AI Reading for Sunday August 17

Are AI foundation model improvements hitting a wall? - Financial Times

Rethinking AGI and superintelligence timelines - The Washington Post

System prompt updated to make ChatGPT a little bit more friendly, sycophantic - reddit

GPT-5 has a hidden system prompt that you can't override, might be different in ChatGPT vs API - Simon Willison’s Weblog

An attempt to extract the GPT-5 system prompt - Reddit

might need to re-extract it and see if anything changed

Understanding LLM post-training via reinforcement learning with human feedback - Notion

Moneyball but for AI talent - Fortune

Reverse acquihires challenge Bay Area startup math and culture, pose conflict of interest challenges for founders and VCs and risks for early rank-and-filers with equity kickers. - The Wall Street Journal

AI agents will disrupt a lot of SaaS that doesn't evolve to be agentic - The New Stack

n8n for no-code AI agentic workflows - Geeky Gadgets

Zoom wants you video with AI avatar customer service agents - aimagazine.com

Ex-Waymo CEO: Tesla's doesn't have a real autonomous robotaxi yet - Business Insider

Is Altman targeting Musk with stuff like new BCI venture? - Forbes

Russian intelligence-linked hackers used LLM-built malware in phishing campaigns, as both attackers and defenders increasingly leverage AI for cyber operations. - NBC News

AI-assisted sniper will kill you at long range before you have any chance to detect him - UNITED24 Media

Using AI on routine eye-exam data to predict childhood myopia evolution, in some cases start treatment early to help patients compensate for eye geometry and lens and muscle and neurological issues - Nature

Shockmatrix, the emergency room triage AI - Le Monde.fr

Becoming a super-ager with AI risk scoring for top three diseases: heart disease, cancer, Alzheimer's, AI help with monitoring and quantifying and making lifestyle with wearables. (Summarizing transcripts with AI is your friend.) - YouTube

I Used an AI-powered continuous glucose monitor with an Oura Ring for 2 weeks and learned three key things about my eating habits. - CNET

NASA and Google test onboard AI medical assistant for astronaut missions to the moon and Mars like Star Trek's Emergency Medical Hologram a/k/a The Doctor - Space

Labor unions push state guardrails on workplace AI; proposals include human review of automated decisions, AI monitoring of workers etc. - MSN

Educators warn AI summaries reduce deep reading, threatening literacy and critical analysis. - Fast Company

Film schools teach AI-assisted screenwriting, but does it make them better or just reduce originality and copy others' ideas? - Gizmodo

AI slop sports 'autobiographies?' - Sky News

Your open-weight model might not be implemented properly correctly in the enterprise AI of your choice - Simon Willison’s Weblog

you have one job, guys

Hands-on with Tiramisu, Meta's prototype that combines beyond-retinal resolution, high brightness, and high contrast. - UploadVR

Neil Young stops posting in Facebook and Instragram over Meta's problematic AI chatbot guardrails - Stereogum

Fears of near-term AI catastrophe spur Silicon Valley “super preppers” to build shelters and reprioritize finances. - Business Insider

similar to living during the Blitz or NYC after 9/11, or Camus’s The Plague people react very differently to chaos

Trump mulls a 300% tariff on chips, seems problematic for building stuff in US - Tom's Hardware

Dan Fabulich argues LLMs “generate slop because they avoid surprises by design,” prediction objective forces them to minimize surprise, making them bad at comedy, storytelling, journalism, finding surprising facts, and discovering important mathematical proofs - Medium

AI Hype Is the Product and Everyone’s Buying It - Truthout

Is The AI Mania a Psy-Op? - Activist Post

Would you eat at a restaurant run by a Chef Aiman AI? - Fox News

"The real question is not whether machines think but whether men do." - B.F. Skinner

Follow the latest AI headlines via SkynetAndChill.com on Bluesky