'Any computable problem is a state machine — why is the world betting on autoregression?'

knowrohit07 · x · 2026-09-15

A post argues that any computable problem is fundamentally a state machine, yet the world — researchers included — is pouring money into autoregressive models to solve them, bluntly questioning whether everyone is missing the point. A typical skirmish in the recurring debate over the limits of the autoregressive LLM paradigm.

Original post →

More from AGI Musings

AGI Musings channel →