Buckle up, buttercup. This week in AI was absolutely unhinged — and we mean that in the best possible way. Grab a coffee (or let an AI brew it for you) and let’s dive in.
🧮 AI Cracks an 80-Year-Old Math Mystery
An internal OpenAI model disproved a geometry conjecture first posed by mathematician Paul Erdős in 1946 — discovering an infinite family of examples that yield a polynomial improvement over what mathematicians had believed was optimal for nearly 80 years. It wasn’t trained specifically for this problem. It didn’t retrieve an existing solution. It just… figured it out. Cambridge mathematician Timothy Gowers declared that no previous AI-generated proof had come close to this standard, and that it would merit publication in a top math journal even if performed by humans. Mildly terrifying, extremely cool. Read it on Scientific American → | OpenAI’s official announcement → OpenAIScientific American
💔 Meet Your AI Ex-Girlfriend (Or Boyfriend)
People are uploading chat logs, social media posts, and photos to AI platforms, which use a tool called Ex-partner.skill to generate a virtual replica of their ex — then refining it further with personal memories like trips, eating habits, and past arguments. The creators insist it’s for “emotional healing.” The app has become a hit in China, with some users integrating it into WeChat, while the rest of the world slowly processes what year we’re living in. No word yet on whether the AI ex leaves you on read. Read it on South China Morning Post → | More on TechRadar → South China Morning PostDexerto
⛪ The Pope Walked Into an AI Conference…
Not a joke. Pope Leo XIV released his first encyclical Magnifica Humanitas — on safeguarding the human person in the age of artificial intelligence — presenting it personally alongside Christopher Olah, co-founder of Anthropic. The 42,300-word document urges governments, corporations, and individuals to slow the rate of technological development and ensure AI remains subject to ethical and political oversight. The Vatican is officially in the AI ethics chat. We love a crossover episode. Read it on TIME → | Vatican News → Vatican NewsTime
🎓 The OG AI Legend Switches Teams
Andrej Karpathy — co-founder of OpenAI and former head of AI at Tesla — has joined Anthropic’s pretraining team, posting simply: “I think the next few years at the frontier of LLMs will be especially formative.” He’ll be building a team that uses Claude itself to accelerate Claude’s own pretraining research — which is either very meta or very genius, possibly both. AI Twitter immediately compared it to Kevin Durant joining the Warriors. Read it on TechCrunch → | Axios → TechCrunchKingy AI
That’s your AI chaos roundup for Wednesday, May 27, 2026. Same time next week — assuming the robots haven’t taken over by then. 🤞