Also, there's a comment in there about AI agents handling on-call duties; and that idea both excites and terrifies me 😅

Also loading up on delicious baked goods today 😍

That's what has prompted my little emergency deep dive into the optimization. Our new CMP project is worked amazing so far for porting over some non-trivial functionality but for a still relatively small project we were already feeling the pain of linking.

Using Parallels is a really interesting idea when targeting those other desktop platforms 😀

Are there hoops to jump through to get it deployed onto those virtualized machines?

Are you using internal and @/HiddenFromObjC together generally? Or do you use a set of team guidelines/conventions to dictate when to use one or both?

Very interesting. I hadn't come across those before or looked into the linker much. Sounds like I should. Thanks for the tip!

"Avoid release at all cost" definitely seems like the right strategy.

This was a timely find in my feed. We've been feeling some of these build time pain points in our new CMP project.

The article confirms a lot of my suspicions about where to start optimizing but also some stuff I hadn't seen before.

Thanks for sharing!

I'm not sure if maybe we should avoid building a Release build type for the simulator target by default since most the time we'd just want the Debug build type 🤔

This has been a problem for us in our other KMP project, and nothing short of throwing more powerful CPUs at it seemed to help much. We eventually just introduced some build flags so we could run most builds against just the JVM target for local dev but our CI build times are still pretty awful.

Just avoid running any task that runs more than one linking task at a time? Maybe paired with doing most local dev again Android or Desktop first then test on iOS?