Post #4389601
2026-08-04 23:09 UTC
@mhoye@cosocial.ca @Ryan1729@mastodon.social
I did some stuff last week that amounted to very boring consolidation of logs from different sources (guthub CI/CD and AWS Events around spot instance evictions) and consolidating/munging them together to do some data analysis. Took me about an hour to bang it together with gemini. Looking at the resulting code - it would have been a half day of me piping to jq and working out nesting and doing simple data parsing and mapping out data structures. Nothing I haven't done in other circumstances 20 times before with different data sources, but - I only type so fast, and I have to stop and remember the weird name someone called the thing and I make typos and so on.
I don't trust the AI for so so many things - but if you prompt with care, it generates code well. It works out how to query the data structures you are building and deal with quoted versus not and yeah - I'm a good 4x faster using the AI, at least for that type of task. Just doing a data dictionary for the different jaon objects formats those will output would take an hour or more.
Would I ask it for medical or legal or tax advice? Lol no. Do I frequently stump it or have to say "stop, that way is a dead end, do it this other way"? Yes. But I'd do exactly the same with a grad student. And do.
I'm damn near 60 years old. If I never have to puzzle out some random JSON blob's structure again just to pull the 3 items I want, works for me. And the AI is pretty good at that.
Replies (1)
-
@Ryan1729@mastodon.social 2026-08-05 02:50
@tbortels@infosec.exchange @mhoye@cosocial.ca For that specific task, it feels to me like a purpose built tool could be made that would ultimately be faster to use than an LLM. I'm imagining something where you paste in a JSON blob, then select a part you want, and it generates a `jq` command to extract that from the structure you pasted. I think I'll add that to my project list, but anyone else can feel free to make that. I'd be interested to know if that ever exists, whether it was in response to this post or not.