Students who use AI generally score worse at school
The OECD's PISA report, based on 2025 data, finds that students who use AI to study tend to perform worse in science, math, and reading. However, students taught to critically assess AI tools showed a slight improvement. The Verge reports on the complex relationship between AI use and academic performance.
Related Tech News
AI CuratedOpenAI's breakthrough solution for the elusive Navier-Stokes problem overshadowed by plagiarism controversy — researcher says OpenAI scraped Codex session and issued career threats
OpenAI announced a breakthrough solution to the Navier-Stokes equations, but researchers Tristan Buckmaster and Levent Alpöge accuse OpenAI of plagiarizing their work on Euler's equations. They allege OpenAI used their Codex sessions without consent and issued career threats. The story raises questions about user data usage and corporate rivalry. OpenAI denies accessing user data and offers compromise. The proof requires peer review. This controversy highlights ethical concerns in AI research.
Suno replaces its AI models with a new one trained on licensed music as copyright suits pile up
TechCrunch reports Suno v6 is built on licensed data as copyright lawsuits continue. It includes base, wild and mini versions, supports editing via prompts or references, and Suno plans to add opt-in remixing with artists. Older models will be retired.
OpenAI claims GPT-6 Astra is an ethereal 'Alien Mind' with AGI-like qualities — company warns of alignment challenges as new frontier leader emerges
OpenAI's GPT-6 Astra is announced as a new frontier model, with claims of AGI from Jensen Huang and a blog post from chief scientist Jakub Pachocki describing it as an 'alien mind'. The article discusses the model's benchmarks, training on 100,000 Blackwell GPUs, and its cheaper per-task cost compared to Anthropic's Fable 5.1. It also covers Pachocki's caution about alignment, calls for international coordination, and the context of rising competition from cheaper alternatives like Deepseek and Google Gemini.
Researcher reverse-engineers infamous Stuxnet malware source code, publishes it on Github for all — attack targeted Iranian nuclear facilities and was the first software of its type to cause physical damage
A security researcher has released a reverse-engineering of Stuxnet's source code on GitHub. The malware targeted Siemens industrial controllers at Iran's Natanz plant and was the first digital worm known to cause physical damage by manipulating centrifuge rotors. It spread via USB zero-days, Print Spooler, and network shares, and used stolen Realtek/JMicron certificates. Build instructions are included, with a warning to run it only in isolated Windows XP/7 VMs.