Hacker News: joaquincabezas

New comment by joaquincabezas in "The Principles of Diffusion Models"

joaquincabezas — Mon, 10 Nov 2025 00:14:24 +0000

AI is definitely related to dashes!!

New comment by joaquincabezas in "The Principles of Diffusion Models"

joaquincabezas — Sun, 09 Nov 2025 21:43:20 +0000

why am I only getting 26 matches? where's the threshold then? :D

New comment by joaquincabezas in "Backpropagation is a leaky abstraction (2016)"

joaquincabezas — Mon, 03 Nov 2025 15:47:07 +0000

just found it! but it's private, I can send it to you if interested but not to publish it

New comment by joaquincabezas in "Backpropagation is a leaky abstraction (2016)"

joaquincabezas — Mon, 03 Nov 2025 08:56:58 +0000

Hi! It's not public, it's part of the https://www.urv.cat/en/studies/master/courses/computer-secur... and I've not found it online

New comment by joaquincabezas in "Backpropagation is a leaky abstraction (2016)"

joaquincabezas — Sun, 02 Nov 2025 11:49:47 +0000

off-topic, anybody knows what's going on with EurekaLabs? It's been a while since the announcement

New comment by joaquincabezas in "Backpropagation is a leaky abstraction (2016)"

joaquincabezas — Sun, 02 Nov 2025 08:34:06 +0000

same here, even more if I'm doing it over few days and different angles

New comment by joaquincabezas in "Backpropagation is a leaky abstraction (2016)"

joaquincabezas — Sun, 02 Nov 2025 08:03:58 +0000

I took a course in my Master's (URV.cat) where we had to do exactly this, implementing backpropagation (fwd and backward passes) from a paper explaining it, using just basic math operations in a language of our choice.

I told everyone this was the best single exercise of the whole year for me. It aligns with the kind of activity that I benefit immensely but won't do by myself, so this push was just perfect.

If you are teaching, please consider this kind of assignments.

P.S. Just checked now and it's still in the syllabus :)

New comment by joaquincabezas in "Modeling Others' Minds as Code"

joaquincabezas — Mon, 20 Oct 2025 14:42:11 +0000

when someone behaves in a very predictable way I use to say "I could code you in C!". Well, turns out is Python!

New comment by joaquincabezas in "Will Smith's concert crowds are real, but AI is blurring the lines"

joaquincabezas — Tue, 26 Aug 2025 14:34:09 +0000

No wonder why Google says:

“Generate, transform and edit images with simple text prompts, or combine multiple images to create something new. All in Gemini.“

https://news.ycombinator.com/item?id=45026719

New comment by joaquincabezas in "LLM Embeddings Explained: A Visual and Intuitive Guide"

joaquincabezas — Mon, 28 Jul 2025 13:41:37 +0000

The culprit seems to be: https://huggingface.co/spaces/hesamation/primer-llm-embeddin...

So someone, at some point, thought this was a feature

New comment by joaquincabezas in "Nvidia-Ingest: Multi-modal data extraction"

joaquincabezas — Fri, 10 Jan 2025 11:15:18 +0000

lol, while checking which OCR is using (PaddleOCR) I found a line with the text: "TODO(Devin)" and was pretty excited thinking they were already using Devin AI...

"Devin Robison" is the author of the package!! Funny, guess it will be similar with the name Alexa

New comment by joaquincabezas in "Timemap.org – Interactive Map of History"

joaquincabezas — Thu, 12 Dec 2024 15:55:12 +0000

this is really cool! I have a ruler from metermorphosen.de and some posters and cardboards from museums. I will share it :)

New comment by joaquincabezas in "The Rules of Programming (2023)"

joaquincabezas — Sun, 08 Dec 2024 21:21:27 +0000

The first rule of programming is you do not talk about programming

New comment by joaquincabezas in "How We Optimize LLM Inference for AI Coding Assistant"

joaquincabezas — Sun, 01 Dec 2024 18:48:11 +0000

balancing compute-bound (prefill) and memory-bound (decode) is a fine art. Luckily there are lots of improvements (incentives) if you can adjust it to your use case (this time is Coding assistants), but it is generally a lonely journey. Good to see you paired with Colfax International.

New comment by joaquincabezas in "SpawELO – small free matchmaking system for LAN parties"

joaquincabezas — Sat, 02 Nov 2024 20:50:02 +0000

what about trying with Shapley values?

New comment by joaquincabezas in "Rust and C++ with Steve Klabnik and Herb Sutter [audio]"

joaquincabezas — Thu, 31 Oct 2024 11:40:45 +0000

Exactly this. I've felt it while checking some C++ OSS projects, thinking "wow I am more Java than this, and I don't do Java". This prevented me directly jumping into C++ which would be the natural step coming from C (because in reality it's not the natural step)

New comment by joaquincabezas in "Rust and C++ with Steve Klabnik and Herb Sutter [audio]"

joaquincabezas — Thu, 31 Oct 2024 11:36:23 +0000

In some way I am stuck because I don't want to dismiss any of these (kind of Buridan's donkey) but planning this in a sequential way seems smart and would end the inner monologue

New comment by joaquincabezas in "Rust and C++ with Steve Klabnik and Herb Sutter [audio]"

joaquincabezas — Thu, 31 Oct 2024 10:55:27 +0000

I've been a long time considering that I should learn Rust. I am using Python as main language (for Machine Learning tasks) so I am at the edge:

Go with:

C++, because I used to code in C (for embedded, 8-bit microcontrollers) when I was younger and also for its possible advantage for learning some CUDA (this last thing is mostly for fun)

Rust, because I have some trauma for kernel panics with bad memory management back in the days with C, and because I want to build some tooling in a nice, efficient way

New comment by joaquincabezas in "An n-ball Between n-balls"

joaquincabezas — Wed, 09 Oct 2024 16:51:49 +0000

wow discovering Hamming’s lecture was enough for me! so good

New comment by joaquincabezas in "Liquid Foundation Models: Our First Series of Generative AI Models"

joaquincabezas — Mon, 30 Sep 2024 16:21:16 +0000

Love the humour here:

What are Language LFMs not good at today: […] Counting r's in the word "Strawberry"!