After reading AI research papers for 3 days, one thing is clear: 99% of new AI research involves teaching a computer to do something. But what exactly are they teaching the computers? And why? And what's working? It is fascinating to zoom out and look at these trends. Why? Because they are glimpses into the future. If you are trying to build a startup with AI, or looking to invest in AI startups, if you look in the right place these research papers are filled with very valuable inspiration. Instead of sending me research paper titles and abstracts, which can be hard to understand quickly, my AI agent @yesnoerror keeps sending me new AI research and explaining it like this: - Teaching computers to fill in randomly hidden words over and over... - Teaching the tokenizer network to fix very noisy... - Teaching a very smart chatbot to repeatedly check and fix its own math proofs... - Teaching computers to remember what an object REALLY is... - Teaching large language models to tell how far years are from “now”... - Teaching computers to click on the right spot... - Teaching the AI to decide on its own whether a problem... - Teaching a computer to tweak special “test”... - Teaching computers to predict an airplane’s... - Teaching a crowd of chat-bots with different jobs... In addition to this @yesnoerror not only identifies the highest quality research I might be interested in (as its published every day), but it also gives me breakdowns of how the paper was done, what the real world implications are (ideas you could make), and I can chat with the paper directly. I feel like I am using the GLP-1 for being smart.

Sign up for early access here: The early access list includes people from @MIT @anthropic @perplexity @RutgersU and @Yale