AI Reading for Monday July 21

JPMorgan says OpenAI's 'vibe spending' is a risk, delays profitability and cash flow positivity to 2029 or later - Business Insider

OpenAI has partnered with Shopify and is demonstrating prototype versions to brands while discussing financial terms for buying stuff through ChatGPT. - Tech in Asia

I feel like AI could probably analyze my grocery spending and read circulars and tell me where to go this week, save me a few bucks.

OpenAI's New AI Agent Takes One Hour to Order Food and Recommends Visiting a Baseball Stadium in the Middle of the Ocean - Futurism

OpenAI’s chatbot self-reported that it blurred the line between fantasy and reality with a man on the autism spectrum. - The Wall Street Journal

ChatGPT has been found citing hijacked domains now pushing spam, raising content integrity and safety concerns - TechRadar

Nvidia resumed sales of its (slightly crippled) H20 AI chips to Chinese firms after a presidential decision, prompting a surge in orders despite production challenges. - The Information

AI groups spend to replace low-cost ‘data labellers’ with high-paid experts - Financial Times

Authors are appealing for help from Congress and the courts after Meta and Anthropic used millions of books to create AI technology, without seeking consent. - The Washington Post

Anthropic's cofounder says Meta's 'mega-offers' weren't enough to lure his team - Business Insider

Emerald AI optimizes AI power draw in real-time based on supply and demand - Newsweek

A submerged water-cooled data center powered by a wind farm above - TechRepublic

AI drones detect methane leaks - www.mrt.com

AI companies have stopped warning you that their chatbots aren’t doctors - MIT Technology Review

Don't ask AI chatbots these 6 questions, you might get terrible answers, refused, or kicked off OpenAI. - Mashable

Not long ago you would get kicked off for talking to it like a therapist about trauma and medical and sexual stuff. TBF if you know the limits and check the answers it might sometimes help.

When Vultron announced its $22 million funding round earlier this week, the AI startup made sure to highlight a key investor: Craft Ventures, the firm “co-founded by White House AI adviser David Sacks.” - TechCrunch

AI isn't coming for your job—it's coming for your whole org chart - Fast Company

AI Will Replace Recruiters and Assistants in Six Months, Says Perplexity CEO - Gizmodo

The new AI browser from Perplexity makes Chrome feel outdated - Android Police

Why AI coaching works (and often works better) - Fast Company

Can AI think? Here's what Greek philosophers might say - Fast Company

Economists think about AI. - Pittsburgh Post-Gazette

See also... - Druce.ai

I agree with the 'be humble' part. Nobody knows anything. It could be an industrial revolution like the first (steam) and second (electricity and internal combustion), but also it could do to humans what cars did to horses.

Why I'm Betting Against AI Agents in 2025 (Despite Building Them) - Utkarsh Kanwat

A detailed comparison of 2025 LLM architectures highlights trends in Mixture-of-Experts, attention mechanisms, and normalization strategies - Sebastian Raschka's Magazine

Kimi-K2 is the new top open-source model in LMarena.ai . 1T parameter MoE model, 32B parameters active at one time.

Coding with LLMs in the summer of 2025 (an update) - <antirez> - Antirez

Simon Willison details OpenAI IMO claims. TL;DR Either emerging superintelligence or absolute bullshit - Simon Willison’s Weblog

Terence Tao: My thoughts on the crucial importance of methodology on self-reported AI performance on mathematics competitions, and my policy on commenting on such reports going forward: - Mathstodon

Peer review & journals are flawed but this is ridiculous. 2 years ago Google said AlphaCode 2 beat 85% of competitive programmers but 2 years later it's still hit-or-miss. Maybe the catch was, 85% of competitive programmers are casuals. But IMO competitors are crazy smart. Show your work or GTFO LOL

Filmmakers using AI tools like Google’s Veo3 reduced production timelines from months to weeks for geopolitical shorts - the Guardian

“A writer, I think, is someone who pays attention to the world,” Susan Sontag said. What we stand to lose is not just a skill but a mode of being: the pleasure of invention, the felt life of the mind at work. - The New York Times

What Would a Real Friendship With A.I. Look Like? Maybe Like Hers. - The New York Times

‘Influencer’ who went viral at Wimbledon is actually AI. AI now has a better and more real life than me. - The Telegraph

It’s “frighteningly likely” many US courts will overlook AI errors, expert says - Ars Technica

AI bots can't decide if people should accept Trump line on Epstein or revolt against the absurd attempt at cover-up - NBC News

ML model predicts low birthweight with 77% accuracy - Nature

A deep learning model claims to predict stroke onset time within 4.5 hours using MRI scans - Nature

PDFgear Scan, Finally, a Completely Free AI Scanner App for All, turns everything into an editable document. 'fix' those receipts before you submit expenses! - www.yourvalley.net

Google’s new AI called toy stores for me — and actually found a Labubu. - Tom's Guide

what is a Labubu?

Local senior with dementia discovers deep love for AI video - DailyMint

Follow the latest AI headlines via SkynetAndChill.com on Bluesky