Yuchen Jin: Opus 5.5 underwhelms, frontier LLM coding has plateaued

Yuchenj_UW · x · 2026-09-23

AI researcher Yuchen Jin shared first-day impressions of Opus 5.5 and came away unimpressed: it got a tricky research question wrong that Astra answered correctly, and Astra still feels better for coding.

He doubles down on his earlier take that frontier LLM coding capability has plateaued, arguing the real battle is now about intelligence per dollar rather than raw capability.

Original post →

More from Models

Models channel →