In-depth articles with more detailed investigations, interesting thought pieces, and more.
A surprisingly simple technique can dramatically improve how well artificial intelligence models answer questions: Researchers at Google have discovered that repeating a question twice in the same prompt boosts accuracy by as much as 76 percent for certain tasks. Carl Franzen reports for VentureBeat that the finding challenges years of increasingly complex methods engineers have …
AI-generated content often feels generic and unreliable, resembling “toys” rather than professional tools. Replit CEO Amjad Masad identifies the core issue: Everything looks the same, from images to code, Taryn Plumb reports for VentureBeat. The problem, known as “slop,” stems from lazy prompting and a lack of individual flavor. Masad believes platforms must expend more …
Artificial Analysis has overhauled how the AI industry measures intelligence, replacing traditional benchmarks with tests that measure whether AI can complete actual work tasks. Michael Nuñez reports for VentureBeat. The independent benchmarking organization removed three widely cited tests from its Intelligence Index, including MMLU-Pro and AIME 2025. The new version 4.0 introduces 10 evaluations focused …
Adam Mosseri, the head of Instagram, believes it will soon be more practical to verify authentic media than to detect AI-generated content. Karissa Bell reports for Engadget that Mosseri expects synthetic imagery to overtake human-made content on social platforms. This shift forces a rethink of how apps identify what is real. Mosseri suggests that the …
More than 20% of videos recommended to new YouTube users consist of low-quality AI-generated content designed solely to generate views and revenue, according to new research. Video-editing company Kapwing surveyed 15,000 of the world’s most popular YouTube channels and identified 278 that contain exclusively “AI slop”, Aisha Down reports for The Guardian. These channels have …
Silicon Valley’s AI enthusiasts are frustrated with public skepticism, but they may be missing the point. While industry insiders celebrate what they see as near-miraculous advances, many ordinary people view AI progress with anxiety or indifference. Sharon Goldman reports for Fortune that the disconnect stems from fundamentally different perspectives. What AI builders frame as thrilling …
Amazon faces a critical decision about AI-powered shopping agents that could reshape the e-commerce industry. The company must choose whether to fight these tools or embrace them as partners. Annie Palmer reports for CNBC that Amazon CEO Andy Jassy has watched competitors like OpenAI, Google, Perplexity and Microsoft release AI agents that allow consumers to …
Mathias Schindler, a longtime Wikipedia contributor and co-founder of Wikimedia Germany, reports on a troubling discovery at the 39C3 conference in Hamburg. While developing a tool to check ISBN checksums in German Wikipedia, he uncovered a significant problem: articles containing completely fabricated literature references generated by large language models. The issue emerged when Schindler found …
The development of large language models has undergone fundamental changes in 2025, marked by new training methods and surprising capabilities that reveal a fundamentally different form of intelligence than expected. AI researcher Andrej Karpathy writes on his blog about six major shifts that defined the year. The most significant change involves a new training technique …
Artificial intelligence has developed a distinctive writing style that readers are learning to identify almost instantly. From student essays to corporate communications, AI-generated text carries unmistakable markers that reveal its algorithmic origins. Sam Wolfson writes about this for the New York Times. He describes how AI writing relies heavily on specific patterns: the “It’s not …