Ethan Mollick: AI Lab Apps Are Buggy and Undocumented, Yet the Models Keep Everyone Hooked

emollick · x · 2026-10-01

Ethan Mollick observes that every AI lab app ships with weird edges, missing docs, surprise updates, no changelogs, and UIs that change overnight — yet everyone keeps using them because the underlying models are that good. His point: a broken early AI product is still often better than a well-built traditional one, since LLMs are unreasonably effective general tools and people have learned to roll with the rough edges.

Related event: Wharton's Mollick: Frontier Models Make Half-Baked AI Products Work(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →