👋

At a high level, you need the 4 things I describe here: bsky.app/profile/bide...

TBH, no vendor (us included) currently offers all 4 to a degree that I'd be fully satisfied with. But you get still value now until the tooling catches up.

First, pick/build an SDK instrument your key workflow.

Two years ago I wrote about how to compare 2 benchmark results (i.e. don't just compare two averages or 2 medians)

Read more here: blog.p-y.wtf/statisticall...

But why

Kinda overstated. 4 hours with 3 "Aufguss"(Infusion?) is more than enough :D

Agreed but Android 7 came out in 2016. If we assume (conservatively) that any android phone gets at least 2 years of software updates, then these phones must be around 10 years old.

Do phones even hold that long?

Is Android 7 still a thing? 😅

Are they though? To me, LLMs still seem way worse at writing code than skilled humans. They are a great assistant though.

Given that LLMs are stochastic parrots of their training data, I assume the simple reason is that LLMs have no idea about machine code.

Also the output is not reliable enough to omit the human in the loop, so do we really want that?

Yeah I think this will happen at some point. If you need to debug, you just convert from machine code to whatever language you want. Meta has this: ai.meta.com/research/pub...

Well that and wiring up DI + Nav + opinionated MVVM on Android requires a ton of know-how and is easy to get wrong.

It has become easier with Compose but there's more than a handful of foot guns along the way. Adopting a thing that takes that off your shoulders is a blessing for medium size teams

Afaik there is no real problem for circuit given that it uses lambda memoization.

The underlying article about data classes can be true regardless.

I was just curious if the team felt restricted by the very opinionated structure of Circuit

How was the experience in the team?

Simple and effective framework, love it!