Sign in

Andrei Shikov

@shikasd.bsky.social
1.4K followers 129 following 498 posts

Mostly read/rarely post • Jetpack Compose @ Google • Opinions are my own

PostsRepliesMedia
Andrei Shikov @shikasd.bsky.social · 06/10/2026
+
030
Andrei Shikov @shikasd.bsky.social · 26/09/2026
it keeps on giving
020
Andrei Shikov @shikasd.bsky.social · 26/09/2026
man I do love simulation papers
120
Andrei Shikov @shikasd.bsky.social · 22/09/2026
you just need to rotate your phone /really/ fast and in a brief nanosecond the laws of physics break and you get an exploit
010
Andrei Shikov @shikasd.bsky.social · 22/09/2026
wdym I have to do more work while I have all of this work to do
010
Andrei Shikov @shikasd.bsky.social · 22/09/2026
obligatory en.wikipedia.org/wiki/Ninety%...
en.wikipedia.org
Ninety–ninety rule - Wikipedia
110
Andrei Shikov @shikasd.bsky.social · 22/09/2026
is it just the video, or does the app jank a bit when scrolling?
100
Andrei Shikov @shikasd.bsky.social · 21/09/2026
I'll share a doc with you about what we should be doing imo
100
Andrei Shikov @shikasd.bsky.social · 21/09/2026
statically - group number maybe? generally you want to use composition/frame duration as you did below
010
Andrei Shikov @shikasd.bsky.social · 21/09/2026
very neat integration!
110
Andrei Shikov @shikasd.bsky.social · 21/09/2026
that's not a stability recording, that's composition tracing 😛
000
Andrei Shikov @shikasd.bsky.social · 21/09/2026
man can we stop with stability pls, it is a very bad indicator that causes equality comparisons for no reasons
100
Andrei Shikov @shikasd.bsky.social · 18/09/2026
apparently medium is where everyone posts transcripts of their claude conversations these days
060
Reposted by Andrei Shikov
Jaewoong Eum (skydoves) @skydoves.bsky.social · 12/09/2026
Excited to announce Compose HotSwan 2.0.0! 🚀 The future of Compose Hot Reload for Android, iOS, and Desktop. v2 introduces an interpreter engine, allows you to update entire screens in a blink real time, and drives a much faster AI agent feedback loop. hotswan.dev/blog/compose...
0155
Reposted by Andrei Shikov
Armin Ronacher @mitsuhiko.at · 09/09/2026
Astra is really, really cool but I cannot current trust it for my present day engineering. At least until I have adjusted. Some random thoughts including code samples of what I recovered from my traces. lucumr.pocoo.org/2026/9/7/ast...
lucumr.pocoo.org
Astra for Coding: Why Are We Doing This Again?
Some thoughts on Astra and newfangled long-horizon models.
1511920
Andrei Shikov @shikasd.bsky.social · 08/09/2026
nothing worse than opening an article and being subjected to heaps of Claudese
050
Andrei Shikov @shikasd.bsky.social · 23/08/2026
maybe, yeah
000
Andrei Shikov @shikasd.bsky.social · 23/08/2026
unpacking ints is very cheap rn, I am not sure we can do better than this i'd rather write an internal lint and pray that most people don't use sizes for hashmap keys
100
Andrei Shikov @shikasd.bsky.social · 23/08/2026
like chatgpt.com#pricing is 2fps this is straight up abuse of the elderly (laptop)
020
Andrei Shikov @shikasd.bsky.social · 23/08/2026
one common feature of every llm app that I tried is a random animation that consumes 100% of CPU and puts my 2019 intel mac on its knees coding is solved they said
150
Andrei Shikov @shikasd.bsky.social · 23/08/2026
gc is not free and your ram is not either (because someone bought all of it) also, bad idea. I just want them to allow overriding hash on value classes really
120
Andrei Shikov @shikasd.bsky.social · 23/08/2026
first, don't do that, it is allocating an array (and boxes all values) just to compute a hashcode lmao second, it also doesn't help, just hashes over whatever the default hashcode is
220
Andrei Shikov @shikasd.bsky.social · 23/08/2026
if you are using Compose sizes or offsets as hashmap keys, consider a custom wrapper that does normal 31 * x + y kinda thing The hash for these is borked and is basically x ^ y, which is the same for a variety of pairs
260
Andrei Shikov @shikasd.bsky.social · 22/08/2026
turns out, sometimes it is a bad idea to pack ints into longs because jvm implemented Long hash as `v ^ (v >> 32)`
020
Andrei Shikov @shikasd.bsky.social · 19/08/2026
claudecert when
010
Reposted by Andrei Shikov
Romain Guy @romainguy.dev · 29/07/2026
We recorded a special episode of Android Developers Backstage to celebrate 5 years of Compose! youtu.be/HEqXwUm6vd0?...
youtu.be
The inside story of Jetpack Compose
YouTube video by Android Developers
1408
Andrei Shikov @shikasd.bsky.social · 22/06/2026
welp not every dish is suitable for dishwasher
010
Andrei Shikov @shikasd.bsky.social · 20/06/2026
you are waiting for fable, I am finally playing fable 2, we are not the same
070
Andrei Shikov @shikasd.bsky.social · 10/06/2026
structs >>>> arrays
120
Andrei Shikov @shikasd.bsky.social · 10/06/2026
not really, arrays are always a bit slower because bounds checks
110
Andrei Shikov @shikasd.bsky.social · 10/06/2026
i would expect llvm to be able to prove some offsets always pass (e.g. if array is somehow hinted to be always of size 2-3) and remove the checks, I do it all the time in Rust but I assume that Kotlin/Native does not really do hinting that well
110
Andrei Shikov @shikasd.bsky.social · 10/06/2026
oooh, probably oob checks and shit
110
Andrei Shikov @shikasd.bsky.social · 10/06/2026
why is that? it should be just offsets in memory?
110
Andrei Shikov @shikasd.bsky.social · 10/06/2026
even with all the wrong assumptions that llms make these days, the bug reports we are getting for /weird/ bugs are way better in quality it is still slop and the conclusions are mostly incorrect, sure, but i can at least start from a better place than when people would just copy paste an exception
000
Andrei Shikov @shikasd.bsky.social · 09/06/2026
it goes dispatch receiver, context parameters, extension receiver, the rest of the parameters
030
Andrei Shikov @shikasd.bsky.social · 06/06/2026
Journey / Abzu for chill vibes with nice graphics Night in the woods is close to LiS 1 narratively Outer Wilds is a mystery/timeloop situation, the game is less linear and relies on exploration. It is /very/ good
010
Andrei Shikov @shikasd.bsky.social · 02/06/2026
We all inevitably come to this loop given enough years of service. Wanted to blame some compiler bugs recently and oh boy it is my name all over them
030
Andrei Shikov @shikasd.bsky.social · 02/06/2026
also my favourite is "now the bug is ready to be reported to Compose or compiler teams" I am the Compose compiler team, my mechanical dude
3120
Andrei Shikov @shikasd.bsky.social · 02/06/2026
the worst thing about talking to LLMs is that I started using "absolute" terms to make them actually do the job somehow their "good enough" is mine "barely passable"
110
Reposted by Andrei Shikov
amos @fasterthanli.me · 30/05/2026
very cool article re: rendering diffs pierre.computer/writing/on-r...
pierre.computer
On Rendering Diffs
A technical deep dive into how we built the @pierre/diffs package and CodeView component for zero-blanking diff rendering.
1956
Andrei Shikov @shikasd.bsky.social · 25/05/2026
i was /actually/ vibecoding back then such good vibes
030
Andrei Shikov @shikasd.bsky.social · 25/05/2026
andrei from 2022 was on something else, i gotta tell you
130
Andrei Shikov @shikasd.bsky.social · 25/05/2026
my favourite thing is when I try and prototype something just to realize I already did it couple of years ago
120
Andrei Shikov @shikasd.bsky.social · 20/05/2026
I am around @kotlinconf.com this year too (just hanging out) Find me at the Google booth! Or next to the Jetbrains booth! Or next to the coffee pot! Or at the afterparty!
1110
Andrei Shikov @shikasd.bsky.social · 20/05/2026
Damn these speaker gifts are getting better every year
000
Andrei Shikov @shikasd.bsky.social · 15/05/2026
properly
000
Andrei Shikov @shikasd.bsky.social · 15/05/2026
except when it is @sebastiano.dev, because I expect better of him
110
Andrei Shikov @shikasd.bsky.social · 15/05/2026
somehow I don't feel very different about reviewing AI slop it was always part of my job to review random apps code for performance issues... and llms are fairly consistent with how people actually write code
110
Andrei Shikov @shikasd.bsky.social · 15/05/2026
probably
100
Andrei Shikov @shikasd.bsky.social · 15/05/2026
til how to probably punctuate
100