Many users criticized Claude for struggling to design multi-harness compatible AI systems because it errs outside its walled garden, makes on-brand mistakes with codex hooks, and gives inaccurate time estimates.
No Digg Deeper questions have been answered for this story yet.
Most Activity
another weird thing is how bad the time estimate is for task, like the thing that take a day probably take less than a hour
asking claude to design a system compatible with multiple harnesses (not just claude code) is a VERY painful experience, the model keeps getting other harnesses features wrong and keeps designing things with only claude code in mind
@eliebakouch LLMs are consistently mid at handling anything outside their specific walled garden.
@eliebakouch claude designing for claude code and then confidently getting the codex hooks wrong is so on brand for this problem
@eliebakouch training data bias. gotta paste the other harness docs into context every single time