Hacker News: gnyeki

Unsolved Problems in MLOps

gnyeki — Wed, 15 Jul 2026 16:15:01 +0000

Article URL: https://spawn-queue.acm.org/doi/pdf/10.1145/3762989

Comments URL: https://news.ycombinator.com/item?id=48923109

Points: 40

# Comments: 4

My Images Are Dithered

gnyeki — Tue, 14 Jul 2026 17:52:58 +0000

Article URL: https://dead.garden/blog/how-my-images-are-dithered.html

Comments URL: https://news.ycombinator.com/item?id=48910576

Points: 2

# Comments: 0

One too many words on AT&T's $2k Korn shell and other Usenet topics

gnyeki — Mon, 08 Dec 2025 10:43:35 +0000

Article URL: https://blog.gabornyeki.com/2025-12-usenet/

Comments URL: https://news.ycombinator.com/item?id=46190796

Points: 31

# Comments: 3

Top-Down Microarchitecture Analysis Method

gnyeki — Wed, 15 Oct 2025 05:14:32 +0000

Article URL: https://www.intel.com/content/www/us/en/docs/vtune-profiler/cookbook/2025-4/top-down-microarchitecture-analysis-method.html

Comments URL: https://news.ycombinator.com/item?id=45588301

Points: 2

# Comments: 0

Storing configuration directly in executable, with no external config (2023)

gnyeki — Mon, 13 Oct 2025 14:53:59 +0000

Article URL: https://retrocomputing.stackexchange.com/questions/26803/storing-configuration-directly-in-the-executable-with-no-external-config-files

Comments URL: https://news.ycombinator.com/item?id=45568980

Points: 1

# Comments: 0

New comment by gnyeki in "Website is served from nine Neovim buffers on my old ThinkPad"

gnyeki — Tue, 19 Aug 2025 06:22:28 +0000

Yes, neither Nginx nor Neovim should be hitting the disk when serving the requests.

One difference I can imagine is context switching due to system calls. If Nginx incurs a context switch when it calls into the kernel for the disk cache, then it suffers a performance penalty. Neovim avoids this because the file contents are loaded into a table. When requests are served, getting the content is done fully in userspace.

I have no idea if this actually accounts for the performance difference though.

Hiding a message in my PyTorch weights

gnyeki — Mon, 25 Nov 2024 15:01:10 +0000

Article URL: https://blog.gabornyeki.com/2024-11-hiding-a-message-in-my-pytorch-weights/

Comments URL: https://news.ycombinator.com/item?id=42236820

Points: 1

# Comments: 0

Borrow Checking, RC, GC, and the Eleven (!) Other Memory Safety Approaches

gnyeki — Mon, 28 Oct 2024 18:07:59 +0000

Article URL: https://verdagon.dev/grimoire/grimoire#the-list

Comments URL: https://news.ycombinator.com/item?id=41974185

Points: 3

# Comments: 1

Linus and Dirk on succession, Rust, and more

gnyeki — Mon, 28 Oct 2024 17:33:40 +0000

Article URL: https://lwn.net/Articles/990534/

Comments URL: https://news.ycombinator.com/item?id=41973782

Points: 5

# Comments: 0

Inside STL: The string (2023)

gnyeki — Thu, 17 Oct 2024 19:14:29 +0000

Article URL: https://devblogs.microsoft.com/oldnewthing/20230803-00/?p=108532

Comments URL: https://news.ycombinator.com/item?id=41872819

Points: 1

# Comments: 0

A Modern Compiler for the French Tax Code (2021)

gnyeki — Mon, 22 Jul 2024 18:14:01 +0000

Article URL: https://arxiv.org/abs/2011.07966

Comments URL: https://news.ycombinator.com/item?id=41037654

Points: 2

# Comments: 0

New comment by gnyeki in "My Python code is a neural network"

gnyeki — Mon, 01 Jul 2024 16:21:37 +0000

I'm glad you found it valuable! Both are good questions and I haven't gone far enough mapping the code to Elman's architecture to know the answer to the second.

For your first question, using three hidden layers makes it a little clearer what the network does. Each layer performs one step of the calculation. The first layer collects what is known from the current token and what we knew after the calculation for the previous token. The second layer decides whether the current token looks like program code, by checking if it satisfies the decision rule. The third layer compares the decision with what we decided for previous tokens.

I think that this could be compressed into a single hidden layer, too. A ReLU should be good enough at capturing non-linearities so this should work.

New comment by gnyeki in "My Python code is a neural network"

gnyeki — Mon, 01 Jul 2024 16:03:51 +0000

This area is covered by non-parametric statistics more generally. There are many other methods to non-parametrically estimate functions (that satisfy some regularity conditions). Tree-based methods are one family of such methods, and the consensus still seems to be that they perform better than neural networks on tabular data. For example:

https://arxiv.org/abs/2106.03253

My Python code is a neural network

gnyeki — Mon, 01 Jul 2024 12:47:38 +0000

Article URL: https://blog.gabornyeki.com/2024-07-my-python-code-is-a-neural-network/

Comments URL: https://news.ycombinator.com/item?id=40845304

Points: 331

# Comments: 67