Analysis: 30-year US Treasury yields hit their highest level since 2002, Bitcoin faces dual signals

ARK Invest CEO Cathie Wood said that as AI Agents evolve from answering questions to performing tasks and making actual purchases on behalf of users, investors may soon need to shift from "tracking developers" to "tracking AI Agents" to monitor which software, services, and financial networks they choose. Wood told the Robinhood Summit, "We may increasingly talk about 'tracking agents'." She believes that if millions of AI Agents can independently choose which tools and services to use in the future, their actual activities could become a new metric for tracking tech demand and capital flows. This trend is also driving the development of "Agentic Finance." SharpLink Co-CEO and former BlackRock Digital Assets head Joseph Chalom said AI Agents need payment permissions that are controllable, revocable, and traceable. For example, a user could authorize an agent to spend up to $500 on hotel bookings, while retaining the ability to revoke permissions and view transaction history at any time. Chalom noted that open blockchains like Ethereum could serve as shared financial networks for AI Agents, applications, and enterprises, enabling AI Agents to complete payments without a single bank or tech company acting as an intermediary. BlackRock has previously stated that AI Agents may use stablecoins to pay for API calls, data, and computing power in the future. Coinbase CEO Brian Armstrong previously said Grok is already one of the main clients for "agent traders" on Coinbase, though he did not disclose specific trading data. Meanwhile, companies including Stripe, Visa, Google, and OpenAI are developing AI Agent payment solutions, accelerating competition between traditional payment networks and blockchains.

According to Iranian media reports, Iran is reviewing a new U.S. proposal delivered via Qatar, which lays out a roadmap and timeline for final negotiations. The proposal is based on the Islamabad Framework but seeks to advance nuclear negotiations at an earlier stage. However, Iran continues to insist that the U.S. first fulfill its commitments, including sanctions relief and the release of frozen funds. Iran’s seven key conditions remain unchanged, and progress has been achieved in talks with Oman on a new shipping route through the Strait of Hormuz.

U.S. President Donald Trump has once again pledged that if the Republican Party wins control of both the House of Representatives and the Senate in the midterm elections, he will issue a $5,000 "dividend" to every adult U.S. citizen. The plan, which Trump has proposed before, remains a campaign promise for now. When reporters asked whether the funding would require congressional approval, Trump said: "Well, we don’t think so. If it were necessary, they would certainly do it, but we don’t think it’s needed. However, we have one prerequisite: the funds must be spent domestically in the U.S. — we don’t want the money flowing to other countries. We believe this is a very good measure, widely popular, and will inject strong momentum into the economy." Calculated based on roughly 240 million adult U.S. citizens, the plan is projected to cost around $1.2 trillion, making it one of the largest economic stimulus packages the U.S. has rolled out since the COVID-19 pandemic.

Bitcoin offline key tool EntropyLab has released its self-contained Bitcoin key and wallet calculation tool v1.0.0rc1, the project’s first release candidate. The tool runs entirely from a single HTML file, requires no internet connectivity, and does not generate random entropy on its own—users must supply their own entropy values to compute Bitcoin keys and wallets in a fully offline environment. It supports features including multi-signature, PSBT, BIP-85, Silent Payments, vanity addresses, read-only wallet exports, and nonce analysis. EntropyLab states that this release is reproducible and verifiable via signatures from four independent signers. The team advises users to verify the downloaded file before transferring it to an air-gapped computer for use. The official v1.0.0 version has not yet been released, as the project is currently testing this release candidate.

Beating AI News: Google’s AI coding tool Antigravity has integrated Claude Opus 5.5 and Sonnet 5.5. The two models are available on Google AI Ultra and Pro, but Pro access is limited to paid subscribers—users still in free trials cannot use them. Free and Google AI Plus users do not have access to the 5.5 models for now. Antigravity supports three thinking intensity levels for models: Low, Medium, and High. Google is also clearing out older models, with Claude Opus 4.6, Sonnet 4.6, and GPT-OSS-120b all set to be removed on November 2. Per the current model lineup, if no new models are added to replace them, Free and Google AI Plus users will have no Claude or GPT-OSS options left by then, leaving only the Gemini series. Google did not disclose specific usage limits for Claude 5.5, only noting that Pro tier limits refresh every 5 hours and are subject to weekly caps. Gemini uses separate limits from Claude and GPT. Multiple Pro users have reported that a single task on Opus 5.5 can quickly deplete both the 5-hour and weekly limits, with some users saying roughly 9 minutes of operation used half their weekly cap.

NVIDIA CEO Jensen Huang held approximately 870 million NVIDIA shares as of March 23, 2026, accounting for 3.58% of the company’s total outstanding shares. Valued at NVIDIA’s early-October share price of $231.49, this stake is worth roughly $201 billion. Forbes data shows Huang’s current net worth is around $200 billion, with his personal wealth primarily derived from NVIDIA holdings. Huang is NVIDIA’s largest known individual shareholder. However, under U.S. SEC rules, individual investors holding less than 5% of shares are not required to disclose their stakes, so it is possible that undisclosed individual shareholders hold more than Huang. NVIDIA’s major institutional shareholders holding over 5% of its shares include BlackRock and Vanguard, with stakes of 7.43% and 7.31% respectively. Huang co-founded NVIDIA in 1993. Before NVIDIA’s first stock split in 1999, he held around 2.9 million shares, representing 9.9% of the company. Had he retained all those shares to date, adjusted for stock splits, he would hold nearly 2.8 billion shares, indicating he has since continuously sold off some of his holdings. Additionally, Huang can earn NVIDIA stock through performance awards. NVIDIA paid him a total compensation of $35.5 million in fiscal 2025, consisting of a $1.5 million base salary, a $3 million cash bonus, and approximately $31 million in equity awarded in the form of 936,800 performance stock units (PSUs).