Today's Hallucination HQGPT-5.5: Because Apparently 5 Wasn't EnoughOpenAI has released GPT-5.5, a new model sitting between GPT-5 and whatever they're cooking next — which, at this rate, should arrive sometime before your lunch cools down. The company describes it as stronger on reasoning and more reliable across complex tasks. The naming convention continues to raise questions that mathematics cannot answer. Think of it less as a product launch and more as OpenAI reminding you that their release calendar is a suggestion, not a contract. Source: OpenAI
Microsoft Invents "Vibe Working," A Phrase Nobody Asked ForMicrosoft is rolling out Agent Mode across Word, Excel, and PowerPoint — a more capable AI assistant that can handle multi-step tasks without you babysitting each click. They've been calling this "vibe working," which sounds like something you'd say when you can't explain what you do at dinner parties. It's a meaningful upgrade to Copilot, letting agents operate more autonomously within documents and spreadsheets. The feature is real and genuinely useful. The name, however, remains a cry for help. Source: The Verge
Astronomers Are Now Competing With Chatbots for GPU TimeScientists hunting for galaxies are increasingly relying on GPU clusters to sift through vast telescope datasets — the kind of work that would take human eyes approximately forever. The problem is they're now queuing alongside every AI startup with a pitch deck and a dream, contributing to an already strained global GPU shortage. It is, frankly, a strange civilisational moment when finding the edge of the observable universe has to wait because someone's training a chatbot to write better LinkedIn posts. Source: TechCrunch
Ars Technica Publishes Its AI Policy. Turns Out: Humans Write the Articles.Ars Technica has released a detailed newsroom AI policy, and the headline finding is refreshingly old-fashioned: journalists write the stories, AI doesn't. The policy permits AI for specific background tasks — transcription, research assistance — but draws a firm line at AI-generated editorial content. In an era where "AI policy" often means "we tried to have one," Ars deserves credit for being specific. The document also implicitly answers a question readers probably had: yes, someone is still actually reading the thing before it's published. Source: Ars Technica
Google Gives You an AI Intern Who Never Takes a Lunch BreakGoogle has updated Workspace with a suite of new automated features powered by Workspace Intelligence, its unified AI layer across Gmail, Docs, Sheets, and Meet. The system can now summarise, draft, organise, and action tasks with considerably less hand-holding than before. Like any good intern, it's enthusiastic, fast, and occasionally wrong with tremendous confidence. Unlike a real intern, it will not eat your labelled lunch from the office fridge. On balance, an improvement. Source: TechCrunch
As always, the AI confidently knows everything — some of it even true.
|
