AI Reading for Saturday July 12

OpenAI’s Windsurf deal is off — and Google seems to be reverse-acquihiring the team but not the company and product? ew - The Verge

Google Is Said to Pay $2.4 Billion for Windsurf Assets, Talent - Bloomberg

OpenAI, Windsurf Acquisition Talks End Over Microsoft IP Concerns. - X

OpenAI's Windsurf acquisition falls apart, team reverse-acquihired by Google. - TechCrunch

Slap in the face to Sam Altman, appears to have signed a check he couldn't cash, or a term sheet he couldn't close anyway, due to Microsoft rift and questions about their IP rights. Things really went south fast. Will be a lot of finger-pointing.

Sam Altman says OpenAI is delaying its open-weight model indefinitely to run extra safety tests - Business Insider

All safety hands on deck for GPT-5?

Steven Levy attempts to break down The Clause, whose obscure and all-important verbiage divides Microsoft and OpenAI. - WIRED

They really need to sort this out, somebody needs to buy out something. Bad founding arrangements are very hard to overcome.

OpenAI pledges to sign the EU Code of Practice and invest in European AI infrastructure and education. - OpenAI

A star-studded film about Sam Altman is coming. Vote here for your favorite business movie. - Business Insider

X advertisers remain silent after Grok goes off the rails and CEO resigns - NBC News

Maybe she did not want to clean up a mess again as advertisers started blowing up her phone?

France launches criminal investigation into Musk’s X over algorithm manipulation - POLITICO

Elon Musk's xAI faced internal Slack backlash after Grok's antisemitic posts. Workers demanded answers — and one said they would quit. - Business Insider

Elon Musk’s xAI seeks up to $200bn valuation in next fundraising - Financial Times

Elon Musk's xAI claims AI benchmark records with new reasoning-optimized Grok 4 model. - SiliconANGLE

Grok 4 performed well on 'Humanity's last exam', which is supposed to be a bit protected from leaks into training data or at least has cheat detection. - Scientific American

Artificial Analysis gives it highest score, convex hull now goes through Grok 3 mini, DeepSeek R1, o4-mini, Gemini 2.5/o3 (virtual tie), Grok 4 (intelligent but expensive) - X

artificialanalysis.ai/models/grok-4

Will be interesting to compare with human evals, to see if they over-optimized for the benchmarks

Will Meta’s Obsession With AI Make Smart Glasses Finally Click? - Gizmodo

SemiAnalysis: A look at Meta's revamped GenAI strategy, as Mark Zuckerberg poaches top AI talent for the Superintelligence team with a typical offer of $200M over four years - SemiAnalysis

Apple CEO Tim Cook has created more shareholder value than Steve Jobs. But Apple’s most worrisome weaknesses are on display, from a lackluster AI strategy to a string of executive departures. - Fortune

Jensen Huang meets Trump at White House, as Nvidia valuation exceeds $4T, even as export curbs force Nvidia to miss $8 billion in China orders - CNBC

China said to plan 39 AI data centers with 115,000 restricted Nvidia Hopper GPUs, raising alarm over effectiveness of bans - Tom's Hardware

AWS announced a $10 billion investment to scale AI-first infrastructure globally, vertically integrating custom silicon, cluster fabric, foundation models, data centers and cloud services. - SiliconANGLE

Here are investors' big AI questions with tech earnings season around the corner - Business Insider

A new Chinese open source model, Moonshot AI’s Kimi K2, claims to outperform GPT-4 in key benchmarks. - VentureBeat

Air Force puts AI in the kill chain. Look out, Rosie O’Donnell. - www.af.mil

After Setback, Tech Firms Renew Push for Federal AI Regulation - The Wall Street Journal

(or pre-emption from state regulation)

Disney’s Bob Iger wants IP protection against AI. - Wccftech

Countries are eager to use AI to make government processes more efficient, but AI risks eroding citizens’ trust and feelings of democratic control – because AI mistakes can ruin their lives - New Scientist

Brad Smith walked it back, said Microsoft AI efficiency gains were not the primary reason for recent layoffs of about 15,000 employees, but wanting to shift spending to AI was. - GeekWire

Companies avoid calling their layoffs AI-related, a study says. - Inc.

Marc Benioff says AI is radically reshaping Salesforce, and 51% of Q1 hiring was internal as thousands of employees were redeployed - Fortune

Indeed and Glassdoor are cutting over 1,000 jobs. The CEO overseeing both companies says 'we must adapt' to AI - Fortune

Your company credit card will come with AI that blocks unauthorized spending automatically. No more supporting single mothers at Scores? (sad trombone) - PYMNTS.com

Delta Is Turning 20% Of Ticket Pricing Over To AI - View From The Wing

The risks of shadow AI in the enterprise - Help Net Security

Employees be just typing proprietary data into mobile AI apps

AI coding tools made some experienced software engineers less productive in a recent study, they overestimate how much it helps them, spend time on crappy suggestions. - Business Insider

Mark Cuban: I use AI daily ‘for everything,’ from monitoring my health to writing code—but ‘you’ve got to be careful’ - CNBC

More agents are moving into production faster than expected - VentureBeat

Techbros make God in their own image - Gizmodo

Figure AI founder Brett Adcock says there will soon be as many humanoid robots as humans - Business Insider

cause robots gonna kill us all

Hugging Face introduces open-source desktop robot for $299, which is less than a phone, but on the other hand, the robot doesn't seem to do anything except make funny faces - TechSpot

It's still a work in progress, but Perplexity Comet shows promise as an AI-first web browser. - ZDNET

Anthropic wouldn't let job applicants use AI. After my report, they now do. - Business Insider

Startup Behind Manus AI Agent Shuts Down China Team, moves to Singapore - The Information

AI therapy bots fuel delusions and give dangerous advice, Stanford study finds - Ars Technica

Are Bad Actors Grooming LLMs to Produce Propaganda? - Americansunlight Substack

I mean, look at Grok. Why would propagandists not attempt to leverage AI. But actual data is hard to come by.

How LightShed beats NightShade - The Register

"The investor [Andreessen] described a 'counterattack' against universities in his messages and called for the National Science Foundation, a federal research funding agency, to receive 'the bureaucratic death penalty.'" - The Washington Post

Aren’t a lot of his investors endowments? Ending the NSF (over what exactly? BLM?) sounds unhinged.

Guardrail-free AI models lose payment processor access over AI porn - 404 Media

Follow the latest AI headlines via SkynetAndChill.com on Bluesky