Article URL: https://www.science.org/content/article/north-korean-nuclear-test-sets-years-earthquakes Comments URL: https://news.ycombinator.com/item?id=49755160 Points: 5 # Comments: 0
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
A team running Backboard CLI on Terminal-Bench 2.1 achieved a score of 85.4% using Claude Opus 4.8 via Amazon Bedrock, outperforming Claude Code's published score of 78.9% on the identical underlying model. The result suggests that the system architecture surrounding an AI model — including context management, tool selection, and failure handling — can significantly influence benchmark performance. Beyond accuracy, the Backboard run cost $280.72, roughly 49% less than the then-verified leaderboard leader, which scored 83.8% at a reported $552.67. The team argues this shifts the key question from which model performs best to which system delivers reliable results at sustainable cost. They also emphasize that well-engineered AI infrastructure should remain adaptable across models, reducing the need to rebuild entire stacks whenever a newer model emerges.
The rapid rise of AI and blockchain technology is intensifying global debates around cybersecurity, data ownership, and privacy rights. A cyberattack on crypto infrastructure firm Haruko exposed operational security vulnerabilities affecting 15 institutional clients, highlighting risks beyond blockchain's core cryptographic protections. Internal discussions among Microsoft employees and an ongoing New York Times lawsuit against OpenAI and Microsoft raise serious ethical and legal questions about whether AI training on scraped data constitutes large-scale intellectual property theft. Meanwhile, the U.S. Department of Homeland Security's predictive policing initiatives have drawn criticism over constitutional concerns related to algorithmic surveillance and civil liberties. Analysts argue these developments are not isolated incidents but reflect deeper structural challenges in how data is generated, owned, and governed in the digital age.
A security audit of Venus Core Pool, a DeFi lending protocol with approximately $1.28 billion in total value locked, has identified nine distinct governance attack vectors and assigned an overall risk rating of 7 out of 10. The review, dated September 18, 2026, focused on the on-chain governance flow covering proposal creation, voting, execution, and contract upgrades. Among the most critical findings are a flash-loan-driven voting power inflation attack scored 9/10 and a timelock bypass via re-entrancy scored 8/10, both considered technically feasible. Auditors noted that the governance subsystem had not been thoroughly examined since the v2.3 upgrade in March 2025, leaving key administrative functions exposed. The report attributes the elevated risk to a combination of a large treasury, powerful admin controls, and a relatively permissive quorum model that could be exploited by a coordinated or well-funded adversary.
A developer has created an AI agent capable of auditing an AWS account for security and cost issues, identifying problems such as publicly exposed S3 buckets, open SSH ports, disabled GuardDuty, and wasted spend on idle Elastic IPs. The entire audit takes roughly two minutes to complete. The agent is deliberately restricted to a read-only IAM identity, meaning AWS itself blocks any write, delete, or modify actions before they can execute. The build relies on just one JSON configuration file and a Markdown checklist, requiring no Python frameworks or vector databases. The full source code has been published on GitHub for others to deploy against their own AWS accounts.
Discussion (0)
Log in to join the discussion and vote.
Log in