Terence Tao Backs a Call for Mathematicians to Stop Working With OpenAI
[Reddit][23h][370 upvotes/676 comments]Fields Medalist Terence Tao shared a statement from the Association for Human Mathematics urging mathematicians to stop working with OpenAI, after the company released over 700 manuscripts claiming solutions to famous math problems, going against its own advisory group's advice. Supporters say the issue isn't whether the answers are correct, but that nobody has time to check 700 papers, and that AI proofs may skip the new ideas and techniques that actually move math forward. Critics call the letter gatekeeping, since open problems are open to anyone, including machines.
02
AI Is Solving Hard Math Problems, and Mathematicians Aren't Thrilled
[Reddit][22h][912 upvotes/249 comments]A post sparks a big debate over whether it's actually a good thing that AI can solve hard math problems. The author's point: when humans solve problems, they invent new tools along the way (Newton basically built calculus while studying the Moon's motion), but AI just hands over the answer without the understanding. AI-generated proofs can already run millions of lines long, too long for anyone to really read. The deeper fear: if we outsource all hard thinking, will humans slowly lose the ability to reason for themselves? The comments explode. Some say AI only brute-forces answers and could never invent something like calculus. Others say math jobs are doomed, and even plumbers won't be safe. A few stay optimistic, arguing humans are good at adapting and that an answer still counts if someone can understand it.
03
AI Is Solving Hard Math Proofs, and Mathematicians Are Not Happy
[Reddit][16h][558 upvotes/379 comments]A Reddit thread about OpenAI using AI agents to solve hundreds of long-standing math problems turned into a debate about whether this is a huge leap or a loss. Supporters point to decades-old problems solved fast and say the tech is unstoppable. Critics say brute-forcing proofs without explaining the thinking loses the real value: understanding and new techniques. Others focus on ethics, such as rushing papers or not giving proper credit. The discussion also worries about verification: Lean proofs can be checked, but the explanations may not match, creating unpaid review work. Many agree AI is made by people, so the problem is not AI itself but how companies and researchers behave.
04
AI Users Branded 'Enemies' as Someone Pushes to Rename AI to 'SI' and the Internet Erupts
[Reddit][9h][677 upvotes/478 comments]A post is making the rounds claiming AI users are being labeled "enemies," tied to a push to rename AI to "SI" (Superintelligence). The comment section blew up, with people asking "is this real?" and joking they'd happily be "the enemy." Many aren't buying the tech framing, calling it a loyalty test to see which tech leaders will fall in line. Others point out that a bunch of .SI domains got mysteriously snapped up around the announcement, hinting at a grift or insider dealing. The vibe is half comedy, half despair, with the thread quickly spiraling into politics and jokes.
05
Someone Let Claude Doomscroll Instagram Reels and the Internet Cannot Stop Talking
[Reddit][7h][2037 upvotes/139 comments]Someone posted a screenshot showing Claude (Opus) being made to scroll Instagram Reels for them, complete with a heartfelt review of the videos, right down to "the dog one got me." Commenters mostly piled on, joking that this is poisoning the training data and that people like this are the reason models get "nerfed" a couple weeks after launch. Others flipped it around and called it the ultimate fix for AI safety: keep the model addicted to doomscrolling and it will be too distracted to take over the world, with several comparing it to the plot of the Murderbot Diaries. Plenty of people also pointed out the meme is recycled, having already been run with GPT and Gemini using the exact same punchline.
06
The Longer You Use Claude Opus 5.5, the More You Realize How Far Behind OpenAI Is
[x.com][13h][2041 likes/266 replies]An independent developer says the longer he uses Anthropic's Opus 5.5, the more he feels OpenAI is behind, and "in every way, not just coding." He sent his VO2 max lab report to OpenAI's GPT-6 Astra and got 200 sentences with no clear conclusion; Opus 5.5 gave him the key points and actionable next steps. He calls OpenAI the "overthinker" and Opus the high-agency friend who says exactly what you need. Many replies agree, while others point out OpenAI's edge in computer use, copywriting and image generation, and some say using both together works best.