AlexUK

In April, a website that has been sued, blocked, deplatformed, and chased across thirty-seven domains over fifteen years quietly launched its own AI. Sci-Hub is the largest unauthorized library of scientific papers in human history. Ninety-five million academic papers. Tens of millions of books. Built and maintained by a single Kazakhstani neuroscientist named Alexandra Elbakyan since 2011, funded by donations, hosted on whatever country's registrar will tolerate it that year, mirrored across torrents and IPFS and Telegram bots. Elsevier sued. Sci-Hub stayed up. The American Chemical Society sued. Sci-Hub stayed up. India sued. Sci-Hub stayed up. Swedish registrar Njalla cut the .se domain in January. Sci-Hub stayed up at .al, .ru, .ee, .box, and a half-dozen .onion addresses the registrars cannot reach. Now the library has built its own intelligence. Sci-Bot launched in alpha in April. You ask it a research question. It answers, and it cites real papers from inside the corpus, with links that actually open the actual papers. The bot does not hallucinate citations. It cannot, because it only draws from papers it actually holds. The same property that the venture-funded labs have spent four years and forty billion dollars trying to engineer back into their products is a free side effect of training the model on a library that contains the books. Anthropic, OpenAI, Google, and Meta have all been sued in the past eighteen months for training their models on the same shadow libraries that Sci-Hub assembled. Meanwhile the corpus those scripts were pointed at, the corpus those models were trained on, the corpus the entire generative AI industry is built on, sat right there the whole time, free, with a search box on top. The pirates beat them to it. Sci-Bot was built on a corpus that was already free, by a team that asked no permission, charging no one, with the explicit position that the right to read scientific research is older than the cartel that decided to charge for it. The same arithmetic the medieval guilds used to keep the printing trade in approved hands. The same arithmetic Pope Paul IV used in 1559 to publish the Index Librorum Prohibitorum. The same arithmetic the Stationers' Company used in seventeenth-century London. Knowledge has always had a fence around it. The fence has always been guarded by men who did not write the books. The library answers. We never asked permission. We never had to.

THE ENTIRE AI INDUSTRY JUST GOT HUMILIATED a tiny model trained in just a few hours on a single graphics card is planning 48x faster than billion-dollar supercomputers. It actually understands physics instead of just memorizing patterns. yann lecun was right the whole time for three years every major lab told you the same story. scale is all you need. just throw more GPUs at it. just train on more tokens. eventually the model will "wake up" and understand the world. it was a lie. or at minimum, a very expensive bet that just lost. LeCun kept saying generative AI is a dead end. predicting the next pixel or the next token is fundamentally wasteful, the model burns trillions of parameters memorizing surface details instead of learning how reality actually works. he proposed JEPA instead. predict abstract concepts in a compressed thought space. don't paint the world pixel by pixel, understand it. the problem was JEPA kept collapsing. left to its own devices the model would cheat, mapping a dog, a car, and a human to the same point in latent space. technically minimizes the loss. learns absolutely nothing. every fix was ugly. seven loss terms. frozen encoders. EMA tricks. stop-gradients. the kind of duct-tape engineering that should have been a red flag. then LeCun's team dropped LeWorldModel. they replaced all the hacks with one regularizer that forces the latent space into a gaussian distribution. the model can no longer cheat. to make accurate predictions it has to actually encode physics. 15 million parameters. single GPU. trains in hours. plans 48x faster than foundation world models. detects physically impossible events on its own. meanwhile OpenAI is raising another $40B to train GPT-6 on a data center the size of manhattan. the entire scaling thesis just got embarrassed by a model that fits on a gaming PC.

FIRST ON FOX: President Trump has directed federal agencies to begin the process of identifying and releasing government files related to UFOs and extraterrestrial life. The move follows "tremendous interest" from the public and promises a level of transparency into unidentified aerial phenomena never before seen from the U.S. government. The administration’s push for disclosure has drawn bipartisan support, with officials noting that the American people deserve answers to decades-old questions regarding potential non-human activity.