Happy Friday! The North Pole’s gone digital, and things are getting weird. No more waiting in line for mall Santas with questionable beards. Now kids can chat for hours with an AI Santa who sees them, remembers everything, and helps pick out the perfect gifts. This slightly cringe St. Nick isn’t just spreading holiday cheer. He’s also reigniting debates about digital safety, data privacy, and how close we really want kids getting to a virtual icon who knows if they’re on the naughty list.
Plus: OpenAI launches GPT 5.2. Yawn.
Darth Vader has joined the ChatGPT multiverse. Disney is making a $1 billion investment in OpenAI as part of a 3 year licensing deal. For the first time ever, fans can create their own AI-generated videos featuring more than 200 iconic characters from Disney, Marvel, Star Wars and Pixar. The agreement lets Sora users generate unlimited short social videos from the IP, with some streaming directly on Disney+. This is a huge leap toward putting you in charge of storytelling.
Vibe coding was 2025. Cursor’s now into vibe designing. Their new Visual Editor lets people build web apps by describing how they should feel. Fonts, colors, and layouts can all be tweaked using natural language, without having to hand off mockups to developers. Cursor’s raised $2.3 billion betting designers will happily ditch Adobe and Figma for this friction-free AI playground. Maybe now sites won’t all look like generic weekend hackathon projects.
SPENDING MADE SMARTER
Ramp makes finance easy. It helps companies of all sizes save time and money with smarter corporate cards, bill pay, and expense management.
- Simple and efficient to use.
- AI agents handle routine tasks & spot savings for you.
- Used by 50,000+ finance teams, with a ⭐⭐⭐⭐⭐ review on G2.
State AGs tell AI giants to fix their delusional bots. After a series of chatbot-induced mental health disasters (including suicides and murders), attorney generals from across the US are demanding that Microsoft, Google, OpenAI, and Meta rein in their hallucinating AI. AGs want clear user warnings, transparent audits, and mandatory safety tests before new models launch. The move escalates a showdown between states and a pro-AI White House already working to limit regulatory power. AI could someday come with warning labels like “Caution: May cause you to question reality.”
The next big leap in AI is spatial intelligence. Imagine walking through a park on a crisp fall morning. Warm sunlight hits your face as leaves crunch beneath your feet. Nearby, your eyes follow the graceful arc of a frisbee, predicting exactly where it will land. Runway’s new world model, GWM-1, immerses AI in hyper-realistic simulations like this. It teaches AI to understand and interact with dynamically complex worlds intuitively, without pre-training for every possibility. Just don’t ask what it’s training for.
The smarter AI gets, the less you can trust it. Sounds backwards, but it’s true. AI is growing incredibly powerful, yet the data feeding it keeps getting noisier, more misleading, and outdated. Reliable, high-quality data is getting harder to find, leaving AI confidently spouting errors and hallucinations. The paradox is that better AI doesn’t need more data. It needs smaller sets of carefully verified data. The real trick is figuring out what actually matters.
TIME’s Person of the Year is… a group of AI billionaires. 75 years ago a computer appeared on the cover. Apparently today building circular economies and sparking a global AI arms race is what it takes. Sam Altman, Jensen Huang, Musk, Zuck and the gang spent 2025 pushing AI into every corner of daily life. It solved unsolvable math problems, chatted with whales, and predicted hurricanes better than humans. Impressive? Maybe. But handing another award to tech leaders is painfully predictable. Somehow Polymarket got it wrong. Perhaps next year “Artificial Intelligence” will win.
fun stats
💋 $50k. Cost for Keeper’s AI matchmaking service to find your “soulmate.”
💍 10%. Success rate for Keeper’s AI turning dates into marriages.
🏁 11x. Speed increase of GPT-5.2 outputs vs previous benchmarks. Faster = better?