Joined October 2022
163 Photos and videos
Pinned Tweet
most important lesson from years of distributed systems: keep everything on a single machine for as long as humanly possible
49
221
4,711
138,342
Piotr Sarna retweeted

7
29
117
33,725
Thank God we managed to generate those 5000000 cheap rip-offs of Minecraft before Fable got banned, Europe is safe now
2
159
Forget restricting "frontier AI research," @AnthropicAI please instead consider blocking requests to reimplement Minecraft with Fable, it's enough
5
426
"there can't be that many syscalls that involve filesystem paths" - me, a few months ago (io_uring* not even supported yet)
10
409
sandboxing internals fun fact: when intercepting syscalls with `ptrace` on x86-64, you can swap one syscall for another just by replacing the syscall number stored in rax - easy. That won't work for arm64, which does not check rax again and just continues to run the original syscall. Solution -- use a pseudoregister called NT_ARM_SYSTEM_CALL to make sure you can overwrite the syscall user wanted to call with your sandboxed one.
1
1
15
1,464
Piotr Sarna retweeted
"Who would want to read what I write?" This insecurity is one of the biggest reasons developers don't start blogging. Tomorrow, @c_a_dunlop and @sarna_dev join us live to discuss common writing roadblocks, finding blog topics, writing as a non-native English speaker, and why you don't need to be an expert to share what you've learned. Register: hubs.la/Q04kktHX0
2
12
2,264
My weekend plan of rooting an old android photo frame turned out somewhat anticlimactic... 1. turn on adb in frame settings 2. adb shell logs in with passwordless `su` 3. sshd installed, the end Where's the challenge in that
1
185
Piotr Sarna retweeted
Laguna M.1 is #1 on @kilocode today. Good to see Laguna getting put to work.
1
5
42
3,900
May in writethat.blog 👇 Flipper One — we need your help // @zhovner mimalloc: A new, high-performance, scalable memory allocator for the modern era // Daan Leijen, @Microsoft When “idle” isn’t idle: how a Linux kernel optimization became a QUIC bug // @estcarisimo & Antonio Vicente, @Cloudflare Lessons learned building high-performance Rust profiler // @_pawurb Natural Language Autoencoders: Turning Claude’s thoughts into text // @AnthropicAI Content for content’s sake // @mitsuhiko
1
7
239
Piotr Sarna retweeted
A lot of developers don't struggle with ideas. They struggle with feeling "qualified" to share them. That's one of the themes @c_a_dunlop and @sarna_dev explore in Writing for Developers. They look at overcoming impostor syndrome to finding practical topics hidden in your everyday work. They'll be discussing it live on June 9th at 3pm ET. Register here: hubs.la/Q04jNWlB0
2
5
1,454
Back in my university days, the only use case for AI was gaming bots and protein folding, and here we are again, except models got way smarter. Beautiful 🤌
The @poolsideai Hackathon is over but I'm continuing to teach Laguna XS.2 to be a better scientist 🧑‍🔬. Our Protein-Ligand Design Gym now has 1000 examples with harder protein/ligand interaction problems.
2
232
Piotr Sarna retweeted
poolside running on satellites! 🛰️🚀
(1/2) Your most sensitive code deserves AI that never leaves your network.
1
3
32
4,851
Piotr Sarna retweeted
We had to blow some stuff up today. 500 holes. 20,000 lbs of explosives. 🤯
11
5
114
33,209
I have notoriously bad intuition for KL divergence values, so I made myself a game to improve. Goal: draw a distribution that hits as close as possible to target KL divergence. KL Zero: klzero.sarna.dev
1
8
1,008
Being mentioned in an AI technical report was definitely not on my bingo card in 2024... Long journey from the open source dist sys bubble
Today we’re publishing the technical report behind Laguna M.1 and Laguna XS.2. This report opens up more of what went into them: Model Factory, pre-training data, distributed training, post-training, agent RL, quantization, and evaluation. poolside.ai/assets/laguna/la…
1
1
17
973
Piotr Sarna retweeted
Laguna M.1 and XS.2 now support 256K context. Laguna M.1 is now live with a 256K context window on the Poolside API and OpenRouter. With this update, it reaches 45.8% on Terminal-Bench 2.0, improving long-horizon performance. Laguna XS.2 is also moving to 256K today, with the updated config already available on Hugging Face. Both models remain free to use. Over 1T tokens have been processed since launch 4 weeks ago. Excited to see what people build with the longer context window.
4
17
78
6,999
Fabulous Adventures in Data Structures and Algorithms by @ericlippert gets very practical, fast. Probabilistic algorithms are experiencing a renaissance in the AI era, so it's a good idea to catch up early
2
6
742
courtesy of @ManningBooks
1
108
My temporary email server just got agentic. Use sorry.idont.date/SKILL.md to autonomously sign up for spammy newsletters 🤌 Bots spam bots, and you get those promo codes you wanted, win-win
1
1
135
Prompt: hey laguna, I just signed up to a xx-redacted-xx newsletter with xx_redacted_xx@idont.date. Be so kind and fetch the verification email from that inbox, verify yourself, and then wait for the email that contains the promo codes, and list them for me. Much obliged
109