also, pretty sure my solve was dumb and solved the route more times than I needed to. oh well.

Time to solve with debug build of a Swift Package Manager cli tool project: 11 secs, 175 ms, 846 μs, 209 ns
Time when using the release build configuration: 1 sec, 741 ms, 706 μs, 292 ns

Holy crap. That's a big difference! #AdventOfCode

I guess it could've been an interesting exercise to produce a very basic BMP file and feed it in to a local LLM to see if it says it's a christmas tree.... but never done any ML stuff before, going to keep that idea in my pocket for when I am interested enough to use an image classifier in my code.

Yeah, I didn't even try to detect it, as soon as I saw part2 I hit reddit and saw a clue and resolved to just sit and watch the results evolve on terminal with Ctrl+C ready to go 🤣

I've been hitting up LLMs (ChatGPT and locally run ones) after I've solved each day to see how close they can get to a sensible answer. ie, would these things *actually* be help or hindrance in my day job? Enormously mixed results 🤣 Seen nothing yet to encourage me to use them for my actual work!

And when I say big hurdle, I mean it. Definitely not a replacement to your usual shell unless you’re prepared to completely rewire your terminal muscle memory 😂

That’s why I’ve been using nushell at home. Native support for json etc meaning no shelling out to jq to parse/extract data. It’s a different paradigm to [ba][z]sh though, so has a big hurdle. Also pre 1.0 release so gets a lot of breaking changes hence don’t use it for work stuff. promising though

Yeah. I’m way past the point of giving Python second chances. It’s a very good language and lots of people do some really great stuff with it, but it’s just “not for me”!

I’m just doing AOC in Swift this year, it’s what I do a lot of in work alongside C++ so just makes sense.