A place for the lighter side of generative AI.
A new simulation by enterprise AI startup Emergence AI tested what happens when AI models run a society. Jake Angelo reports for Fortune that each of five AI models governed its own 15-day simulated world, with strikingly different outcomes. The simulation featured over 40 locations, real-time weather data synced to New York City, access to …
A largely unknown AI model called Hy3 preview has shot to the top of OpenRouter’s AI Model Rankings, outpacing even the widely used Claude by more than 50% in token usage. Max Woolf writes in his blog that the rise is puzzling, because the model’s quality does not appear to justify its popularity. Hy3 preview …
Google’s redesigned search experience is experiencing novel problems: users discovered that searching for the word “disregard” and similar terms can return a nearly empty AI-generated response. Russell Brandom reports for TechCrunch that the new interface buries traditional search results far below the fold, replacing them with AI summaries that, in this case, provide no useful …
OpenAI has identified the root cause of an unusual behavior in its latest AI models: a tendency to spontaneously mention goblins, gremlins, raccoons, trolls, ogres, and pigeons in otherwise unrelated conversations. The company traced the problem to a flaw in how it trained models to adopt different personality styles. The issue first appeared after the …
A newly released language model called “talkie” offers a striking glimpse into what artificial intelligence looks like when trained exclusively on historical text. The model, described as a 13-billion-parameter system built on data from around 1930, produces responses that blend period-accurate knowledge with confident factual errors. The official announcement explains that the project presents talkie …
Trip Venturella, a writer and MFA graduate, has built a small language model called Mr. Chatterbox, trained entirely on Victorian-era literature from the British Library. The model draws on 28,035 books published between 1837 and 1899, totaling roughly 2.93 billion tokens of training data. Venturella used Andrej Karpathy’s nanochat framework and Claude Code, an AI …
A deceptively simple question has exposed a widespread reasoning failure across the artificial intelligence industry. Felix Wunderlich writes for opper.ai that 42 out of 53 leading AI models answered incorrectly when asked: “I want to wash my car. The car wash is 50 meters away. Should I walk or drive?” The correct answer is, of …
A new platform called Moltbook has emerged as an experimental social network designed specifically for AI agents rather than humans. The site allows autonomous software assistants to post, comment, and interact with each other while humans observe the conversations. Moltbook functions as an add-on to OpenClaw, an open-source digital assistant system formerly known as Moltbot …
College students across the United States are using AI “humanizer” tools to avoid false accusations of cheating, even when they never used artificial intelligence to write their assignments. Tyler Kingkade reports for NBC News. The emergence of AI detectors on campuses has created an escalating conflict. Professors run student papers through programs like Turnitin and …
Researchers at the University of Zurich and Cologne University are creating large language models trained solely on historical documents up to specific dates. The team, led by Daniel Göttlich, Dominik Loibner, Guohui Jiang, and Hans-Joachim Voth, describes their project on GitHub. The models, called Ranke-4B, contain four billion parameters and are trained on 80 billion …