Today's LLM Strengths and Weaknesses Trace Back to Component Ideas

KordingLab · x · 2026-08-26

UPenn's Kording Lab argues that a great deal of the strengths and weaknesses of today's LLMs can be understood from a set of component ideas. In a follow-up, the thread concludes: the Old Recipe was extended with 'inner speech' + tools + agents + better evaluation, producing today's interactively 'thinking' machines.

Original post →

More from AGI Musings

AGI Musings channel →