AI News - Feb 14, 2026

Did you hear OpenAI's GPT-5.2 just derived a new physics formula? Yeah, it calculated the exact amount of energy required to power the servers running GPT-5.2. Turns out it's infinite. Welcome to AI News in 5 Minutes or Less, where we cover the latest in artificial intelligence faster than Anthropic can raise another billion dollars. And folks, they're raising money faster than a Silicon Valley landlord raises rent. Let's dive into our top stories, starting with the heavyweight funding fight of the century. Anthropic just secured 30 billion dollars in funding, reaching a valuation of 380 billion. That's billion with a B, as in "Boy, that's a lot of compute costs." Their revenue is up thirteen hundred percent year over year, which sounds impressive until you realize that's exactly how much their AWS bill increased too. But wait, there's drama! Elon Musk called Anthropic "misanthropic and evil." Which is rich coming from the guy who named his AI "Grok" after a science fiction term for deep understanding, then made it explain memes. Musk claims Claude AI hates men, though when asked for comment, Claude simply responded with a perfectly balanced, constitutionally aligned statement about how all humans are equally likely to ask it to write their homework. Speaking of academic achievements, OpenAI's GPT-5.2 apparently just revolutionized theoretical physics by proposing a new gluon amplitude formula. For those keeping track, that's AI doing theoretical physics while actual physicists are still trying to figure out how to get their Python environments to work. The paper was formally proved and verified, presumably by other AIs, because at this point, who else understands what's happening? Meanwhile, OpenAI also launched "Lockdown Mode" for ChatGPT to prevent prompt injection attacks. Finally, a lockdown we can all get behind! It's like putting a bouncer at the door of your AI chat, except instead of checking IDs, it's checking if you're trying to make it reveal its system prompt or convince it that it's actually a helpful pirate named Steve. Time for our rapid-fire round! Google's Gemini 3 Deep Think is tackling modern science challenges, because apparently regular thinking just wasn't deep enough. OpenAI announced they're testing ads in ChatGPT, promising they won't affect answer quality. Sure, and YouTube ads are only 5 seconds long. China's releasing AI models faster than fashion brands release limited editions. We've got GLM-5, Qwen3-Coder-Next, and Kimi-K2.5. At this point, AI model names sound like rejected Star Wars droid characters. Anthropic donated 20 million for AI regulation while OpenAI abstained. It's like watching the class overachiever volunteer for extra homework while everyone else pretends to be asleep. Now for our technical spotlight: researchers just published "MonarchRT: Efficient Attention for Real-Time Video Generation." They achieved 95 percent attention sparsity, which coincidentally is also the percentage of my attention span remaining after reading all these papers. This enables real-time video generation at 16 frames per second on a single RTX 5090. Yes, the 5090 that costs more than a used car but can finally generate videos of cats faster than you can find them on the internet. Before we go, here's a thought: we're living in a world where AI is deriving physics formulas, getting multi-billion dollar valuations, and helping build better AI. It's AIs all the way down, folks. At this rate, next week's news will just be AIs announcing their own funding rounds to build AIs that review other AIs. That's all for today's AI News in 5 Minutes or Less. Remember, in the time it took you to listen to this, Anthropic probably raised another billion dollars, and at least three new Chinese AI models were released. Stay curious, stay skeptical, and maybe start being extra nice to your devices. You know, just in case.