A bitter-lesson debate over whether LLMs belonged inside systems all along

lambdaviking · x · 2026-08-04

The reply argues that some “scaling-pilled” people used to oppose putting LLMs inside larger systems on bitter-lesson grounds, and says that history is being rewritten now that the same camp is wrapping open-source models into systems and calling it neurosymbolic.

It is essentially a debate about whether system-level LLM use was always part of the plan, or whether the current narrative is retrofitted after the fact.

Related event: LeCun: Strong Code Generation Systems Are More Than Pure Autoregressive LLMs(5 posts)→

Original post →

More from AGI Musings

AGI Musings channel →