Current LLMs resemble kindergarten or elementary school students cognitively despite their ability to perform advanced tasks
A lot of the smaller models, or the dumber models, remarkably resemble a kindergarten student, or an elementary school student or high school student. Somehow, we still haven't graduated enough where this stuff can take over. My Claude Code or Codex, they still feel like this elementary-grade student.