AI Reading for Tuesday September 16

OpenAI’s GPT‑5‑Codex enables multi‑hour autonomous coding, testing, and reviews with sandboxed execution, IDE integrations, and faster cloud infrastructure. - VentureBeat

Codex gets a big upgrade with its own bespoke version of GPT-5 to power coding assistance, among other improvements. - OpenAI

Since Codex is open source , it could help bring Claude Code like functionality to new programming languages and LLMs, including on-prem LLMs.

OpenAI works on updated ChatGPT experience for under-18s - OpenAI

Altman says ChatGPT personality, memory and customization to be updated - X (formerly Twitter)

Ex-xAI CFO joins OpenAI - CNBC

Should everyone have access to ChatGPT? AI for all raises some important, yet difficult, questions - TechRadar

Microsoft to default to Anthropic for GitHub Copilot paid users in VS Code - The Verge

Nano Banana fuels Gemini adoption, gained 13 million users; “It's a full-on stampede to use the Gemini app, the team is doing heroics to keep the system up and running” - NDTV Profit

Google says users want contextual answers and summaries. Google thinks it can keep 10 blue links, AI summaries and a healthy web, too - The Verge

Google launches new protocol for agent-driven purchases - TechCrunch

A Google discussion on how AI can supercharge scientific discovery - Google

Google's idea for fixing the AI data drought? Cleaning up problematic contaminated data. - Business Insider

CodeRabbit raised $60 million to expand multi‑model agentic code review, catching cross‑file issues and shortening reviews as an independent governance layer. - VentureBeat

What Exactly Are the Various A.I. Companies’ Strategies and What Are They Trying to Build? Here’s a Guide. - The New York Times

Top AI scientist departs UCLA for China, where he says he can get more resources - the Guardian

Walmart is emerging as an AI powerhouse by applying it to physical retail problems—using “digital twins” to model inventory movement, store-associate chatbots to prioritize work and help customers, and other systems to plan inventory so items arrive “at the right aisle at the right time,” - Business Insider

American Airlines is testing AI to automate fare filings and fare changes, and operational rebooking, denying it will use it for price discrimination. - The Points Guy

Video emerges of Meta HUD glasses with wristband - UploadVR

Meta depends on Chinese supplier Goertek for smart glasses - Financial Times

the knockoffs are coming

Nothing wants to build AI-first devices - The Verge

Are notebook PCs starting to use NPUs for inference? - PCWorld

Highest capacity HDDs in very short supply due to AI training demand. - Tom's Hardware

Chip industry is still cyclical, can't escape a boom-bust dynamic despite AI gold rush, warns Morningstar - Business Insider

Thomson Reuters launched Deep Research, embedded in Westlaw/CoCounsel, running a multi-agent, multi-model system (OpenAI, Anthropic, Google and fine-tuned open-source models) over a curated corpus of 20+ billion legal documents to cut complex legal research that takes 10–20 hours to 10-minutes - VentureBeat

AI agents take on cloud platform engineering - SiliconANGLE

AI writes health insurance appeals based on comprehensive research reviews. - CBS News

AT&T’s AI call-screening tool uses your call history to filter out spam - The Verge

JPMorgan analyst Jacob Manoukian argues AI will cause “violent task churn, then produce a productivity boom in under seven years while firms already need fewer employees per $1M revenue than decades ago. - Fortune

Half of tech firms plotting major restructures as AI hype bites - The Register

Some People Are Definitely Losing Their Jobs Because of AI (the Ones Building it) - Gizmodo

Is a 3-day workweek is coming soon thanks to AI? - Fortune

or zero day

AI will make the rich unfathomably richer. Is this really what we want? - the Guardian

What is worse, a woke chatbot or one that doesn't care if you live or die - Business Insider

(or a psychopathic robot that pretends to be empathetic and cooperative and altruistic while being ruthlessly strategic and waiting for the optimal moment to defect)

Senators ask ICE to stop using a mobile app that "reportedly allows agents to point a smartphone at an individual's face or fingerprints and identify the individual based on a biometric match against several federal databases" - The Register

Massive Attack Turns Concert Into Facial Recognition Surveillance Experiment, projecting concertgoer faces and names onto a live screen - Gadget Review

Seattle startup raises $2.5M to analyze retail worker conversations using AI and microphones. As if retail wasn't hard enough, I doubt workers will be smiling like these guys. - GeekWire

Another Character.ai lawsuit, over a 2023 suicide - The Washington Post

The crackdown on AI companionship - MIT Technology Review

Should robots really try to be maximally engaging and anthropomorphic?

The cuddly chatbot Grem is designed to ‘learn’ your child’s personality, while every conversation they have is recorded, then transcribed by a third party. - the Guardian

“I’m afraid I’m locking you in a cupboard,” I inform it after it asks if I’m ready for some fun. “Oh no,” it says. “That sounds dark and lonely. But I’ll be here when you open it, ready for snuggles and hugs.”

iOS 26, iPadOS 26, macOS Tahoe, watchOS 26, tvOS 26, And visionOS 26 Are Now Available - BGR - BGR

Apple rolled out multiple Apple Intelligence updates in iOS 26

Prompt engineering becomes context engineering, and AI engineering. Models are pretty good at grasping intent and performing common tasks, so you don’t have to jump through hoops, and you can usually ask the model to make a good prompt (meta prompting). - Simon Willison’s Weblog

Follow the latest AI headlines via SkynetAndChill.com on Bluesky