Latest / UpNext AI / Thrive Holdings’ ADT Deal, Mistral Large 4, and Meta’s Style Sketch
Transcript
- Welcome to the UpNext AI podcast. It's Friday, October 30th, 2026, and here's what matters in AI today. Thrive Holdings is buying home-security company ADT in a deal with a total enterprise value of $15.3 billion. TechCrunch reports that the transaction values ADT’s equity at about $7.6 billion, or $10 per share. OpenAI is an investor in and partner to Thrive, not the company acquiring ADT. Thrive Holdings grew out of Thrive Capital, the venture firm founded by Josh Kushner. Its strategy is to hold businesses long-term and bring AI into their operations. According to TechCrunch, OpenAI supplies models, technology, and engineers to Thrive’s companies. ADT is slated to become the foundation of Thrive’s smart-home security business, with plans to invest in and release new products. That puts a familiar household-security company inside an investment strategy built around applying AI to established businesses. ADT already works with Apple and Google, and TechCrunch reports those partnerships are expected to continue. The companies have not announced specific new AI-powered security products. A quick disclosure: we use models from OpenAI to help write the UpNext AI podcast. From a security business to the models organizations might use to protect one: Mistral has released the weights for Mistral Large 4, its trillion-parameter model. TechCrunch reports that businesses and governments can download and deploy it themselves, including on their own infrastructure. The weights are available through Hugging Face, and the model is also offered through Mistral’s API. Mistral is pitching that deployment control for cybersecurity. Its argument is that organizations can keep access to a model used to detect, analyze, and respond to threats across code and IT systems, rather than depend on continued access to an outside provider’s closed model. That choice could matter for teams deciding where sensitive security work runs and who controls the system. The company also says Mistral Large 4 uses 47 percent fewer tokens per task than the open-weight model GLM-5.3 while delivering comparable performance. That is Mistral’s comparison, not a general measure of cyber-defense effectiveness. For a very different use of AI, TechCrunch reports that Meta has quietly launched Style Sketch, an outfit-planning app. It connects to Instagram to assemble a virtual closet from photos users have posted. People can also add clothes by taking pictures, recording a short video, or uploading images from their camera roll. Style Sketch suggests combinations using items already in that wardrobe. According to the app description, users can start with a particular piece they own, preview outfits on a model matched to their body shape and skin tone, and plan what to pack for a trip based on the itinerary and weather. The app is available on iOS and Android in Indonesia, Singapore, and Thailand. TechCrunch says Meta has not indicated whether it will launch Style Sketch more widely. AI can also propose mathematical proofs, which creates a harder question: how do you judge them reliably? Researchers examining submissions to the Ramanujan Challenge looked at research-level number-theory problems. Some proposed formulas can be checked numerically, while establishing a proof takes more work. The researchers studied submissions made with AI models, computer algebra systems, and formal tools. They report finding correct submissions that differ conceptually from the official solutions, as well as a proof of one published open conjecture. To assess work presented as prose, code, or formalized mathematics, they built an AI-assisted evaluation pipeline. Against a one-shot language-model judgment, that pipeline reduced false-positive verdicts on natural-language proofs from 23 percent to 6 percent. Its accuracy when judging supporting code rose from 46 percent to 58 percent. The result speaks to evaluating these challenge submissions: as AI tackles harder mathematics, the method used to check an answer matters alongside the answer itself. ...Are you building apps with voice? Elevate your app's voice capabilities with ElevenLabs. Their API is a game changer for embedding dynamic, responsive voice interactions in your applications, providing unprecedented realism, flexibility and latency. Visit up next dot fm slash eleven to check out their latest offerings. ... A few more developments before we go. Simon Willison notes that eligible OpenAI subscription plans now include matching monthly API credits. A subscription priced at $100 per month, for example, comes with $100 in credits for an API organization, giving subscribers a way to use the API under their existing plan. The Information reports that DeepSeek’s funding round this month raised nearly $16 billion. Its account says the investors included Chinese technology companies and state-backed institutions; Tencent and battery maker CATL remain its two largest investors. After Google’s quarterly earnings, its stock fell as much as 10 percent, according to The Decoder. The outlet reports that cloud revenue more than doubled to $30.7 billion; the market reaction alone does not establish what AI search contributed to the result. And in game development, Simon Willison highlights Spawn, a multiplayer world where players prompt an in-game agent to change their surroundings. Founder Jacob Sansbury says Fable-powered agents also work on Spawn’s live code and ship changes hourly—an unusually tight development loop for a game. Before we wrap up, a quick note: this podcast is generated with the assistance of AI and is intended for informational purposes only. All referenced articles, research, and commentary remain the property of their original authors and publishers. And that's your briefing for today. If you enjoyed this episode, share it with a friend or colleague—and follow UpNext AI on Apple Podcasts, Spotify, or wherever you get your podcasts! Full source links are in the episode notes, and we'll be back Monday with what's up next!