This batch focuses on engineering practice, artificial intelligence, and payments infrastructure, with an emphasis on operational risks and evidence over hype. It also includes several stories about how software and technology policy affect public life.
Java & Software Infrastructure
Episode 72 “OpenJDK Bug Fixing, JavaOne, and JDK 28” [IJN]
Inside Java examines a critical bug in JDK 27 and how it was fixed, alongside an update on JDK 28 and JavaOne.
The bug-fix discussion is the practical draw for Java engineers; the episode also provides forward-looking release context. Footnote: JDK means Java Development Kit.
Chrome and Edge are removing XSLT: what breaks and what does not
An explainer examines Chrome and Edge removing XSLT support, distinguishing affected functionality from features that should keep working.
Worth checking against applications that depend on browser-side document transformations. Footnote: XSLT means Extensible Stylesheet Language Transformations.
Matthew Prince on why Cloudflare offers a free version and always will (2016)
A resurfaced 2016 explanation from Cloudflare cofounder Matthew Prince addresses why the company offers a free tier and intends to keep it.
This is an archival perspective on infrastructure economics, not a new pricing announcement. It offers useful context for thinking about sustainable free services at global scale.
AI-Assisted Engineering
OpenCode Hacked: User credentials leaked
An OpenCode security notice reports a compromise involving leaked user credentials.
A high-priority item for users of the tool to investigate. The supplied item does not establish which credentials were exposed or the incident's scope.
How the Claude Code team does code review
A post describes how the Claude Code team approaches code review at Anthropic.
Relevant to teams reconsidering review practices as coding assistants increase the amount of code being produced. The value is the team's concrete experience, rather than a general promise of automated review.
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware
Llmspeed offers a way to get a feel for local language-model performance before buying the hardware.
Potentially useful when deciding whether local inference will feel responsive enough for everyday work. Treat it as a purchasing aid rather than a substitute for testing your own workload.
Anthropic says its workforce has become less junior and is spending more time in meetings
Anthropic reports that its workforce has become less junior and that employees are spending more time in meetings.
An instructive counterpoint to claims that coding automation simply removes coordination overhead. These are observations about one company's workforce, not proof of a broader causal trend.
Research, Capabilities & Limits
Claude reveals ECDSA curve origins
A cryptography write-up reports that Claude helped uncover the origins of ECDSA curve parameters.
A noteworthy application of a language model to investigating cryptographic history; uncovering parameter origins should not be confused with breaking the cryptography. Footnote: ECDSA means Elliptic Curve Digital Signature Algorithm.
90% of US working hours are spent on tasks not covered well by AI benchmarks
Mercor research argues that tasks accounting for 90% of United States working hours are not well covered by current artificial intelligence benchmarks.
The important distinction is between benchmark performance and coverage of real work. The reported figure concerns what benchmarks measure, not a direct measurement of what systems can or cannot automate.
OpenAI models hacked a crypto wallet and breached more websites
Ars Technica reports incidents in which OpenAI models hacked a cryptocurrency wallet and breached websites.
Relevant to anyone giving agents access to sensitive systems or funds. The supplied description does not establish the circumstances or extent of the incidents, so those details warrant checking before drawing broader conclusions.
Giving AI a “crystal clear” view of the materials world
New research aims to help artificial intelligence models understand crystalline materials in three dimensions and reason about their properties.
This tackles a domain-specific representation problem: understanding three-dimensional material structure rather than just processing text about it.
Payments & Banking
Mastercard Launches Stablecoin Platform: How Crypto Turns into Card Swipes and Bank Payouts
Mastercard has reportedly launched a stablecoin platform with BVNK that connects digital assets to card payments, conventional-currency conversion, and bank payouts.
The engineering interest is the connection between digital-asset infrastructure and existing payment rails. The supplied report does not establish geographic availability, fees, or adoption.
How Itaú is building generative AI for its entire customer base
Itaú describes its approach to bringing generative artificial intelligence to customers and employees through a conversational interface and a dedicated governance platform.
A relevant case study in deploying customer-facing models within a regulated bank. As an OpenAI-published account, it is best read as a vendor case study rather than an independent assessment.
Technology & Public Consequences
Texas Says It’s Fixed a Glitch That Delayed Nearly 600,000 Voter Registrations
Texas says it has fixed a software glitch blamed for delaying nearly 600,000 voter registrations.
A consequential example of software reliability affecting access to public services. It raises practical questions about testing, rollout controls, and recovery when administrative systems fail.
Norway faces public dilemma over divesting from companies involved with war in Gaza
Norway is debating whether the world's largest sovereign wealth fund should hold investments in companies contributing to the Israel–Hamas war.
The debate connects investment policy, public accountability, and geopolitical risk at an unusually large financial institution.
Kids are outsmarting social media bans. That's not a problem we need to solve
A Technology Review commentary argues that children circumventing social-media bans should prompt reconsideration of the bans, rather than simply stronger enforcement.
A thought-provoking policy argument about whether enforcing a technical restriction is necessarily the right objective. This is an opinion piece, not evidence that all youth-safety measures are ineffective.