Most of the AI in the news today runs on a single approach: large language models. ChatGPT, Claude, Gemini — all variations on the same underlying idea. They’re impressive. They are also not the only path forward, and some of the field’s most influential researchers think they’re not the right path forward.
This page tracks the technical frontier — what’s being built, what’s being questioned, what’s coming next. The goal is to help people who aren’t AI researchers see past the hype cycle to what’s actually happening underneath.
Beyond the language model
Yann LeCun — Meta’s chief AI scientist and a Turing Award winner — has spent the past several years publicly arguing that large language models alone will never produce real intelligence. He’s now raised a billion dollars to build something different: AI systems based on “world models” rather than next-token prediction. Whether he’s right or wrong, the bet itself tells you something important about where parts of the research community actually think we are.
More research deep-dives to come.