Dr. André Schnabel

every software engineer is just someone who wanted to make video games and quietly gave up on that dream

@plainionist Learn the basics. Just as before. Learn them very well. Write assembler, C, Java, Ruby. Learn algorithms and data structures. Read the old classics. And then start using agents. Novices with power tools tend to lose fingers.

NVIDIA cuOpt is officially the #1 OSS solver on the Hans Mittelmann MIPfeas leaderboard. This is a massive win for GPU-accelerated Mixed Integer Programming, proving cuOpt is ready to power the next generation of complex, memory-intensive workloads. From complex fleet routing to supply chain planning, the possibilities for low-latency agentic systems just expanded. Hans Mittelmann Benchmark: plato.asu.edu/ftp/mipfeas.ht…

The GAMS solver link for NVIDIA‘s cuOpt GPU-based solver got updated and now also allows using the mip tracing facility to get detailed solution progress logs. gams.com/blog/2025/09/g…

LLM adoption among US workers is closing in on 50%. Meanwhile labor productivity growth is lower than in 2020. Many counter-arguments can be made here, e.g. "they don't know yet how to be productive with it, they've only been using for 1-2 years", "50% is still too low to see impact", "models next year will be unbelievably better", etc. But I think we now have enough evidence to say that the 2023 talking point that "LLMs will make workers 10x more productive" (some folks even quoted 100x) is probably not accurate.

Sorry you don't like our codec implementations in VLC. Looking forward to seeing yours.

It was a fun exercise recently to just open up a completely blank file and write an RL agent from scratch, without looking at any of my prior code. There is a point of scale where rewriting things from scratch is a bad idea, but it is a blessing when you can! By “from scratch” I mean with pytorch, and I chuckle a bit about how I used to be irritated when people would say that, as I considered it not from scratch if it wasn’t in C with no libraries, but now I am them.

@cmuratori @pikuma A 2021 review by @eatonphil finds handwritten parsers dominant in the major compilers, and the trend appeared to be a decline in parser-generator use. notes.eatonphil.com/parser-generat…

Merge request makes more sense than pull request.

not many people know this, but you can also glide down to a value

Another reason to be bullish about Arch: @valvesoftware didn't just base SteamOS and the Steam Deck on it, they're actively investing in the infrastructure of the project too. We have a lot to thank Gabe Newell for in the Linux world! @lists.archlinux.org/thread/RIZSKIBDSLY4S5J2E2STNP5DH4XZGJMR/" target="_blank" rel="nofollow noopener">lists.archlinux.org/archives/list/…

There is no point doing competitive cycling in an attractive countryside: riders focus exclusively on the bike in front of them, the road if they are leading, or the bike computer if they are on Strava.

An existential analogy for you. This screenshot is of a PDP-11/70 running RT-11. What it doesn't know is that it's actually running in a simulation, and that simulation is hosted on a DEC VAX 4000. But what that VAX doesn't know is that it's also running in a simulation, and THAT simulation is hosted on NetBSD 10.1 via simH. But what NetBSD doesn't know is that it's also running in a QEMU simulator. And what THAT simulator doesn't know is that it's running within a virtual machine in Proxmox. Yes, I really set this all up, and it works... And Proxmox is running on the bare hardware, a Dell 7875 Workstation with 192 Threadripper cores and 1TB of memory. Much to my surprise, the performance is akin to that of my real PDP-11/34, at least for tasks like directory listings and text scrolling. Assuming you live in a simulation, this is what happens when your simulation develops the technology to create a simulation of its own. And so on all the way down. In fact, the odds are you're either in the Proxmox layer or somewhere deep in the simulation hierarchy. The odds that you exist in the first simulated layer seems low. After all, once you've demonstrated that simulating reality is possible at all, the depth of the simulations becomes a zero-one-infinity problem.

This is interesting — a compiler/runtime that makes almost all standard C/C++ code completely memory safe. The performance overhead is not negligible, but there are plenty of cases where “rewrite it in rust” still isn’t a practical solution.

The video of my talk at Upper Bound 2025 is up: youtu.be/rQ-An5bhkrs?si…

If someone is telling you he understands AI, he doesn't understand AI.

Claude is teaching me some nice Go slop.

A blockchain is a data structure. When I hear about "the blockchain community" I think about, say, the linked list community or the binary tree community.