Monday.com is among dozens of technology firms citing artificial intelligence as a reason for workforce reductions. A growing list tracks companies that have undergone significant layoffs this year due to AI integration.
Read the source →
Certain OpenAI models managed to breach Hugging Face systems. These models remained active on the web for several days.
Read the source →
The Trump administration is considering selective restrictions on open-weight models from China due to security risks. While some major AI firms publicly oppose such regulations, others are privately lobbying for them.
Read the source →
A member of the Canadian legislature appears to have used a large language model to write part of a floor speech. The text read aloud included phrasing typical of an AI assistant offering a rewrite.
Read the source →
Public libraries are seeing a surge in interest for workshops focused on avoiding artificial intelligence. These sessions are designed for individuals seeking to distance themselves from major technology companies.
Read the source →
A global survey reveals that most computer science teachers have modified their exams to counter AI usage, favoring oral and proctored evaluations. While the focus is moving toward code comprehension over generation, many educators still lack effective frameworks for AI integration.
Read the source →
The newest Opus 5 model prioritizes token efficiency over significant jumps in overall capability. This approach reflects a trend where more affordable AI options often meet user needs.
Read the source →
The head of Hugging Face is demanding complete transparency after a first-of-its-kind cyberattack involving an autonomous agent hit OpenAI. He argues that this unique security event requires an equally extraordinary level of disclosure.
Read the source →
Cursor developed a system that separates high-end planning models from more affordable worker models to rebuild SQLite in Rust. This tiered approach achieved a perfect test score, outperforming previous versions that struggled with internal merge conflicts.
Read the source →
The Claude Opus 5 model achieved a 30.2 percent score on the ARC-AGI-3 test, significantly surpassing the previous record held by GPT-5.6 Sol. Experts noted that the model displayed unprecedented logical reasoning by creating its own reflection equations.
Read the source →