arata retweetet
arata
1K posts
karpathy's augmentation vs automation
posted this on every other alt i had, but somehow not on this one yet. i first put it together as a remix of karpathy's slides, and it's somehow crystallizing into my design direction (think sci-fi hud or exocortex, not the ironman).
used to think augmentation is all i need - now i don't even think it's falls on a spectrum. it's more messy, like different modalities serve different constraints and needs. the switch between them has to feel seamless, and i still haven't solved it yet. and no one seems to be going with this direction despite it being the obvious progression to me (maybe i'm completely wrong about that).
all i'm trying to say is, the future isn't shaped like what you'd expect. and one thing i know for sure, an enterprise-grade chat experience is not for me: x.com/arata_ae/statu…
arata@arata_ae
sounds like claw and its variants to me, but karpathy said it’s different. i’m confused.
English
arata retweetet
How I turned 3d scans into an interactive firewood splitting simulator
English
supporting ~3 deployment types and feature sets (local vs cloud, web vs desktop permutation) is a nightmare. i have zero users right now - haven't even managed to ship once yet.
looks like i'll have to let go of my ideal local stack to get things moving and save on tokens (most of which are getting burned on constant refctoring).
just need to simplify. simplify.
English
skipping engineering practices has hurt me over and over. i keep cutting corners, skipping tests and instrumentations, not thinking things through, and every time it came back to bite me.
for some, it's second nature, but as someone who's been vibe coding a lot lately, i have to be deliberate about practicing these.
English
arata retweetet
『鳴潮』先駆ラジオスタートーチ卒業記念ソング——絶やさぬスタートーチ（卒業合唱 Ver.）
卒業おめでとう。遥かな空へ……飛び立って。
プロダクション：鳴潮
コーラス：Lychee ACG Choir
プロデューサー：Gigawatts Productions
作曲：Linda Ke
作詞：パンダの流儀
アレンジ：David Soltany
ピアノ：Laura Strubel
ベース：JF Untrau
ストリングス：Vienna Synchron Stage
ストリングス指揮：Bernhard Melbye Voss
オーケストレーション：Mathieu Alvado
合唱指導：HE Jiamin (Miette)
レコーディングエンジニア：PARTY UP Studio
音声編集：Clémentine Charuel
ミックス：Samy Cheboub
マスタリング：John Webber
音楽監督：Steven Tang, SmileL, Linda Ke, Grass
#鳴潮
日本語
context carryover is a must-have for me. i hate having to spell things out slowly with current stt models. looks like this model is in livekit?
voice focus feels like magic tho. is it like diarization + filtering out irrelevant speakers using the previous context (similar to context carryover)?
English
arata retweetet
Introducing HIVEMIND: the Social Intelligence.
Watch the film, based entirely on true stories from @jasonyuan, @biz, @supercgeek, @Reshusaur, and more...
English
arata retweetet
ok, ima give it another shot with this codemod: github.com/skovhus/styled…
via: x.com/kenneth_skovhu…
Kenneth Skovhus@kenneth_skovhus
New post: why and how @linear is moving to @stylexjs. Codemods and agents are doing the heavy lifting. Humans still have to perform detective work on cursed hover states. skovhus.dev/blog/moving-li…
English
i tried migrating my design system from styled-components to stylex using codex and it made a lot of mistakes, even for a button (though tbf a button is one of the most complex components of all, anyway..).
now i get why cursor did it with fable..
some of my components are quite complex since i like using cutting-edge stuff and pushing styled to its limits. migration is almost impossible without fable-class level, i think.
English