Claude Opus 4.7

Anthropic released Claude Opus 4.7, positioning it as a direct upgrade to Opus 4.6 with particular gains in advanced software engineering. The company describes it as capable of handling "the hardest coding tasks that previously required greater supervision," with improvements to instruction-following, multimodal support, file-system-based memory, and vision resolution. The release includes a new "X high effort" level, which causes the model to generate more output tokens—particularly on later turns in agentic (multi-step, automated) settings. Anthropic notes a tokenizer change that increases token consumption by roughly 1 to 1.035x depending on context, a practical cost consideration for enterprise API users. Benchmark comparisons place Opus 4.7 ahead of Opus 4.6, GPT-5.4, and Gemini 3.1 Pro on agentic coding tasks, though behind Anthropic's own Claude Mythos Preview on broader capability. Mythos Preview remains the more powerful system but is restricted to select businesses due to its demonstrated ability to identify and exploit vulnerabilities across major operating systems and browsers. Anthropic explicitly experimented with "differentially reducing" Mythos's cyber capabilities during training—a notable disclosure about deliberate capability suppression as a safety measure. Opus 4.7 ships with automated safeguards to detect and block high-risk cybersecurity requests. The company frames this deployment as a learning exercise toward eventual broad release of Mythos-class models. Anthropic has also established a more predictable two-month release cadence for the Opus line: 4.5, 4.6, and now 4.7 have each arrived roughly eight weeks apart, compared to a three-month gap between 4.1 and 4.5. Claude Code's Auto mode is now available to Max Plan subscribers, and a new "Ultra review" command runs a dedicated code review session flagging issues a careful human reviewer would catch.

Simultaneously, Anthropic has quietly introduced government ID verification for certain Claude users. The policy, which went live April 14, requires a physical passport, driver's license, or national identity card—photocopies and mobile IDs are excluded—along with a potential live selfie. Anthropic selected Persona, a know-your-customer (KYC) infrastructure provider used across financial services, as its verification partner. Anthropic characterizes the measure as targeted: triggered by "activity that indicates potentially fraudulent or abusive behavior" rather than applied universally. The company draws a technical distinction between data custody and control—ID data goes to Persona's servers, not Anthropic's, with Anthropic acting as "data controller" setting terms. The data is described as encrypted, excluded from model training, and not shared with third parties for marketing. The rollout has generated significant backlash, particularly given that Anthropic attracted a surge of privacy-conscious users in February after declining a Pentagon classified-network contract that OpenAI accepted. Daily signups reportedly broke records at that time, with free users up 60% since January. Critics, including commentary circulating on X, note that the verification requirement is not mandated by any regulation—it is a voluntary business decision. The discussion raises the broader question of whether AI identity controls will become an industry norm ahead of formal legal requirements. The security risk of third-party ID custody is also flagged: a breach at Discord in October 2025 exposed approximately 70,000 government IDs submitted for age verification, illustrating that even reputable KYC vendors carry custody risk.

TSMC reported Q1 revenue growth of 35.1% year-on-year and net income growth of 58.3%, both above analyst estimates. More consequential for the AI investment thesis: the company's CEO confirmed that TSMC directly checked with customers—including Nvidia and Apple—about AI demand durability amid an ongoing Middle East conflict (described as the "Iran war"), and received reassurance sufficient to raise guidance. TSMC now projects capital expenditures at the high end of its $52–$56 billion range for the year and expects full-year revenue growth exceeding 30% in dollar terms, up from an earlier "around 30%" estimate. Gross profit margin reached 66% in Q1, described as the highest level in more than 20 years. The company is accelerating investments including construction of additional clean rooms. Supply chain risk remains a live concern. TSMC's Vice President called on Taiwan to increase strategic reserves of helium, hydrogen, and natural gas—all critical to chip fabrication. Qatar is a significant helium supplier, and the company noted that specialty gases are sourced from multiple regions. Near-term production disruptions are not anticipated, with LNG supplies secured through at least May.

A new Alaska securities filing—required for holders of 5% or more—reveals that Google LLC held a 6.11% stake in SpaceX at the end of 2025. Following SpaceX's February merger with xAI, Bloomberg calculates Google's stake has diluted to approximately 5%, which would be worth $100 billion at a $2 trillion IPO valuation. SpaceX has filed confidentially to go public targeting a June listing, with reports of up to $75 billion in fundraising in what would be the largest IPO on record. Google first invested in SpaceX in 2015 alongside Fidelity in a $1 billion round that valued the company at $10 billion. Elon Musk holds approximately 40% of SpaceX. A successful $2 trillion listing would, per analyst commentary, deliver "career-defining returns" for investors who entered as recently as 2021, with estimated 20x returns from that vintage. X (formerly Twitter) separately launched Cashtags, a feature displaying real-time stock and crypto price data in user timelines, beginning with iOS users in the US and Canada. X will not execute trades but intends to build financial data tools; Canadian users see a direct link to brokerage Wealthsimple. The feature aligns with Musk's stated ambition to develop X into a super app, alongside the planned April launch of X Money, a peer-to-peer digital wallet. --- **Key takeaways:** - Anthropic is shipping Opus model upgrades on a reliable two-month cadence, with Opus 4.7 optimized for autonomous coding tasks, but enterprise users should budget for higher token consumption due to tokenizer and effort-level changes. - Anthropic's voluntary rollout of government ID verification represents a significant policy shift that contradicts its recent positioning as a privacy-first alternative—and sets a potential industry precedent ahead of formal regulation. - TSMC's raised guidance and record margins constitute the clearest recent signal that AI infrastructure spending has not softened, even amid geopolitical disruption; energy and specialty gas supply chains remain the primary operational risk. - Google's ~5% SpaceX stake, worth an estimated $100 billion at IPO valuation, illustrates how early-stage infrastructure bets in the 2010s are now generating returns that dwarf core business lines for major tech firms. - X's Cashtag launch and X Money rollout represent incremental but concrete steps toward financial services integration, with the platform explicitly positioning itself as a destination for finance and crypto communities rather than a regulated brokerage.