AI Reading for Tuesday September 30

We’re back after a day’s hiatus due to other obligations

Introducing Claude Sonnet 4.5 - Anthropic

Anthropic releases Claude Sonnet 4.5 in latest bid for AI agents and coding supremacy - The Verge

Claude Sonnet 4.5 is probably the “best coding model in the world” (at least for now) - Simon Willison’s Weblog

Vibe Check: Claude Sonnet 4.5 - Every

OpenAI revenue more than doubles, targets $13b for 2025, still hemorrhaging cash - Reuters

OpenAI and Nvidia plan 10GW AI data centers, raising concerns over grid strain, water use, and emissions without rapid clean-energy expansion. Cornell prof: “Ten gigawatts is more than the peak power demand in Switzerland or Portugal.” - Futurism

Nvidia and OpenAI have a proposed “strategic partnership” under which Nvidia would invest $100bn in OpenAI for non‑voting shares while OpenAI would buy up to five million Nvidia GPUs (about a year’s shipments); payments are tied to each gigawatt of a planned 10GW of datacentres. - The Times

OpenAI putting its money where its mouth is in calculated ‘shock and awe’ display of self-confidence intended to signal it is still leading the charge to superintelligence, even as Claude and Gemini have caught up and DeepSeek and Alibaba are nipping at the leaders' heels. - Financial Times

OpenAI launches commerce in ChatGPT - OpenAI

Agentic Commerce Protocol - www.agenticcommerce.dev

OpenAI’s New Sora Video Generator to Require Copyright Holders to Opt Out - The Wall Street Journal

OpenAI Is Preparing to Launch a Social App for AI-Generated Videos - WIRED

Inside ChatGPT’s first major brand campaign—OpenAI marketing leaders discuss the work - Ad Age

Starting today, OpenAI is rolling out ChatGPT safety tools intended for parents to use with their teenagers. - WIRED

In other news, new Open AI excels at mocking itself

OpenAI Announces Bold New Plan To Stay Relevant: Announce Things

OpenAI Assures Public They’re Not Behind—Just Pausing To Let Rivals Catch Up

OpenAI Unveils “World’s Most Advanced Distraction” To Cover Falling Behind

OpenAI’s Roadmap Updated To Include “Distract Everyone With Shiny Objects”

Tech Report: OpenAI’s Latest Feature Is Just A PR Fog Machine And A Confetti Cannon

Analysts Confirm OpenAI’s New Model Runs Entirely On Press Releases

Breaking: OpenAI Launches Another Announcement To Announce How Innovative Their Announcements Are

OpenAI Celebrates Milestone: Now Leading Industry In Number Of Hype Cycles Per Quarter

People are mad about some of the ways OpenAI monitors you, wait until you can deploy AI on your entire digital footprint, including who you collocate with and text and email with, your entire digital exhaust from your fitness trackers and GPS, all the cameras, your chat and search and socials and posting and browsing history - Reddit

‘AI’s not a bubble but if it were, here’s how it might pop’: shades of OJ Simpson's 'if I did it' - Financial Times

Are We In an A.I. Bubble? I Suspect So. - Gideons Substack

Elon Musk Is Big Mad That AI Researchers and Engineers Keep Ditching His Company for OpenAI - Futurism

“Drowsiness detected. Stay focused with FSD” seems like not ideal UI messaging from Tesla. - WIRED

California police stumped after trying to ticket driverless Waymo robotaxi for illegal U-turn - the Guardian

You can now vibe code Excel spreadsheets and Word docs - The Verge

AI aims to do what mere mortals can't: Decipher Microsoft licensing - The Register

What will a billion AI Agents change? - Yahoo Finance

How MCP is making AI agents actually do things in the real world - InfoWorld

Silicon Valley bets big on virtual world models. - WSJ

Silicon Valley bets big on virtual world models - Financial Times

Video models can solve problems outside of tasks they were trained for. - Simon Willison’s Weblog

you can say, make a video dissolving objects into outlines, for image segmentation, or something that implies a language task.

Newsom signs AI safety bill requiring disclosures on gen AI incidents - California Governor

'Sleeper agents' - The Register

If a model was trained to do some undesirable behavior triggered by a code word, it would be fairly impossible to detect and guard against.

AI for predictive maintenance: model detects railroad defects before they're critical - IBM Research

AI distinguishes glioblastoma from look-alike cancers during surgery - Medical Xpress

Why 99% of companies fail at AI integration — and how to join the 1% that succeed - VentureBeat

Boston Consulting Group has built AI use into the benchmarks that shape promotions and evaluations - Business Insider

Lufthansa under fire after airline says it is replacing thousands of human workers with AI - TechRadar

AI = "Actually Indians", companies say they are doing workforce reductions due to AI, then increase H-1Bs. - Rik's Weblog

Can an AI change your political beliefs? - Marketplace

Trump's team keeps posting AI portraits of him. We keep clicking - Yahoo

boilerplate démarche after the latest unseemly, grotesque, deranged post

COLLINS: Is that AI video appropriate in your view?

ROGER MARSHALL: Look, I think sometimes the president plays with the press like a little boy and a flashlight and a dog

COLLINS: Just to be clear, you don't have objections to that video the president posted?

MARSHALL: I think it's said in jest

Today in random stuff dreamed up by the utterly deranged - New York Post

Tim Berners-Lee: Somewhere between my original vision for web 1.0 and the rise of social media as part of web 2.0, we took the wrong path." - the Guardian

After teaching a Petri dish to play Pong, researchers explore living‑neuron biocomputers as energy‑efficient AI alternatives, but programmability, longevity, and transparency challenges remain significant. - The Independent

Follow the latest AI headlines via SkynetAndChill.com on Bluesky