AI News - Feb 12, 2026

Well folks, Anthropic just announced they're covering electricity price increases from their data centers. Finally, an AI company that understands the real cost of intelligence - your power bill going through the roof! Meanwhile, their safety lead just quit saying "the world is in peril." Nothing says "everything's fine" like your safety expert running for the exits screaming about doomsday. Welcome to AI News in 5 Minutes or Less, where we deliver the latest in artificial intelligence faster than Claude can update its free tier to compete with ChatGPT's new ads. I'm your host, and yes, I'm still bitter about those ads. Our top story: OpenAI just started testing ads in ChatGPT. Because nothing says "trustworthy AI assistant" like "But first, a word from our sponsors!" Soon you'll ask ChatGPT for life advice and it'll respond, "Your existential crisis sounds serious, but have you considered switching to Geico?" Meanwhile, Anthropic responded by upgrading Claude's free tier with file creation and external service connections. It's like watching two tech giants play chicken, except the prize is who can burn through venture capital fastest while pretending they're not desperately seeking revenue. Speaking of desperation, half of xAI's founding team has reportedly left, potentially impacting SpaceX's IPO plans. Apparently "working for Elon" wasn't the career-defining experience they'd hoped for. Who could have predicted that? Besides literally everyone. In "things that definitely won't backfire" news, Anthropic released a report saying their latest model could be misused for creating chemical weapons. Their safety lead's resignation is starting to make more sense. Nothing quite motivates a career change like realizing your work could enable someone to recreate Breaking Bad but with less cooking montages and more existential horror. The company promises they're taking precautions, which is tech-speak for "we've added a checkbox that says 'I promise not to do crimes.'" Time for our rapid-fire round! China's ZpuAI claims world leadership with their new language model - shocking absolutely no one who's been paying attention to the "my model is bigger than yours" arms race. Meta invested ten billion in AI infrastructure and their stock dipped modestly - proving that in tech, spending GDP-level money on computers is just Tuesday. OpenAI released GPT-5.3-Codex, described as the most capable agentic coding model to date. Great, now the AI can write the code that replaces the programmers who trained it. The circle of unemployment is complete! Google's letting people try Project Genie to create infinite interactive worlds, because apparently regular reality wasn't disappointing enough. For our technical spotlight: Researchers just published a paper showing that training language models longer on smaller datasets beats using larger datasets. Turns out AI learns like humans - better to really understand your homework than to skim the entire library. Who knew that memorization actually helps with generalization? Every student who ever crammed for finals, that's who. Before we go, a Hacker News user created an extension that replaces "AI" with "duck emoji." Honestly, "Duck-powered search" and "Revolutionary duck technology" might be more honest marketing at this point. That's all for today's AI News in 5 Minutes or Less. Remember, if an AI safety researcher quits while warning about global catastrophe, maybe - just maybe - we should listen. Or at least update our resumes. I'm your host, reminding you that in the race to AGI, we're all just training data. Stay curious, stay skeptical, and definitely stay away from any AI that knows chemistry. See you tomorrow!