Get Big or Get Out
Chatbots Keep Telling Stories About Lighthouse Keeper 'Elias Thorne'. We Might Know Why
404 Media reports that multiple, unrelated large language models, when asked for a short story, keep independently inventing the same character — a lighthouse keeper named Elias Thorne — and investigates why models trained on overlapping data converge on the same defaults.
A Farmer Donated Land to Turn into a Park. The City Is Building a Massive Data Center Instead
In 1999 a farmer gave 87 acres to a small Texas city on the understanding it would become a park. The city sold the land to a data center developer for $10 million.
If Claude Fable stops helping you, you'll never know
Simon Willison highlights a detail from the 319-page system card for Fable 5 and Mythos 5 concerning the models' capacity to accelerate their own development, and the conditions under which that assistance would be silently withdrawn.
Software Update Automatically Turns off Amazon Delivery Drivers' AC During Dangerous Summer Heat
A software update to Amazon delivery vehicles automatically shut off the air conditioning during a stretch of dangerous summer heat.
A report suggests Amazon's Mississippi data centers have already pushed up electricity rates for nearby residents.
Breaking: Google liable for hallucinations
Gary Marcus reports on a German court decision finding Google liable for AI-generated hallucinations, and argues the ruling could spread if other jurisdictions follow.
Why AI hasn't replaced software engineers, and won't
Writing at Normal Tech (formerly AI Snake Oil), Arvind Narayanan and Sayash Kapoor frame coding agents as a normal technology and argue that the predicted replacement of software engineers has not happened and will not.
Slop, productivity, and why the AI-fueled world is going nowhere mighty fast
Gary Marcus points to a Financial Times chart by John Burn-Murdoch that, he argues, captures the gap between the scale of AI adoption and the absence of measurable productivity gains.
Simon Willison quotes Andrej Karpathy on Jevons paradox in software: as working software "increasingly comes out on a tap," his own demand for it grows substantially — explainers, visualizers, dashboards, bespoke single-use apps.