The Hallucination HQ

Archives
Log in
Subscribe
17 April 2026

The robot had a think about it

The Hallucination HQ

AI news that's actually fun to read — 2026-04-17

Today's Hallucination HQ

The Robot That Figures Things Out — Without Being Told

Physical Intelligence has released π0.7, an AI model designed to give robots a general-purpose brain — one that can work out unfamiliar tasks without explicit training. Think less "robot follows instructions" and more "robot has a think about it." It's an early step toward the long-sought goal of a truly adaptable machine, which is either tremendously exciting or the premise of every cautionary science fiction film ever made. Possibly both.

Source: TechCrunch


Anthropic Quietly Drops Its Most Capable Model Yet

Claude Opus 4.7 has arrived, and Anthropic is being characteristically understated about it — which, given that most AI companies announce minor updates with the energy of a moon landing, is genuinely refreshing. The model sits at the top of Anthropic's lineup and has already attracted 1,652 upvotes and 1,173 comments on Hacker News, which is the internet's way of saying people have Opinions. Full details are on Anthropic's site, which is worth a read before the discourse does it for you.

Source: Hacker News / Anthropic


AI Is Sending Shoppers to Retail Sites — and They're Actually Buying Things

Adobe has found that AI-driven traffic to US retail websites surged 393% in the first quarter of this year, with a further 269% jump in March alone. More tellingly, these visitors convert better and spend more than shoppers arriving via traditional channels. So it turns out that asking an AI "what should I buy?" and then buying exactly what it says is not just a thing people do — it's apparently the thing people do. Retailers, understandably, are delighted.

Source: TechCrunch


Gemini Will Now Generate Images Using Your Own Photos

Google's Gemini can now pull from your Google Photos library to create personalised images — powered by a model called, with remarkable confidence, Nano Banana 2. The feature sits within Google's Personal Intelligence suite, which lets Gemini tap into your apps to give contextually relevant responses. In short: show Google enough of your life and it will start illustrating it back to you. Whether that sounds convenient or slightly eerie probably says something useful about your relationship with technology.

Source: The Verge


OpenAI Names a Research Model After a Pioneer Who Was Famously Uncredited

OpenAI has introduced GPT-Rosalind, a model built specifically for life sciences research — named, one hopes deliberately, after Rosalind Franklin, the scientist whose X-ray work was crucial to discovering DNA's structure and who received rather less recognition than the occasion demanded. It's a meaningful nod. The model is designed to assist with biological and medical research, which is a serious undertaking. Naming it well was the least they could do.

Source: Hacker News / OpenAI


As always, the robots are learning. Do try to keep up.

Was this forwarded to you? Subscribe here

Written with AI assistance, edited by humans. • Privacy • Unsubscribe

Don't miss what's next. Subscribe to The Hallucination HQ:
← Newer The shoe company that became an AI firm Older → Today's AI News from The Hallucination HQ
Powered by Buttondown, the easiest way to start and grow your newsletter.