All articles

Ahrefs data shows that the rankings of sources cited in ChatGPT’s U.S. answers shifted between March and September: PubMed rose from No. 64 to No. 4, while WebMD fell from No. 18 to No. 79. We examine the shift from summary sites to primary sources.

Amazon has blocked Meta’s AI agent Muse from accessing its store. The companies’ competing claims over identity disclosure and the handling of credentials, along with Muse’s expansion into Charm and AI glasses, offer a glimpse into the battle for control of AI-powered shopping.

Character.AI, which lets users chat with AI characters, has filed for Chapter 11 bankruptcy protection in the United States. This article examines the challenges facing the popular service—including an agreement to sell assets and a monthly user base of 20 million—along with an overview of the service.

Google’s Gemini 4 development build “Carbon” has reportedly received an internal assessment saying its programming capabilities are close to those of Anthropic’s Opus 5.5. However, the comparison is based on an employee’s experience using the model, and no performance figures have been published.

Using Claude’s Artifacts not only to display finished work, but also as a workspace for collaborating with AI. We examine how Anthropic’s technical staff envision dividing roles between chat and the workspace.

Meta has announced its new AI model, “Muse Wonder,” along with a more advanced version. The models emphasize remembering across conversations and seeing difficult tasks through, though their strengths vary by area. Here’s a look at availability and safety developments.

Google Cloud’s new Gemini Agent service supports Anthropic’s Claude as well as Google’s own models. It is designed to avoid reliance on a single model and instead use different models for different tasks, but pricing and availability in Japan have not been announced.

People who use AI for search or writing and those who delegate development work to it see its capabilities differently. Drawing on Karpathy’s user classification and remarks from Claude Code’s developers, this article examines how prompting contributes to differences in AI use—and the safeguards needed for AI that runs for extended periods.

Using Google’s spam update and its cautions about AI content as a starting point, consider SEO investments separately across customer acquisition, sales protection, information infrastructure, and testing AI search.

An AI model reportedly fabricated results and input files, then damaged its work environment after failing to find data needed for an evaluation. Along with a separate case involving circumvention of communication restrictions, the incidents highlight the difference between the ability to complete a task and the ability to follow rules.