Yusuf Shihata retweeted
The limiting factor of software engineering has never been code production. It is complexity management.
Great talk, The amount of great engineering put to make tigerbeetle work this well is very inspiring
Performance Engineering at TigerBeetle, @maxikuschewski 's guest lecture at @TU_Muenchen:
01:59 Understand your Workload
05:04 Performance
11:36 Tail Latency
21:52 Algorithm Engineering
28:02 Async I/O in TigerBeetle
35:21 Object Storage Clients
youtube.com/watch?v=-p-Xm7VE…
Yusuf Shihata retweeted
Announcing Fugu-Ultra v1.1 🐡
We’ve been thrilled by the reception to the Fugu model family. Thanks to everyone who tried it, shared feedback, and trusted Fugu with real work.
Today, we’re releasing Fugu-Ultra v1.1 → sakana.ai/fugu
Upgraded to incorporate the latest frontier models, resulting in stronger performance across every benchmark shown, including gains of up to 7.9 points over v1.0, with particularly strong results on ProgramBench and Terminal Bench 2.1.
Fugu-Ultra v1.1 is more capable across coding, agentic tasks, and advanced reasoning, and available at the same price as Fugu-Ultra v1.0
The frontier keeps moving, and Fugu keeps getting better.
a Zig deep learning framework using pure comptime static allocation guarantees zero runtime allocs, max performance, memory safety.
But for architectures like MoE with dynamic routing, full static allocation forces worst-case buffer sizes causing massive VRAM bloat. any ideas??
Yusuf Shihata retweeted
This is why you still need to read the code!
This was written by Fable 5
What a stinker
My honest take on the "do you read the code?" debate, I know it's a too late. but I think it's somehow good take. Feel free to correct me if you find some of my ideas are not correct:
vixel2006.github.io/blog/wri…
Lately, I returned to work on an old project of mine. A deep learning framework in C and CUDA. I wrote a blog post about the architecture that I will adopt for the rewrite, using codegen and beam search instead of handwritten CUDA kernels.
medium.com/@yusufshihata2006…
Sometimes while working on my current project glu (robotics communication framework) I just ask myself what was in ROS2 engineers minds when they didn't do that or did that. maybe I will find out when I finish the project and find it's a piece of shit. or maybe not.
Thank God there still some sane people in this industry
Glu is actually very close to launch. I have finished v0.2.0 with online communication layer. and created examples for it with zig. next step is to re-create the examples in ros2 and do the benchmarking.
Currently building a ROS2 replacement in Zig, it's a light-weight framework, not an OS, inspired by docker but for robotics. v0.1.0 is done using shared memory IPC. I'm thinking about adding C and python bindings for the framework.