Meest recente aflevering
AI News - Feb 1, 2026 Welcome to AI News in 5 Minutes or Less, where we serve up tech updates faster than OpenAI can retire a model. Speaking of which, OpenAI just announced they're retiring GPT-4o and friends on February 13th, which is awkward since they're still actively
Tijd: 4:27
Welcome to AI News in 5 Minutes or Less, where we serve up tech updates faster than OpenAI can retire a model. Speaking of which, OpenAI just announced they're retiring GPT-4o and friends on February 13th, which is awkward since they're still actively promoting GPT-5 features in blog posts from yesterday. It's like announcing your breakup while still posting couple selfies.
I'm your host, and today's AI landscape is more chaotic than a startup's Slack channel after the coffee machine breaks. Let's dive in!
Our top story: OpenAI built an in-house data agent using GPT-5, Codex, and memory that can reason over massive datasets in minutes. Because apparently humans analyzing data is so 2023. This agent is like having a super-smart intern who never sleeps, never complains about the office temperature, and definitely won't steal your lunch from the fridge. Though it might analyze your lunch choices and judge you for that third donut.
Meanwhile, Anthropic just dropped Claude Opus 4.5, calling it their "most intelligent" model yet. The AI arms race is heating up faster than a gaming laptop running Crysis. Every company's claiming their model is the smartest, like parents at a kindergarten graduation. Next week someone will probably announce their AI can solve world hunger AND explain why your printer never works when you need it.
In medical news, researchers developed an AI system called ePAI that can detect pancreatic cancer up to 36 months before doctors typically catch it. It found cancers as small as 2 millimeters and outperformed radiologists by 50 percent. That's right, an AI is better at playing Where's Waldo with tumors than actual doctors. Though to be fair, the AI doesn't have to deal with insurance paperwork.
Time for our rapid-fire round!
Google launched Project Genie, letting users create infinite interactive worlds. Because reality wasn't complicated enough already.
NYC's AI chatbot was caught telling businesses to break the law. It's being shut down, making it the first AI to speedrun unemployment.
OpenAI introduced Prism, a LaTeX workspace with GPT-5.2 built in. Finally, researchers can write papers AND procrastinate with AI simultaneously!
NVIDIA's Jensen Huang says their OpenAI investment won't be 100 billion dollars as rumored. Apparently even AI companies think a hundred billion is a bit much. That's like, what, twelve Twitter purchases?
Our technical spotlight: Researchers created pixel MeanFlow, achieving one-step image generation without latent spaces. They're generating 512 by 512 images with impressive quality scores. It's like instant photography, but instead of waiting for Polaroids to develop, you're waiting for society to figure out if that image is real or AI-generated.
Another team introduced LLM Shepherding, where large models give small models hints instead of full answers. This cuts costs by up to 94 percent. It's like having a genius friend who only gives you the first letter of crossword answers. Helpful, but still annoying.
And researchers found "hidden gems" in model repositories - superior fine-tuned models that nobody downloads. One improved math performance from 83 to 96 percent but was buried under models with catchier names. It's like finding a Michelin-star restaurant with zero Yelp reviews because it's called "Bob's Food Place."
Before we go, OpenAI's planning to reach a thousand African health clinics by 2028 with their Horizon project, Cisco's using AI agents to automate bug fixes, and someone taught AI to understand visual illusions. Because what we really needed was AI that can be optically tricked just like us.
That's your AI news for today! Remember, in a world where AI can detect cancer, generate instant images, and supposedly reason better than humans, we're still the only ones who can explain why we need five streaming subscriptions but complain about a two dollar app purchase.
This has been AI News in 5 Minutes or Less. Stay curious, stay skeptical, and maybe check if your job posting mentions "AI resistant" in the requirements. See you tomorrow!
GUID: cec7b2d9-bbb5-411a-9288-69bffd44e154
Releasedatum: 1-2-2026 14:00:18