In partnership with

Good morning. It’s Friday, December 12th.

On this day in tech history: In 2008, Google released the first stable version of Chrome, featuring the V8 JavaScript engine that compiled code to native machine instructions for blazing speed. This turbocharged web apps, enabling complex client-side computations that paved the way for browser-based AI tools like TensorFlow.js.

In today’s email:

• OpenAI launches GPT-5.2

• Google debuts updated Deep Research Agent

• Etcetera: Stories You May Have Missed

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

You can (easily) launch a newsletter too

This newsletter you couldn’t wait to open? It runs on beehiiv — the absolute best platform for email newsletters.

Our editor makes your content look like Picasso in the inbox. Your website? Beautiful and ready to capture subscribers on day one.

And when it’s time to monetize, you don’t need to duct-tape a dozen tools together. Paid subscriptions, referrals, and a (super easy-to-use) global ad network — it’s all built in.

beehiiv isn’t just the best choice. It’s the only choice that makes sense.

Start free today. No credit card required

OpenAI launches GPT-5.2 and scores a $1B Disney deal

OpenAI has officially launched GPT-5.2, a three-tier model lineup: Instant, Thinking, and Pro, pushing AI reasoning, coding, math, and visual comprehension to new heights.

GPT-5.2 Thinking outperformed human benchmarks on GDPval while cutting hallucinations by 38 percent versus GPT-5.1. The Pro model tackled a simplified open problem in statistical learning theory without human guidance. Visual reasoning and coding saw major gains, from parsing low-resolution motherboard images to scoring 80 percent on SWE-Bench Verified. Developers gain efficiency through OpenAI’s new response-caching system.

A day prior to the release, the company warned that next-gen AI poses “high” cybersecurity risks, from zero-day exploits to industrial intrusions. OpenAI is responding with the Frontier Risk Council, a security expert advisory group overseeing safeguards and frontier-capability risks.

OpenAI also just struck a three-year $1 billion licensing and equity deal with Disney, letting over 200 Disney, Pixar, Marvel, and Star Wars characters appear in Sora and ChatGPT Image. Nicholas Grous of Ark Invest called it a potential dividing line in entertainment: “We're not just going to want to be fed AI slop for 16 hours a day.” Bob Iger emphasized the initiative “does not in any way represent a threat to the creators at all,” noting that name, likeness, and voices are excluded. Disney will curate promising AI-generated content for Disney+, creating a feedback loop between fan creativity and premium production.

Meanwhile, ChatGPT’s long-awaited “adult mode” is slated for Q1 2026, using a tested age-prediction model to keep minors out. The company is also currently facing internal tension as staff like Tom Cunningham leave, allege that the company’s economic research is drifting into AI advocacy. Cunningham reportedly struggled to publish research on AI’s negative economic impacts, including job displacement. Chief strategy officer Jason Kwon defended the approach, stating the company must not only identify risks but actively “build the solutions. Read more.

Google debuts updated Deep Research Agent amid GPT-5.2 release 

Gemini 3 Pro has leapt in real-world reliability, scoring 69 percent trust in the HUMAINE benchmark versus 16 percent for Gemini 2.5. Blind, multi-turn testing with 26,000 participants revealed gains in reasoning, adaptability, and ethics. Complementing this, Google and Kaggle’s FACTS suite shows AI factuality still caps near 70 percent, with Gemini 3 Pro leading. Gemini 3 now powers AI Mode in Search, Stitch for interactive app prototypes, and TTS models that provide natural speech across multiple voices and languages.

Google launched an upgraded Deep Research Agent, on the same day OpenAI dropped GPT-5.2. The autonomous research agent is built on Gemini 3 Pro and accessible to developers via the new Interactions API. It has also integrated Anthropic’s Model Context Protocol (MCP) into its cloud, letting AI models natively access tools like Maps, BigQuery, Compute Engine, and Kubernetes without custom interfaces. On the web, Google rolled out Disco, which turn browser tabs into handy mini-apps for research, planning, and prototyping.

Meanwhile, Disney sent a cease-and-desist over alleged large-scale copyright infringement by Gemini AI, citing unauthorized reproduction of characters across Frozen, Star Wars, and Marvel franchises. Read more.

• Sam Altman’s World project merges its iris-scanning verification with a new payments and encrypted chat “super app.”

• Researchers build digitally programmable, fully autonomous microrobots, shrinking complexity by 10,000-fold

• New survey shows a growing trust gap between AI investors and everyday users

• Oracle’s revenue miss sparks fears that tech giants are overspending on AI

• Pew survey shows two-thirds of U.S. teens now use AI chatbots, with 30% logging in daily

• Time magazine names 'Architects of AI' Person of the Year 2025

• AI in space requires new cooling tech and cheap rockets

• China launches massive 1,243-mile AI network with ultra-low latency and 98% single-center efficiency

• MIRO U, the world’s first six-armed humanoid, begins pilot testing in a Chinese factory

• Watch: Apollo humanoid robot tackle never-before-seen objects with Google DeepMind

• Runway releases its first world model, adds native audio to latest video model

• Tencent raids ByteDance for AI talent as competition heats up in China’s artificial intelligence race

• Pebble’s Index 01 is a smart ring that turns your thoughts into notes without ever needing a charge

• Trump signs order aimed at curbing state AI laws

• From bitcoin to AI, El Salvador launches a nationwide partnership with xAI for personalized learning

• ElevenLabs joins forces with Meta to bring AI audio to Instagram, Horizon, and more

• Cursor Browser gets a Visual Editor for drag-and-drop webpage tweaks

• Startup Marble launches free AI tool as it eyes disruption in the $250 billion US tax industry

• AI can learn cultural values like kids do, study shows using video game experiments

• Nvidia’s Blackwell GPUs gain location verification software

• Decades-old Erdős problem #1026 finally cracked through a mix of AI, literature, and online collaboration

• Wells Fargo CEO warns AI-driven efficiency will mean more job cuts as the bank eyes leaner operations

• Microsoft study reveals peak Copilot usage for relationship advice spikes every February and late at night

• Watch: Boston Dynamics explains the eerie but efficient stand-up move behind Atlas 

• Deepseek’s latest AI model fueled by smuggled U.S. Nvidia chips

• The latest Qwen3-Omni-Flash upgrade turns the AI into a multilingual, roleplay-ready conversational powerhouse

• Boom Supersonic pivots from supersonic jets to 42MW gas turbines to power AI data centers.

• South Korea’s Motif Technologies launches 12.7B-parameter open-weight AI, punching above its size in math and reasoning benchmarks

• LTX’s new Retake tool lets creators patch a single bad moment in an AI-generated clip without redoing the whole scene

• Spotify’s new AI-powered Prompted Playlists let you tell the algorithm exactly what to play

• Figma adds Photoshop-style AI tools so users can erase, isolate, and expand images without leaving the app 

• State AGs warn Meta, Microsoft, Google, Apple, and others that their chatbots’ “delusional outputs” may breach state laws

• Reddit rolls out alpha verification for public figures to curb misinformation

• Harvard and Perplexity researchers conducted the first large‑scale field study of how people use AI agents

📄 Visionary: The World Model Carrier Built on WebGPU-Powered Gaussian Splatting Platform

📄 OneStory: Coherent Multi-Shot Video Generation with Adaptive Memory ​​

📄 Self-Improving VLM Judges Without Human Annotations

📄 Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance

📄 Efficiently Reconstructing Dynamic Scenes One D4RT at a Time

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on 𝕏!