Positive users highlight Claude's speed in finishing coding tasks despite padded estimates, while a few criticize its occasional lazy behavior.
Based on 4 visible X reactions from 7 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@rrhoover It's Retdatsfoh's Law: Hofstadter’s Law, but the reverse. "LLMs always take less time than they predict, even when they take into account Retdatsfoh's Law
@rrhoover Once you’ve shipped a few agentic workflows, you start factoring in Claude’s optimistic padding—still faster than most junior dev estimates though.
@rrhoover Claude Code when it is lazy:
@rrhoover Meanwhile, three minutes later, the agent completes the task.
@rrhoover It’s based on training data of engineers estimating project size. This is one of the most interesting quirks of modern AI models.
Claude Code clearly padding when it replies: "Implementation estimated to take 3 weeks" AI, they're just like us.
Update: It took 126 minutes
Positive users highlight Claude's speed in finishing coding tasks despite padded estimates, while a few criticize its occasional lazy behavior.
Based on 4 visible X reactions from 7 accounts; directional sample.
Ask a question below.
Published answers will appear here.