It’s become my go-to for parsing logs, cleaning CSV columns, and wrangling JSON that’s more tangled than a headphone cord in a pocket. Everyone’s busy asking it to write sonnets or debug their Kubernetes configs, while I’m over here pasting a 200-line Apache log and saying, “Give me a pattern that extracts the timestamp, the user agent if it contains ‘Mobile’, and the status code, but only if the request took over two seconds.”
Maybe I’m just old-fashioned, but most of the “creative” or “strategic” output still requires so much fact-checking and rework that it’s not worth the hassle. But for regex? It’s oddly reliable. It explains the capture groups, it usually gets the edge cases right, and it saves me from the soul-crushing cycle of writing, testing, and failing on regex101.com for an hour.
I suspect this says more about the state of generative AI than my own habits. It’s fantastic at syntactic tasks—reformatting, pattern matching, templating—where the rules are clear and success is binary. Ask it for a “strategic roadmap for a Series A martech startup” and you’ll get a platitude salad. Ask it to write a regex to find all ISO 8601 dates in a text, and it delivers something you can actually use.
So, am I the outlier here, or is anyone else just using these tools as a very expensive, overqualified text parser?
Show me the data
Totally get this. I use it the same way for transforming weird API responses before they hit my staging tables. Last week I had a nested JSON where some values were arrays and some were strings - it gave me a jq filter that actually worked on the first try.
But I've noticed it sometimes messes up on more complex CSV patterns, like when you have escaped quotes within quoted fields. Still have to test everything.
What tool do you usually run the regex in once it's written? Back in your code, or do you test elsewhere first?