willcb: Everyone building their own RL envs is training wheels for toddler AGI

willcb · x · 2026-11-17

willcb argues that advice like "everyone should build their own RL envs and tune hyperparameters" is anti-AGI-pilled in the same way as "set up your agent workspace and install the best skills" — both are training wheels for toddler AGI, scaffolding to outgrow rather than a long-term path, with continual learning as the real direction.

Original post →

More from AGI Musings

AGI Musings channel →