Actually Be OpenAI Itself? OpenAI in the Robotics Field, Could It Actually Be OpenAI Itself? OpenAI, originally known for language models like GPT, is emerging as a significant force in robotics—potentially becoming the "OpenAI of robotics" itself. Recent experiments showcase GPT-6 Astra acting as a robotic "brain." A $150 SO-101 robotic arm, guided by Astra, successfully drew the Golden Gate Bridge by planning strokes and using camera feedback. In another test, Astra controlled a dual-arm system to pick up a block and place it in a bowl with 95% success, outperforming other models. This challenges the conventional approach of training specialized foundation models for robotics from scratch, as pursued by companies like Physical Intelligence and Skild. Instead, Astra leverages its vast pre-existing knowledge of language, code, and vision to reason about physical tasks. It typically provides high-level commands (like end-effector pose), while traditional controllers handle low-level joint movements. However, limitations remain. Astra struggles with precise "last-millimeter" tasks like fitting a puzzle piece, where fine contact and force control are critical. Its inference speed also isn't yet suitable for real-time, low-level control. The future likely involves a hierarchy: models like Astra for planning and understanding, paired with dedicated low-level policies for execution. OpenAI, which once paused its robotics work due to data scarcity, is now revisiting the field. Sam Altman has confirmed OpenAI's interest in humanoid and other robot forms, with the AI "brain" being the key focus. The question is no longer just who will build the robot brain, but how the most advanced general AI models will integrate with capable physical bodies. For the First Time, GPT-6 Astra Cracks a 'Major Progress'-Level Challenge, Mathematicians Reduced to Prompt Engineers For the First Time, GPT-6 Astra Cracks a 'Major Progress'-Level Challenge, Mathematicians Reduced to Prompt Engineers For the first time, OpenAI's GPT-6 Astra, in collaboration with three human researchers, has solved a "Major Progress" level open problem on the FrontierMath benchmark. The challenge, open since 2017, was to find a counterexample in approval-based committee elections where the "core"—a concept from cooperative game theory guaranteeing a stable, fair committee—is empty. Instead of finding a counterexample, GPT-6 Astra proved the core is *never* empty, demonstrating that a perfectly fair committee always exists. The AI did not merely provide a non-constructive proof. It invented a novel voting rule based on a "harmonic entropy" objective function and provided a polynomial-time algorithm to compute the committee. The proof showed that a locally optimal solution under this harmonic entropy function is guaranteed to lie within the core. This elegant mathematical construct, described in a 20-page arXiv paper, effectively overturned the premise of the original problem. The researchers noted that this breakthrough was the result of intense, days-long collaborative iteration. While human mathematicians provided direction and logical rigor, the core conceptual leap—the harmonic entropy framework—came from GPT-6 Astra. This achievement prompted the FrontierMath organizers to create a new "Human + AI" solution tag, acknowledging the pivotal role of AI in deriving the core idea. The result marks a significant step for AI from being a problem-solving tool to becoming an active co-researcher capable of discovering new mathematical principles. Crypto Bill Falls, Fed Hikes Rates, Yet Bitcoin Rises? Crypto Bill Falls, Fed Hikes Rates, Yet Bitcoin Rises? "The 'Clarity Act,' a long-awaited crypto industry bill, failed in the U.S. Senate, while the Federal Reserve raised interest rates by 25 basis points. Surprisingly, Bitcoin's price rose by 6%, climbing back above $81,000 after initially dipping below $75,000. The market's resilience was further bolstered by an SEC order granting a five-year exemption, allowing tokenized U.S. stocks to trade via liquidity pools on public blockchains. Analysts note that Bitcoin now faces a significant resistance wall between $83,000 and $86,000, where approximately 1.07 million coins are held by long-term investors. The path forward hinges on three key indicators: Bitcoin sustaining a weekly close above $86,000, consistent large inflows into U.S. spot Bitcoin ETFs, and a decline in the 30-year Treasury yield below 5%. Currently, none of these conditions are met. The situation draws parallels to mid-2023 when regulatory optimism was eventually outweighed by macroeconomic pressure, leading to a significant price drop. However, the current market structure is different, supported by substantial ETF assets and stablecoin market capitalization. The debate continues between analysts who focus on increasing money supply (bullish for asset prices) and those focused on the high cost of capital (a bearish pressure). The coming weeks' ETF flow data will be crucial in determining whether this is the start of a sustained breakout or a temporary rally before another macro-driven decline." 市场分析 8,336 news items 项目动态 6,435 news items 监管政策 5,912 news items Политика регулирования 1,509 news items Analyse de marché 3,945 news items AI 3,463 news items Prediction 443 news items Cross-Domain 2 news items World Models 12 news items Foundation Models 2 news items