On the other hand it seems like LLMs' inability to generate perfect boilerplate, requiring me to go through and fix all the bugs and nits, is all that's keeping my "sharpness" at this point.