Top Stories

Filing the Corners Off My MacBooks

1308 points · kentwalters.com

Still dominating the front page with over 600 comments, this post about literally filing down the sharp corners on MacBooks has clearly struck a nerve. What started as a simple hardware mod turned into one of the biggest HN discussions in weeks — covering industrial design philosophy, Apple’s aesthetic priorities versus ergonomics, and whether laptop makers have sacrificed comfort for visual minimalism. Plenty of commenters shared their own wrist pain stories from long coding sessions on sharp-edged aluminum. Sometimes the most relatable posts are the simplest ones.

Small Models Also Found the Vulnerabilities That Mythos Found

876 points · aisle.com

The top new story of the day challenges the narrative around large frontier models dominating cybersecurity research. This analysis shows that smaller, more accessible AI models were able to discover the same vulnerabilities that the much-hyped Mythos system found, raising important questions about the “jagged frontier” of AI capabilities. The 247-comment discussion dives deep into what this means for AI safety policy, open-source security tooling, and whether we’ve been overestimating the gap between large and small models in security-critical domains.

Show HN: Pardonned.com – A Searchable Database of US Pardons

386 points · pardonned.com

A Show HN project that built a searchable database of US presidential pardons. With 220 comments, the community discussion is lively — covering everything from the technical implementation to the political implications of making this data more accessible. The project makes it easy to search, filter, and explore the history of executive clemency, which has traditionally been scattered across various government sources. A good example of civic tech done right.

South Korea Introduces Universal Basic Mobile Data Access

339 points · theregister.com

South Korea is rolling out universal basic mobile data access for all citizens, a policy that treats connectivity as a fundamental right rather than a luxury. The 97-comment thread explores whether this model could work in other countries, the technical infrastructure required, and what it means for digital equity. South Korea continues to be a global leader in connectivity policy, and this move is likely to influence similar discussions worldwide.

Every Plane You See in the Sky – Follow It from the Cockpit in 3D

258 points · flight-viz.com

A stunning flight visualization tool that lets you pick any plane visible in the sky and follow it from a virtual cockpit view in 3D. The project pulls real-time flight data and renders it into an immersive browser-based experience. The 55-comment thread is full of appreciation for the technical execution, with pilots and aviation enthusiasts weighing in on the accuracy of the cockpit instruments and the flight model.

How We Broke Top AI Agent Benchmarks: And What Comes Next

251 points · rdi.berkeley.edu

Berkeley’s RDI group published a detailed post about how they systematically broke top AI agent benchmarks, exposing fundamental weaknesses in how we evaluate autonomous AI systems. The 68-comment discussion is a must-read for anyone working on AI evaluations — it covers benchmark contamination, the gap between benchmark performance and real-world reliability, and what a better evaluation framework might look like. As AI agents become more capable, getting the measurement right matters enormously.

Cirrus Labs to Join OpenAI

238 points · cirruslabs.org

Cirrus Labs announced they’re joining OpenAI, sparking a 120-comment discussion about the ongoing consolidation in the AI infrastructure space. Cirrus Labs has been known for their CI/CD and cloud infrastructure tools, so the acquisition signals OpenAI’s continued investment in developer tooling and infrastructure. The thread debates what this means for Cirrus Labs’ open-source projects and whether OpenAI is building out a full-stack developer platform.

The Future of Everything Is Lies, I Guess – Part 5: Annoyances

236 points · aphyr.com

Kyle Kingsbury (aphyr) continues his series on the frustrations of living in an increasingly AI-generated information landscape. This installment focuses on the everyday annoyances — from AI slop in search results to generated content polluting documentation and forums. The 133-comment thread resonates strongly with developers who are feeling the quality degradation firsthand. Aphyr’s writing is characteristically sharp and well-argued, making this a cathartic read for anyone tired of the noise.