Ten years ago I wrote my own regex for log parsing, now I just ask the model to do it
Back in 2014 I spent a whole weekend at my kitchen table in Lansing building a regex to pull timestamps out of nginx logs. I had a legal pad full of crossed out patterns and I finally got it working around 2am on a Sunday. For years that stupid 80 character regex followed me from job to job, and I broke it at least four times whenever the log format shifted. Last month I pasted 200 lines of messy server output into a model and asked it to give me the columns in order, and it spit out a clean table in about 9 seconds. I did not type a single bracket or backslash. The part that gets me is I no longer even care how it works, I just check the output and move on, which feels a little wrong but also saves me hours. My old regex still lives in a gist I refuse to delete. Anyone else hanging onto a tool they built by hand that AI just wiped out in one sitting?
Oh man, I read a piece last year about how coders are basically becoming editors instead of writers, and your gist story nails that feeling exactly. I have a bash script from 2016 that renames files and I still run it just out of habit even though I could ask a model in two seconds. It is weird to not care how the thing works anymore, but if the table comes out clean, who cares how the sausage got made.