🚀 OpenAI drops GPT-5.2 after an internal “code red” scramble
OpenAI rolled out GPT-5.2, billing it as sharper at general reasoning, coding, and long-context work - basically the stuff that starts to feel like magic… until it doesn’t. The vibe is: fewer dropped threads, more “finish the whole project” energy.
They also framed it as more useful for real business-y tasks like building spreadsheets and presentations, plus handling multi-step workflows without face-planting halfway through. And yes, it’s clearly part of the “please don’t let Gemini eat our lunch” storyline - even if OpenAI says the panic was a bit overblown.
🔗 Read more
🎬 Disney puts $1B into OpenAI - and licenses Star Wars/Marvel/Pixar for Sora
Disney is investing $1 billion in OpenAI and letting the company use major characters inside Sora for AI video generation. That’s… huge, honestly - it’s like Hollywood cautiously stepping into the pool while still gripping the railing.
The deal includes guardrails (no “inappropriate situations”) and specifically avoids using talent likenesses or voices. Unions reacted warily anyway - because of course they did - with compensation and creative control looming over everything.
🔗 Read more
🔎 Google ships a beefed-up Gemini Deep Research agent for developers
Google reworked Gemini Deep Research and is now putting it in developers’ hands via a new Interactions API. The pitch: let apps embed an autonomous research agent that plans, searches, reads, fills gaps, searches again… the whole “intern that never sleeps” thing, except it’s a model.
They also open-sourced a benchmark called DeepSearchQA to measure how well research agents actually perform on messy, multi-step web tasks. Google says it’s tuned to reduce hallucinations and improve report quality - which is the exact promise everyone makes, but here it comes with more concrete eval talk behind it.
🔗 Read more
🇬🇧 UK government and DeepMind expand partnership around science, services, safety
The UK announced a new partnership with Google DeepMind aimed at pushing AI into science breakthroughs, cleaner energy, and “smarter public services” - the sort of phrase that sounds boring right up until it quietly changes how your life works.
It also mentions an automated research lab planned for the UK and an expanded relationship with the UK’s AI Security Institute, which reads like: “Yes we want the speed, but also please no chaos.” Or so it seems.
🔗 Read more
🧩 Nvidia’s China chip sales spark calls for CEO testimony
Sen. Elizabeth Warren called for Nvidia’s CEO (and the U.S. Commerce Secretary) to testify after President Trump moved to greenlight sales of Nvidia’s H200 AI chips to China. The concern is national security - and the timing is awkward alongside a Justice Department crackdown on chip smuggling.
The White House drew a line between illegal smuggling and licensed exports to approved buyers, while Nvidia emphasized any China sales still require U.S. government licenses and represent a small share of its advanced chips. This is one of those stories where everyone says “controls are in place,” and nobody fully relaxes.
🔗 Read more
🤖 1X and EQT plan a massive humanoid rollout into factories and warehouses
Robotics company 1X announced a strategic partnership with investor EQT to make up to 10,000 NEO humanoid robots available across EQT’s portfolio companies. The target zones are the unglamorous-but-real stuff: logistics, warehousing, manufacturing, facility ops, even healthcare.
It’s framed as “robots working alongside people,” not replacement - the classic “superpowers” metaphor that’s a little corny, but also… kinda sticks. If they actually scale deployments safely, this could be one of those quiet shifts that turns out loud later.
🔗 Read more