kalomaze's Empirical Account: Opus 5.5 Is Bad at Shallow Instructions
kalomaze · x · 2026-11-21
Open-source training researcher kalomaze tweets that people sometimes discuss Opus 5.5 as if it lacks observable basic problems — and that he was recently accused of being an OpenAI shill. He shares an empirical account arguing the model is "kinda bad at listening to even shallow procedural instructions," with a link to the write-up.
Related event: Researcher Reports Opus 5.5 Fails Basic Instruction Following(2 posts)→
More from Models
- Model candidly admits it picked outdated MoGe-2 because MoGe-3 postdates its training cutoff — eigenron · 2026-11-21
- Redditors tout uncensored open models: DeepSeek V4.1 Flash runs 200 tok/s and never says no — Ontopoftheworld_ay · 2026-11-21
- NDIF offers free white-box access to Inkling 2 Mini for activation patching and steering — simonguozirui · 2026-11-21
- Inkling 2 mini impresses as a local model on M5 Max, token-efficient at low effort — LiTianleli · 2026-11-21
- Users push Google to open Gemini subscriptions to third-party agent harnesses — WolframRvnwlf · 2026-11-21
- Reddit user buys $200 Claude Max for $29, wonders how resellers profit — mehdiweb · 2026-11-21