@Levinarcc@vozercozer I agree that it was super impressive, it's just funny seeing everyone talk about optimization as if there were no drawbacks to having such limited amounts of RAM
@madsbrodt@thdxr Agreed this is exactly where I'm at too. Plus I feel like users can clock AI user interfaces pretty easily and I don't want to ship that.
@dexhorthy Yes yes yes. I've let AI models run wild in various code bases and they crush it in vertical slice repos. They just add the slice and wire it up in main, much easier to review. I don't even have to explain vertical slices it just sees that's what I'm doing and follows the pattern
@JustJake The literal Claude Code product gatekeeps it. I'm trying to run agentic loops and it just stops halfway through and asks me for input about stuff it can find on its own.
@unclebobmartin Sometimes it depends on the domain.
I work on Minecraft servers and it almost always tells me to STOP making architectural improvements because "it's just a Minecraft server bro".
Like no, I have a lot of paying players and there's a lot at stake. This needs to scale!
@ccallac7 Yesss it's like how Minecraft servers attempt to spawn monsters every single tick by default. I lowered ts to like 100 ticks and my server performed so much better
@ItsKieranDrew I rarely outsource my writing for this exact reason. Having to write clear and concise docs stimulates the hell out of my brain, way more than coding even. Plus no one wants to read AI slop.
@GuyFronten13778@housecor In my experience, sometimes it does miss those extra things. I wish it didn't. Then I have to prompt again or go into the code manually anyways.
But to your point, sometimes it's worth it since I can go do something else while the AI runs.
@rickyfm this is hilarious when you realize that prettier basically is a non-magical, explicit function that you need to pass your code into in order to get formatted output