Indepth Research

Provide in-depth research reports and independent analysis, leveraging data, technology, and economic insights to deliver a comprehensive examination of the blockchain ecosystem, project potential, and market trends.

GPUs Have No 'Price': Four Indices Clash, the Compute Market Is More Chaotic Than You Think

The GPU compute market lacks a clear price benchmark, with four major indices on Bloomberg Terminal showing significant divergence—differing by over $2 and moving inconsistently. This reveals a fragmented and inefficient market for AI compute power, particularly for H100 and B200 GPUs. While demand surges and supply constraints are real, pricing is highly opaque and varies by contract type, provider, and location. The absence of standardized contracts, reliable benchmarks, and liquid secondary markets leads to hoarding and informal subleasing. Key issues include inconsistent pricing methodologies, unstandardized agreements, quality assurance gaps, and no forward pricing mechanism. Without coordinated improvements in infrastructure and market design, establishing a true price for GPU compute remains impossible.

marsbit04/08 09:24

GPUs Have No 'Price': Four Indices Clash, the Compute Market Is More Chaotic Than You Think

marsbit04/08 09:24

Polymarket: 84% of Traders Are Losing Money, 0.033% of People Take the Majority of Profits

A recent analysis of 2.5 million Polymarket wallet addresses by on-chain researcher Andrey Sergeenkov reveals that 84.1% of traders are losing money, with only 16% achieving any profit. A mere 2% of addresses have accumulated over $1,000 in profits, while just 0.033% (840 addresses) have earned more than $100,000. The study, which improved on prior methodologies by accounting for token splits and mergers, shows a significant increase in unprofitable traders compared to a previous 70% estimate. Profit concentration is extreme: only 1.25% of addresses make over $1,000 monthly, 0.26% earn above $5,000, and just 0.13% surpass $10,000. Most high-earning addresses are short-lived, with 53% of top traders active for only one month. Research indicates that automated strategies—such as arbitrage bots and high-frequency trading systems—dominate profits, exploiting public on-chain data and speed advantages that manual retail traders lack. Despite Polymarket’s accuracy in prediction (94% a month before outcomes), the platform exhibits a stark wealth disparity, raising questions about its function as an information aggregator versus a zero-sum game favoring sophisticated players.

marsbit04/08 03:49

Polymarket: 84% of Traders Are Losing Money, 0.033% of People Take the Majority of Profits

marsbit04/08 03:49

A Detailed Explanation of Tempo Chain and MPP Machine Payments Protocol

The global payment system is undergoing structural transformation, driven by the explosive growth of stablecoins and the rise of the AI agent economy. This creates a pressing need for next-generation payment infrastructure. AI agents have five core payment requirements: autonomy, micro-payments, high frequency, interoperability, and atomic settlement. Tempo, a payment-native blockchain by Commonware, addresses these needs. It features the Simplex BFT pipelined consensus for sub-second finality, dedicated block space, a stablecoin-native gas mechanism, and the Machine Payments Protocol (MPP) for end-to-end autonomous payments. Its technical architecture is payment-optimized. Key innovations include: - **Simplex BFT Consensus:** A pipelined design reducing confirmation latency to one network round-trip (1Δ). - **BLS Aggregate Signatures:** Minimizes bandwidth and computational overhead. - **Parallel Transaction Execution:** Enabled by custom EIP-2718 transaction types and an expiring nonce system. - **Dedicated Payment Lanes:** Protocol-reserved block space to shield payments from network congestion. - **Stablecoin-Native Design:** Stablecoins are first-class citizens for gas and on-chain exchange. MPP, co-designed with Stripe, is an open standard like "OAuth for payments." It enables AI agents to pay autonomously via a standardized HTTP challenge-response flow. Its core innovation is a session mechanism for efficient, continuous resource consumption without per-action on-chain confirmations. MPP is payment-rail agnostic, supporting various networks like Tempo, Stripe, and Lightning. Application scenarios include cross-border corporate settlements, 24/7 tokenized deposit clearing, commercially viable micro-payments, and autonomous AI agent transactions. Compared to rivals like Circle's Arc and Stable, Tempo differentiates through its EVM compatibility and Stripe partnership. Versus general-purpose chains like Ethereum L2s and Solana, Tempo's advantage lies in its payment-semantic native design, not just superior performance. The success of autonomous AI payments hinges on resolving regulatory uncertainty around agent identity and compliance. Tempo's core contribution is rethinking payment infrastructure at the protocol level, focusing on precision in payment semantics, pluggable compliance, and agent authorization models.

marsbit04/07 13:31

A Detailed Explanation of Tempo Chain and MPP Machine Payments Protocol

marsbit04/07 13:31

Everyone Is Waiting for the War to End, But Does the Oil Price Suggest a Long-Term Conflict?

This article argues that the market views oil price volatility as a consequence of the ongoing conflict, but the real insight is understanding how the war itself is being priced through oil. With the Strait of Hormuz blocked, global crude supply is being restructured, with Asian buyers shifting to US oil, causing WTI to surpass Brent—a sign of structural changes in pricing and trade flows. While short-term price differences can be explained by contract timing, the deeper issue is a fundamental shift in who can supply oil. The market's key misjudgment is not on price, but on time. Futures curves still imply the conflict will end soon and supply will recover. However, the more likely path is a prolonged war of attrition. This means high oil prices are not a temporary shock but a new structural reality, with a range of $120-$150. In this framework, oil is no longer just a commodity but the "upstream variable" for all assets. Its repricing will ripple through interest rates, currencies, equity, and credit markets. The market has priced in the war's occurrence, but not its persistence. The analysis concludes that a prolonged conflict is the base case scenario, and any pullbacks in oil prices present an opportunity as the market has yet to fully price in a long-term disruption.

marsbit04/07 07:30

Everyone Is Waiting for the War to End, But Does the Oil Price Suggest a Long-Term Conflict?

marsbit04/07 07:30

Xianyu, the OpenClaw More Suited for Chinese Users

The article "Xianyu: The OpenClaw More Suited for Chinese Users" argues that while OpenClaw (a hypothetical AI agent) has generated significant buzz for enabling AI to perform digital tasks, the Chinese second-hand marketplace Xianyu serves as a more practical and "localized" alternative for many. The author contrasts OpenClaw's high computational costs (token consumption) and complex setup with Xianyu's extremely low-cost, human-powered services. On Xianyu, users can find people to perform a vast array of tasks—from booking hotels and buying tickets to AI subscriptions, hiring help, and even matchmaking—often at a fraction of the cost of using an AI. The platform is portrayed as a massive, flexible ecosystem where unconventional supply meets niche demand, creating a parallel economy with over 200 million monthly active users. The piece draws parallels between the two platforms: OpenClaw acts as a leverage tool using AI and tokens, while Xianyu leverages cheap and abundant human labor. Both systems facilitate a form of arbitrage—OpenClaw with digital tokens, and Xianyu through bilateral transactions between buyers and sellers. Ultimately, the article suggests that for China, with its specific socio-economic context and different attitude towards privacy, Xianyu's human-network-based model is a more effective and accessible "agent" platform than an AI-driven solution like OpenClaw.

marsbit04/07 06:12

Without KPIs, Can They Still Lead OpenAI? An 'Internal Observer' Perspective on How Top AI Labs Operate

The article explores the unique organizational and research strategies of top AI labs like DeepMind, OpenAI, and Anthropic, based on an interview with Sebastian Mallaby, author of a biography on DeepMind’s Demis Hassabis. Mallaby highlights DeepMind’s hybrid model of "free exploration" and focused "strike teams" that enabled breakthroughs like AlphaGo and AlphaFold. He attributes DeepMind’s success to Hassabis’s vision, competitive drive, and ability to balance long-term research with practical execution, supported by Google’s resources. The discussion also contrasts the labs’ approaches to safety, scalability, and AGI development, noting differences in risk tolerance—with OpenAI being more aggressive, Anthropic more cautious, and DeepMind居中. The piece concludes that AI progress relies on both scaling compute and algorithmic innovation, with the future of AGI depending on sustained research investment and strategic leadership.

marsbit04/07 05:27

Without KPIs, Can They Still Lead OpenAI? An 'Internal Observer' Perspective on How Top AI Labs Operate

marsbit04/07 05:27

The Small-Town Youth Labeling AI Giants

In China's hinterland cities like Datong, Shanxi, thousands of young people are working as data annotators—the invisible workforce behind AI development. They perform repetitive tasks like drawing bounding boxes on images or rating AI-generated responses, earning piece-rate wages as low as a few cents per task. These workers, mostly from rural areas or small towns, endure intense labor conditions: strict monitoring, high error tolerance thresholds, and mental exhaustion. Despite the cognitive nature of their work, they are often paid meager salaries, with some earning as little as ¥30 ($4) for a day’s work. As AI industry evolves, even highly educated workers—including master’s graduates—are being drawn into similar precarious freelance roles, evaluating complex AI outputs under vague and shifting standards. Yet the industry is structured through layers of outsourcing, where most profits flow to tech giants like OpenAI and Microsoft, while annotators see dwindling incomes. Worse, as AI models become more self-sufficient, the demand for human annotators is declining. Companies like Li Auto have slashed annotation costs by using AI-powered tools that complete in hours what used to take humans years. These annotators, who helped train the very systems now replacing them, face an uncertain future—a stark contrast to the booming valuations and optimistic narratives of the global AI industry. No one seems to see a problem with any of this.

marsbit04/07 04:37

marsbit04/07 04:37

The New Yorker In-Depth Investigation Analysis: Why Do OpenAI Insiders Believe Altman Is Untrustworthy?

"The New Yorker investigation, based on internal documents and interviews with over 100 sources, reveals deep internal distrust in OpenAI’s leadership, particularly toward CEO Sam Altman. Key allegations include a pattern of dishonesty, undermining safety protocols, and prioritizing commercial interests over OpenAI’s original non-profit mission to develop AI safely. Chief Scientist Ilya Sutskever compiled a 70-page dossier accusing Altman of repeatedly lying to the board—for instance, falsely claiming GPT-4 features had passed safety reviews. Anthropic co-founder Dario Amodei’s private notes further detail how Microsoft’s investment deal effectively neutered OpenAI’s safety commitments. The report also highlights unfulfilled promises, such as allocating only 1-2% of promised computing resources to critical safety teams. Internal conflicts extend to CFO Sarah Friar, who opposed Altman’s aggressive IPO timeline amid financial concerns. Microsoft executives compared Altman to fraudsters like SBF, citing a tendency to distort facts and renege on agreements. Critics argue that Altman’s unchecked authority and alleged disregard for transparency pose significant risks given OpenAI’s powerful, potentially dangerous AI technology. The company’s transformation from a safety-first non-profit to a profit-driven entity raises fundamental questions about its governance and ethical commitments."

marsbit04/07 03:40

The New Yorker In-Depth Investigation Analysis: Why Do OpenAI Insiders Believe Altman Is Untrustworthy?

marsbit04/07 03:40

Research: $35 Trillion in Annual Stablecoin Transaction Volume, How Much is Real Payment?

A joint report by McKinsey and Artemis reveals that only about 1% of the annual $35 trillion stablecoin transaction volume represents real payments, totaling approximately $390 billion. Among these, 58% are B2B transactions—such as cross-border supplier settlements and financial operations—which grew by 733% year-over-year. Consumer usage, including card payments and remittances, remains negligible. The analysis identifies five structural reasons for this institutional dominance: 1) Greater financial efficiency gains for businesses compared to consumers; 2) Programmability benefits B2B workflows but lacks retail applications; 3) Regulatory frameworks favor institutional adoption; 4) Closed-loop B2B systems avoid the chicken-and-egg problem of consumer-merchant networks; 5) Corporate incentives prioritize internal efficiency over downstream consumer adoption. While B2B stablecoin payments are growing rapidly, consumer adoption may remain limited without breakthroughs in usability and regulatory clarity. The report suggests stablecoins may evolve primarily as an institutional settlement layer rather than a mainstream consumer payment tool.

marsbit04/07 03:09

Research: $35 Trillion in Annual Stablecoin Transaction Volume, How Much is Real Payment?

marsbit04/07 03:09

70-Page Confidential Document's First Allegation: 'Lying', Altman Told the Board 'I Can't Change My Character'

In a major investigation, Pulitzer winner Ronan Farrow and Andrew Marantz reveal two previously undisclosed documents: a ~70-page confidential file compiled by former OpenAI chief scientist Ilya Sutskever and over 200 pages of internal notes by Anthropic CEO Dario Amodei from his time at OpenAI. Sutskever’s file, which opens with the accusation that Sam Altman exhibited a "pattern of lying," alleges he misled executives and the board on safety protocols and corporate matters. Amodei’s notes similarly claim “the problem at OpenAI is Sam himself,” citing instances like Altman denying agreed-upon terms in Microsoft’s $1 billion deal. Key revelations include: - No written report was produced from the post-reinstatement independent investigation into Altman. - OpenAI’s superalignment team received only 1-2% of the promised computing resources, mostly on outdated clusters. - In 2018, executives considered a "National Plan" to auction AI tech to nations including China and Russia. - Microsoft executives expressed strong distrust toward Altman, with one comparing his risk profile to figures like Bernie Madoff. During a board call after his firing, Altman reportedly said, "I can’t change my personality," which a director interpreted as an admission of persistent dishonesty. Altman denies intentional deception, attributing his behavior to "well-intentioned adaptation" and conflict avoidance.

marsbit04/06 14:24

70-Page Confidential Document's First Allegation: 'Lying', Altman Told the Board 'I Can't Change My Character'