How much does prompt bloat matter with Claude?
Heat trend
Collecting trend data
The percentage is based on available heat signal, not comment count or independent people.
A user on Reddit's dev_community is questioning the impact of "prompt bloat" when using Claude. They note that prompts have become unintentionally long due to continuous additions of instructions to prevent unwanted behaviors or address edge cases. While these lengthy prompts still work well, the user is concerned about sending unnecessary context, especially for frequently run tasks. They are seeking advice on whether trimming down mature prompts or comparing them to smaller versions affects quality or token usage.
Hi guys
I have been going back through some of the prompts we use with Claude and realized a few of them have gotten way longer than I remembered and it wasn’t really intentional either like usually Claude does something we don’t want, we add an instruction to prevent it next time then a few weeks later there’s another edge case and another instruction gets added and eventually you end up with a massive prompt where we aren't sure which parts are doing anything useful.
They still work well so I’m hesitant to start deleting things just for the sake of making them shorter but I’m also wondering how much unnecessary context we’re sending over and over again (especially for stuff that gets run pretty frequently).
Have any of you guys gone back and trimmed down mature prompts or compared them against a much smaller version and I wanna know whether you noticed any difference in quality, token usage or both.