Research Engineer @ Google. Progenitor of Jetpack Compose (May 2017), Compose Multiplatform. Now: AI research @GoogleDeepMind, teaching AI to write code.

Joined May 2019
68 Photos and videos
Pinned Tweet
28 Jul 2021
More than four years after I first proposed it and wrote the first line of code, we've finally released #JetpackCompose (stable) version 1.0.0! 🎉 Incredibly grateful to everyone who helped me make it happen. Exciting to have built the future of #Android together!
71
164
1,173
Jim Sproch retweeted
The leading theory is that I'm actually just an AI, and I've allocated all my GPU compute for code generation instead of video rendering.
2
1
13
511
Super interesting UX paradigm for the future. Making UIs that feel 3d by tracking the user's position relative to the device and adjusting what gets rendered accordingly. Obviously this scene is extreme, but a more subtle version could be built into existing design systems.
Head-tracked “Window Mode.” Your front camera finds your head. The view reprojects in real time so the screen feels like a window into the 3D scene. True3D, no glasses.
2
18
2,407
We're already at the point where the majority of text being generated by LLMs is never being looked at by a human. It's being used to drive tool calls that do work behind the scenes.
7
1,297
Video was an incredible way to learn about the world. To gain an intuition about physics, animal behaviors, and more. The intuitions of the next generation are going to be so screwed up, as essentially everything they'll have seen is a hallucination. x.com/rd_mech/status/1973038…

30 Sep 2025
Community note
2
11
2,340
11 Aug 2025
LLMs are a brilliant way to test usability. LLMs are literally average users distilled into an automated form. They make the same kinds of mistakes assumptions as real humans. You can radically improve usability of your app by fixing failures encountered by an LLM using your app.
3
15
1,779
It turns out that using LLMs is much more environmentally friendly than the alternative (ie. human labor). 😉😜 nature.com/articles/s41598-0…
1
4
20
2,268
13 Jul 2025
Inspiring Animation; built with Compose! Building animations like this is a great way to uplevel your Compose skills and demonstrate your abilities. Especially the matrix animation, leaves lots of room for interpretation and creativity. This one is very well done!
I created this amazing MATRIX RAIN EFFECT using Jetpack compose. Jetpack Compose is truly a game changer. I think the only limit of Jetpack Compose is our imagination. #android #jetpackCompose #androidDev #kotlin #animation
1
5
37
3,262
22 Jun 2025
What is the best way to be notified when something big occurs, without being bombarded with news notifications on slow days? I just want occasional phone notifications when something big happens like the US bombing Iran. I don't want notifications for ordinary day-to-day stuff.
6
1
33
3,448
Did you know: There is a phenomenon called "ai smuggling" where many people at big companies are secretly using AI to perform their job, and hiding that fact from their company.
7
1
26
3,520
Reads like an Onion headline, but it's actually real. The tables have turned, humans are now stealing AI jobs 🤣 An AI startup (builder.ai) backed by Microsoft turns out to be 700 Indian employees pretending to be chatbots.

5
2
41
3,097
Until AGI, AI will increase (not reduce) employment opportunities. Augmenting human with AI enables human to do more; previously impossible problems become tractable. People have unlimited problems they want solved; no limit on jobs available for those willing to solve problems.
7
6
44
4,842
20 May 2025
We've released Jules (an AI coding agent, similar to OpenAI Codex, based on Gemini), available for free! Check it out!
9
5
97
6,684
24 Apr 2025
I'm pretty sure ChatGPT just randomly selects some large percentage of image edit requests and claims they're a policy violation, just to gaslight the users. They want to publicly demonstrate how seriously they take safety.
3
1
19
3,013
19 Dec 2024
Efficiency is a hallmark of intelligence. Deep learning achieves SOTA on many benchmarks, but measuring accuracy irrespective of compute leads to bigger models instead of smarter models. Optimizing the wrong metric means running in a misguided direction. x.com/JimSproch/status/18697…

19 Dec 2024
Replying to @fchollet
We should be optimizing a metric like "intelligence per floating-point operation" (intelligence per unit of compute) or "intelligence per watt" (intelligence per unit of energy). This would lead to better architectures instead of bigger models.
2
13
4,583
24 Mar 2025
Glad to see this finally happening! I've been advocating for this for a while now! x.com/fchollet/status/190426…

A crucial point that everyone should be internalizing: in the age of test-time search, it's pretty much always possible to reach any level of capability by simply expending more compute. So it’s not just about, "can you do it?". The key is how efficiently you can do it. Intelligence is not just the ability to demonstrate high skill. It's the efficiency with which you acquire and deploy these skills. Both data efficiency and compute efficiency. Going forward we will report ARC-AGI scores on a 2D plot – score on one axis and cost per task on the other. In general, this is the only sane way to report performance for test-time reasoning models.
5
1,438
The broader public (both left and right) is shockingly inept at analyzing geopoliticals. If you want a non-partisan overview of strategy diplomacy geopolitics, I've found William Spaniel's videos to be truly excellent: youtube.com/@Gametheory101/v…
1
8
3,034
Only after you've watched some of his videos on the main channel linked above, if you're still hungry for more, he has an extras channel here: youtube.com/@LinesonMapsExtr…
2
1,999
Super cool tool! Great job @csinco! Love seeing these open source tools built using Compose for Desktop!
27 Feb 2025
Inspired by this article, I built a small tool in Compose Desktop to more easily iterate through building fancy mesh gradients from @sinasamaki's Modifier recipe! Code is here github.com/c5inco/Mesh with the code export assuming you are using the same Modifier recipe
2
42
2,645
Jim Sproch retweeted
19 Dec 2024
Replying to @jordwalke
A challenge that people will need to grapple with is how a newborn bootstraps their position in such a world. Human labor-hours is an asset granted in roughly equal proportions to all humans at birth. What happens when that grant has zero value? Worse, human life consumes energy.
3
1
8
1,787