Hacker News
1John Ternus to become Apple CEO
- Apple announced Tim Cook will become executive chairman and John Ternus will become CEO effective September 1, 2026. Cook joined Apple in 1998 and became CEO in 2011, growing the company's market cap from $350B to $4T. Ternus joined in 2001 and currently serves as senior VP of Hardware Engineering.
Apple software quality decline
Users express frustration with Apple's software becoming worse over time, citing issues with macOS bugs, poor UI design, problematic apps like Finder and Xcode, and a general decline from the polished experience of earlier versions. Many hope new CEO Ternus can address these problems.Hopes for new leadership direction
Discussion centers on expectations that hardware-focused CEO John Ternus might bring the quality and attention to detail from Apple's successful hardware division to improve the company's struggling software and services, potentially reversing recent negative trends.Tim Cook's legacy assessment
Mixed evaluation of Cook's tenure, praising his business success (1000%+ market cap growth), operational excellence, and employee-friendly policies, while questioning whether his logistics-focused approach maintained Apple's innovative product vision and values.
2All phones sold in the EU to have replaceable batteries from 2027
- EU rules require smartphones and tablets sold in Europe to have user-replaceable batteries starting February 18, 2027. Users must be able to remove and replace batteries without specialized tools, with replacement batteries available for at least 5 years after last sale.
Exemption for high-performance batteries
The EU regulation includes an exemption for batteries that can maintain 80% capacity after 1000 cycles, which Apple likely meets. This has sparked debate about whether it's a reasonable standard or a loophole that undermines the regulation's intent.Trade-offs between repairability and design
Users debate whether replaceable batteries are worth potential compromises in phone thickness, water resistance, and structural integrity. Some argue modern phones can achieve both, while others prefer current sealed designs.Market demand versus regulatory intervention
Discussion centers on whether this regulation addresses a real problem or forces unwanted changes. Some argue the market would provide replaceable batteries if consumers truly wanted them, while others see regulation as necessary to counter manufacturer preferences.
3GitHub's fake star economy
- A peer-reviewed study found 6 million fake GitHub stars across 18,617 repositories, with stars selling for $0.03-$0.85 each on dozens of websites. VCs use star counts as funding signals, creating a shadow economy where manipulated repos show distinctive patterns like high zero-follower rates and low fork-to-star ratios compared to organic projects.
Stars are a poor quality metric
Developers debate whether GitHub stars are useful for evaluating libraries. Many prefer examining commit history, issue handling, and code quality over star counts, viewing stars as bookmarks rather than quality indicators.VCs using flawed investment signals
Discussion criticizes venture capitalists for making investment decisions based on GitHub stars and other easily manipulated metrics, comparing it to drafting NFL players based on Instagram followers rather than actual performance.Systematic gaming of metrics
Users highlight how fake star markets are part of a broader ecosystem where all meaningful metrics get gamed, from social media followers to top app lists, making authentic evaluation increasingly difficult.
4Kimi K2.6: Advancing open-source coding
- Moonshot AI has released Kimi K2.6, an open-source AI model with advanced coding, long-horizon execution, and agent swarm capabilities. The model shows significant improvements in complex programming tasks, can autonomously optimize software systems over extended periods, and coordinates up to 300 sub-agents simultaneously for comprehensive project execution across multiple formats and applications.
Chinese open source AI models vs US closed models
Discussion centers on the irony that China is leading in open-source AI development while the US focuses on closed models. Users debate whether this represents strategic competition, economic warfare, or simply practical business decisions.Kimi K2.6 performance and benchmarks
Users share technical evaluations showing Kimi K2.6 performing competitively with top models like Claude Opus in coding tasks, though mixed results on reasoning benchmarks. Performance varies significantly across different programming languages.Pelican riding bicycle SVG test
Ongoing tradition of testing new AI models with a creative SVG generation task. Users debate whether this test remains valuable or has become repetitive, with some defending it as a useful benchmark for creative capabilities.
5At long last, InfoWars is ours
- This appears to be a satirical announcement from "Bryce Tetraeder" of Global Tetrahedron claiming to have acquired InfoWars. The piece uses dark humor to mock conspiracy media, describing plans to transform InfoWars into an absurd platform of misinformation, scams, and psychological manipulation designed to exploit visitors for profit.
Legal uncertainty and approval process
The Onion's takeover of Infowars requires judicial approval and is structured as a licensing deal ($81,000/month) rather than outright purchase. Multiple commenters note the deal isn't final until a Texas judge signs off, with Alex Jones able to appeal any ruling.Difficulty distinguishing satire from reality
Users express amazement that Infowars content is so absurd it's hard to tell from parody. One commenter thought the current site was already an Onion joke. The transition may be barely noticeable given how outlandish the original content appears.Creative vision and public service aspect
Tim Heidecker will serve as creative director, initially parodying Alex Jones before transforming the site into experimental comedy. Commenters see this as both entertainment and public service, preventing Jones from continuing to cause harm while benefiting Sandy Hook families.
6Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving
- Qwen has released Qwen3.6-Max-Preview, an early version of their next proprietary AI model that shows significant improvements over Qwen3.6-Plus in coding capabilities, world knowledge, and instruction following. The model achieves top scores on six major coding benchmarks and is available through Alibaba Cloud Model Studio, though it's still under active development with further improvements expected.
Chinese vs Western AI models
Users debate the actual performance differences between models like GLM 5.1, Qwen, and Claude Opus beyond benchmarks. Many report Chinese models performing surprisingly well for coding tasks, with some users switching from Claude subscriptions entirely due to cost and performance considerations.Open source strategy and geopolitics
Discussion centers on Chinese companies' open-source AI strategy as potential state-sponsored competition against Western proprietary models, with concerns about long-term sustainability and whether this benefits or threatens global AI development.Model pricing and value proposition
Users compare costs across different AI services, with many finding Chinese models offer better value despite recent price increases. Discussion includes concerns about rate limits on premium Western services and the economics of model subscriptions versus API usage.
7Atlassian enables default data collection to train AI
Atlassian's default data collection for AI training
Atlassian is automatically opting all customers into AI training using their Confluence pages, Jira tickets, and other content. The opt-out setting doesn't exist yet and won't be available until May 2026, with enforcement starting August 2026. This affects even paid enterprise customers and includes highly sensitive business data.Product quality deterioration and technical issues
Users report widespread bugs across Atlassian products including broken search, UI glitches, outdated infrastructure, and non-functional new features. Former employees cite poor engineering talent, lack of focus, and minimal maintenance staffing as root causes of the quality decline.Rumors of Anthropic acquisition
Speculation that Anthropic may acquire Atlassian primarily for access to business training data. This has prompted some users to consider data poisoning efforts and raised concerns about privacy implications for companies using Atlassian products.
8NSA is using Anthropic's Mythos despite blacklist
- The NSA is using Anthropic's powerful Mythos Preview AI model despite the Pentagon classifying Anthropic as a "supply chain risk" and pursuing legal action against the company. This highlights tensions between cybersecurity needs and ongoing disputes over AI usage restrictions.
Government hypocrisy and contradictions
Users debate the administration's contradictory stance on Anthropic's AI models, with some arguing this demonstrates government hypocrisy while others note surveillance and intelligence gathering have long histories predating current politics.Marketing strategy and artificial scarcity
Discussion centers on whether Anthropic's restricted release of Mythos is a brilliant marketing tactic creating artificial scarcity to force government adoption, or genuine safety concerns about the model's cybersecurity capabilities.Surveillance state and civil liberties
Commenters express alarm about accelerating surveillance capabilities enabled by AI tools like Mythos, drawing connections to existing programs revealed by Snowden and concerns about pre-crime detection systems.
9Stop trying to engineer your way out of listening to people
- The software world often struggles with communication because people aren't truly listening to each other. Common listening pitfalls include assuming technical knowledge is uniform, projecting your own resources and capabilities onto others, treating people as static rather than evolving, and making judgments that prevent genuine understanding. Effective listening requires recognizing that people have different expertise, constraints, and perspectives that change over time. When teams fail to listen properly, they miss valuable insights that could drive success and minimize technical debt, as every misunderstanding creates future code complications.
Ai corrupting precise technical documentation
A developer's carefully written RPC documentation was passed through AI by their manager, resulting in completely fabricated endpoints and parameters that caused widespread confusion and wasted time for the receiving team.Communication styles and listening challenges
Discussion about how effective communication requires vulnerability, accepting that people may not want to accept new information, and focusing on understanding pain points rather than jumping to solutions.Technical vs non-technical communication barriers
Debate over whether engineers should bear the burden of improving communication skills, with some arguing that non-technical stakeholders should also learn to articulate their needs more effectively.
10Changes in the system prompt between Claude Opus 4.6 and 4.7
- Anthropic released Claude Opus 4.7 with updated system prompts showing several key changes: the developer platform is now called "Claude Platform," expanded child safety instructions, reduced verbosity requirements, new disordered eating guidelines, protection against simple yes/no responses to complex issues, and removal of specific language about Trump's presidency as the model's knowledge cutoff now includes January 2026.
Acting vs clarifying behavior
Users strongly prefer Claude to ask clarifying questions rather than making assumptions, but Anthropic added instructions for Claude to "make reasonable attempts" without interviewing users first. Many developers use custom prompts to force more clarification.System prompt bloat and restrictions
The 60,000+ word system prompt includes specific restrictions for eating disorders and other behaviors, leading to concerns about incremental censorship, token costs, and reduced functionality for legitimate use cases like coding.Agent-agnostic development approaches
Developers are creating neutral codebases that work across multiple AI providers (Claude, GPT, Gemini) using standardized file structures, skills, and hooks to avoid vendor lock-in as model capabilities fluctuate.