hazel★

@fractal.beauty

software engineer; local-first, programming languages, p2p, rust; fractal.beauty 💙🧡

we (i) have poor debugging tools (for react native). the most reliable (and tedious) method (that i know of) to find bugs is to recompile native code in dependencies etc with a bunch of logging. that can't be right

i'm generally hesitant to call LLMs AI, but it's true that LLMs with tool calling are something different, so maybe "agent" is actually a useful term

i don't understand pbt. in theory i get the appeal but i'm not sure how to apply it to the type of software i'm writing (~applications with interfaces? that are fairly impure?)

so I've been looking at my finances, and all my income sources have plummeted I'm gonna have to shift my focus away from #half-edge and giving talks, because unfortunately they're huge money sinks I can't afford if you can, I'd appreciate your support, especially now ❤️ www.patreon.com/c/acegikmo

Get more from Freya Holmér on Patreon

Get more from Freya Holmér on Patreon. Creating video games & media. Support Freya Holmér and get exclusive access to their work.

patreon.com

my main issues with worktrees: - duplicate build artifacts; would a shared content-addressed build cache help? otherwise working in an interpreted language probably works well - if you need to do manual testing it can be slow to switch (ex. building/installing a mobile app). u need a fast test cycle