Top HN Daily Digest · Fri, May 22, 2026

The Trump administration has issued a new policy requiring most green card seekers to leave the United States and apply through consulates abroad, limiting "adjustment of status" within the country to only extraordinary circumstances. [src]

Commenters argue that requiring green card seekers to apply from abroad is a "malevolent" disruption of the only practical path for skilled workers, who cannot realistically maintain US employment while waiting years for uncertain consular processing [0][3]. While some note that leaving a country to renew a visa is common internationally [9], others contend this policy is a cynical attempt to restrict legal immigration by reinterpreting "adjustment of status" as a discretionary grace rather than a standard procedure [2][7]. The discussion highlights a deep frustration with the complexity and perceived hostility of the legal system, with users warning that these hurdles threaten the US technological lead and may inadvertently turn legal residents into "illegal" immigrants [5][6][8].

Anna’s Archive has published an `llms.txt` file inviting AI models to access its data via bulk downloads and APIs rather than scraping, while requesting donations to support its mission of preserving and providing open access to human knowledge. [src]

The discussion centers on Anna’s Archive (AA) appealing to Large Language Models for donations, sparking a debate over the ethics and definition of "ownership" regarding pirated content [0][1]. While some users argue that AA cannot claim the data as "theirs" since it belongs to the original authors and publishers [1][4][8], others contend that "possession" is a valid linguistic interpretation of the word [9] or argue that intellectual property is a flawed concept to begin with [3]. Many commenters defend the service as a necessary response to aggressive DRM, high academic costs, and market failures, noting that piracy often becomes the preferred option when legal alternatives are more difficult to use [2][5][6][7].

Japanese companies are highly diversified because their "J-firm" structure—characterized by lifetime employment and horizontal coordination—prioritizes long-term survival and employee retention over shareholder profit, allowing them to master niche, high-precision industrial sectors through decades of incremental refinement. [src]

The discussion highlights a divide between Westerners, who often idealize Japan’s corporate diversification as a product of "mastery" and lower executive pay [3][4][6], and East Asian observers who attribute the model to rigid classism, collectivism, and a "zombie company" problem [0][9]. While some argue the structure is a survival mechanism for lifetime employment [2], others suggest Western fascination is a result of decades of cultural familiarity and media-driven biases [1][5]. Ultimately, commenters disagree on whether these business models represent a superior alternative or a stagnant system built on vertical hierarchies and outdated workflows [0][7][8].

Apple cofounder Steve Wozniak received applause at Grand Valley State University's graduation after joking that students possess "actual intelligence" (AI), contrasting with other tech executives who were recently booed for promoting artificial intelligence during commencement speeches. [src]

Steve Wozniak’s commencement speech was praised for its human-centric approach, contrasting sharply with Eric Schmidt’s recent address which many viewed as resentful, chiding, or a form of propaganda intended to make a specific corporate future seem inevitable [0][5][9]. While some commenters argue that Wozniak is merely telling students what they want to hear to soothe economic angst, others contend that human value should be based on sentience rather than mere economic utility [1][2]. However, skepticism remains regarding the "race to the bottom" in software quality and the limited agency young people actually have in shaping the trajectory of AI [4][8].

The yt-dlp maintainers have deprecated and limited Bun support to versions 1.2.11 through 1.3.14, citing security concerns with older releases and stability risks following Bun's recent AI-assisted Rust rewrite. Support may be dropped entirely if maintenance becomes too burdensome. [src]

The decision to deprecate Bun support has sparked a debate over whether the move is a pragmatic engineering choice or an ideological reaction to "vibe coding" and AI involvement [0][2][7]. Critics argue the decision is speculative and "hysterical" since no specific technical regressions have been identified in the new Rust rewrite [5][8][9], while supporters contend that maintainers cannot responsibly support a million-line codebase generated by AI that they did not directly write [3]. Ultimately, some users view the move as a valid political or ideological stance against the direction of Bun's development, even if it prioritizes "feelings" over current performance data [1][4][6].

DeepSeek has announced that the current 75% discount on its V4 Pro model API pricing will become permanent following the scheduled end of the promotion in May 2026. [src]

DeepSeek’s permanent price reduction and extremely low caching costs have sparked debate over whether the company is pursuing a "long game" to bankrupt US competitors through non-viable unit economics [2][8]. While users praise the model's performance and "chains of thought" in coding tasks, some express significant privacy concerns regarding potential data leaks or state-level surveillance by the Chinese government [3][6]. Despite these concerns, many are integrating DeepSeek models into third-party tools like Claude Code, noting that the combination is highly effective and significantly cheaper than using domestic alternatives [0][1][4].

Microsoft has begun canceling licenses for Claude Code, an AI-powered coding tool, following its recent decision to discontinue the experimental project. [src]

Microsoft reportedly canceled Claude Code licenses after developers overwhelmingly preferred it over GitHub Copilot, frustrating management's goal of validating their own product [0][5]. While some users find Claude's performance declining or its high token consumption financially unpredictable for enterprise use, others argue that its superior capabilities justify the cost, especially when compared to cheaper alternatives like DeepSeek [1][3][4][8]. A significant point of contention is the pressure on developers to prioritize speed over token costs, leading many to stick with the most effective models to protect their job security [2][7].

After a 42-day journey across 12 countries, a Congolese refugee in Uganda successfully received a donated MacBook for his computer science studies despite facing bureaucratic hurdles, customs seizures, and a final delivery pickup at a local hardware store. [src]

The discussion highlights the extreme friction and corruption inherent in shipping goods to developing nations, where government taxes and bribery often stifle progress [0][7]. While some argue that Western aid can inadvertently entrench these "shady practices" or that outside interference should be avoided entirely [3][6], others suggest that a "Western approach" fails to account for how small amounts of money could have bypassed local bureaucratic pain [5][9]. Ultimately, commenters debate whether these systemic issues stem from a lack of government foresight regarding economic friction or a deeper lack of local agency and accountability [4][8].

Anthropic’s Project Glasswing has identified over 10,000 high-severity vulnerabilities in critical software using its Claude Mythos Preview model. While the AI significantly accelerates bug discovery, the initiative highlights a growing bottleneck in human capacity to verify, disclose, and patch these flaws before they can be exploited. [src]

The discussion reveals a sharp divide between users who find AI security tools "essential" and highly accurate [0][6] and skeptics who argue the technology produces false positives or offers only marginal improvements over existing scanners [3][8]. While some see a "step change" in vulnerability discovery that could eventually automate the entire development lifecycle [0][4][6], others point to high-profile cases like *curl* to suggest the tools may not yet outperform traditional methods in well-scrutinized codebases [3][9]. Ultimately, the debate centers on whether these models represent a fundamental shift in risk management or are simply an expensive, token-intensive evolution of the standard software "OODA loop" [1][5].

Paul Graham argues that a 1% wealth tax is mathematically equivalent to a 20% income tax based on a 5% rate of return, suggesting that politicians underestimate the significant financial impact of wealth tax proposals. [src]

Critics argue that Paul Graham’s conversion math is flawed because it assumes income is derived solely from wealth, ignoring the reality that a wealth tax would not affect laborers with zero savings [0][1]. While some commenters highlight the destructive potential of wealth taxes on illiquid assets like homes [3], others suggest these issues could be mitigated through progressive brackets that exempt the middle class [7]. Ultimately, the debate centers on whether such a tax is a necessary tool to curb runaway inequality or an unfair "grab" targeting the most productive members of society [4][5][6].