OpenAI says it’s turned off app suggestions that look like ads

Key Takeaway: OpenAI has temporarily turned off ChatGPT app suggestions that felt like ads after paid users complained, while insisting it isn’t actually running or testing advertising yet.

More Insights:

• Paying ChatGPT users reported seeing promotional-style messages for brands like Peloton and Target, sparking anger and confusion.

• OpenAI claims these were app recommendations from its new app platform, with “no financial component” and no real ad program behind them.

• Chief research officer Mark Chen admitted the company “fell short,” saying anything that feels like an ad must be handled carefully.

• OpenAI says it has turned off this kind of suggestion while it improves targeting and plans better user controls to dial suggestions down or off.

• A recent “code red” memo from Sam Altman reportedly shifts focus to improving ChatGPT quality, delaying ad-focused products despite hiring ad-savvy execs like Fidji Simo.

Why it matters: This showdown over “not-ads” is really about trust—if users feel their AI assistant is quietly shifting from helpful guide to stealth salesperson, they’ll question every answer, and once that trust erodes, no amount of new features or revenue streams can easily win it back.

Learn AI tools, agents & automations in just 16 hours (End of Year offer)

It’s almost 2026, and if you still haven’t learned how to use AI to upgrade your career, this is your wake-up call. 2025 shook the job market. AI moved fast, companies moved faster, and many people struggled to keep up

But here’s the good news — you’re not late

You still have a shot at turning AI from a threat into your biggest advantage. 💡

Outskill can teach you AI in just 2 days — the skill that can make you smarter, faster, more confident and absolutely irreplaceable in 2026.

The 16-hour AI course is everything you need to enter 2026 as an AI powered Professional

🧠 Live sessions- Saturday and Sunday
🕜 10 AM EST to 7PM EST

The Math Legend Who Just Left Academia—for an AI Startup Run by a 24-Year-Old

Key Takeaway: At 57, famed number theorist Ken Ono is leaving a tenured professorship at UVA to join his former student’s AI startup, Axiom Math, betting that AI-powered “mathematical superintelligence” will transform how deep mathematics is done.

More Insights:

• Ono went from AI skeptic to daily power user after seeing cutting-edge models solve advanced math problems in areas outside his expertise.

• He’s joining Axiom Math, founded by his 24-year-old former student and prodigy Carina Hong, which has raised $64 million and hired leading AI researchers from Meta.

• Axiom’s goal is to build an “AI mathematician” that can reason about open problems, generate new ones, and verify proofs with formal rigor.

• Ono’s role as “founding mathematician” is to design hard, representative problems and benchmarks that stretch the models and map their true capabilities.

• Disillusioned by politics and funding pressures in academia, Ono sees startups like Axiom as a place where pure math can directly shape real-world systems—from chip design to finance.

Why it matters: This move signals a deeper shift: when top, established mathematicians leave secure academic posts to help build AI that might one day outperform them, it suggests we’re not just automating routine tasks—we’re beginning to rewire who does fundamental discovery itself, and what a “career in thinking” will look like in the age of machines that think with us.

AI Pioneer LeCun Says Meta Won’t Invest in ‘World Model’ Startup

Key Takeaway: Yann LeCun is leaving Meta to build a Paris-based “world model” AI startup, without Meta’s investment, aiming at systems that understand and predict the physical world rather than just generate text.

More Insights:

• LeCun confirmed Meta will not be an investor in his new, still-unnamed company.

• The startup focuses on “advanced machine intelligence” trained on visual and sensory data to predict real-world outcomes.

• He criticizes Silicon Valley’s obsession with generative models and LLMs, saying that real-world prediction needs to be built “outside the Valley, in Paris.”

• World models target physical systems—robots, autonomous vehicles, industrial processes—rather than just word prediction.

• Mark Zuckerberg supports the project, but both concluded its applications go far beyond Meta’s core interests.

Why it matters: This marks a clear split between today’s LLM-centric AI and a possible “next wave” focused on embodied, real-world intelligence—hinting that the future of AI may be decided not just by bigger text models, but by whoever can give machines common sense about the physical world.

Other stuff

• Pat Gelsinger wants to save Moore’s Law, with a little help from the Feds

• Prosecutor Used Flawed A.I. to Keep a Man in Jail, His Lawyers Say

• In comedy of errors, men accused of wiping gov databases turned to an AI tool

• Google details security measures for Chrome’s agentic features

• We gave 5 LLMs $100K to trade stocks for 8 months

• Chatbots are now rivaling social networks as a core layer of internet infrastructure

• Jony Ive's OpenAI Device Barred From Using 'io' Name

Hope you enjoyed today's newsletter

⚡️ Join over 300,000 people using the Superpower ChatGPT extension on Chrome and Firefox.