Thanks a lot. It helped me get a much more detailed view of turboquant than a few youtube videos that I watched. Also, the choice of color is excellent as it serves both light and dark mode. I'll try to use it in my sites. Kudos!
Thanks a lot <3
TurboQuant: A first-principles walkthrough
31–40 of 78 posts
Re: TurboQuant: A first-principles walkthrough
#32I feel like I've gotten really good at noticing which model generates what type of site and this oozes codex
Hey, thanks for the pointer. Had I known this, I would have used codex (as a matter of fact, I have never used it before and this prompts me to use it if I can get something like this much quicker with codex). I think making codex copy this for a new content will be much easier now. The issue was with making things the way I exactly want, the exact intuition, the exact primers, and the exact visuals to drive the point home.
Re: TurboQuant: A first-principles walkthrough
#33Re: TurboQuant: A first-principles walkthrough
#34TurboQuant is a restricted version of EDEN quantization (NeurIPS 21, ICML 22). It lacks the optimal scale derivations, which makes the TurboQuant variant considerably less accurate than those works. We show this thoroughly in a new note at https://arxiv.org/abs/2604.18555 . We were the first to introduce post-rotation distribution-aware quantization in 2021. This was later implemented in many fields, including federa…
Thanks a lot for pointing this out. I will update this explainer to properly add the prior literature so that there is a proper attribution.
Re: TurboQuant: A first-principles walkthrough
#35what did the author used to create the site?
I did a bunch of things :D I am not a frontend engineer (I am MLE) so I don't have the prowess to create things like this. I am heavily inspired by 3blue1brown and I love creating interactive explainers for ML concepts like this. I previously created this as well arkaung.github.io/interactive-eigenvector/. I heavily used Claude to get to the the exact design, typography, and style I want (there was a lot of hand holding to get to this state). I heavily influenced Claude on how I want the explainer to flow, how I want to make things intuitive, the kinds of mathematical concepts I want to visualize (and how). So all in all, a lot of hand holding for the Coding agents to get to where I want and exactly how I want.
But at the end of the day it is just vanilla HTML, CSS and JS without anything fancy :D MathJax 3 was used to render math stuff.
Re: TurboQuant: A first-principles walkthrough
#36what did the author used to create the site?
The fonts, the cards, the copy are all hallmarks of Claude Code.
While the aesthetic doesn't spark joy for me, the overall execution is great, the presentation flow and interactive boxes are very nice.
Re: TurboQuant: A first-principles walkthrough
#37I feel like I've gotten really good at noticing which model generates what type of site and this oozes codex
Hey, thanks for the pointer. Had I known this, I would have used codex (as a matter of fact, I have never used it before and this prompts me to use it if I can get something like this much quicker with codex). I think making codex copy this for a new content will be much easier now. The issue was with making things the way I exactly want, the exact intuition, the exact primers, and the exact visuals to drive the poin…
What did you use?
Re: TurboQuant: A first-principles walkthrough
#38TurboQuant is a restricted version of EDEN quantization (NeurIPS 21, ICML 22). It lacks the optimal scale derivations, which makes the TurboQuant variant considerably less accurate than those works. We show this thoroughly in a new note at https://arxiv.org/abs/2604.18555 . We were the first to introduce post-rotation distribution-aware quantization in 2021. This was later implemented in many fields, including federa…
Thanks a lot for pointing this out. I will update this explainer to properly add the prior literature so that there is a proper attribution.
Thanks for the quick response and for being willing to update the explainer. I really appreciate the clarification.
Re: TurboQuant: A first-principles walkthrough
#39I feel like I've gotten really good at noticing which model generates what type of site and this oozes codex
Hey, thanks for the pointer. Had I known this, I would have used codex (as a matter of fact, I have never used it before and this prompts me to use it if I can get something like this much quicker with codex). I think making codex copy this for a new content will be much easier now. The issue was with making things the way I exactly want, the exact intuition, the exact primers, and the exact visuals to drive the poin…
Woah very cool, yeah I think the cards and heading/subheading structure is very similar to what codex outputs, but I can tell the different visualizations definitely require your own personal touch
Re: TurboQuant: A first-principles walkthrough
#40I am fascinated by this and similar research (RotorQuant, etc). It seem by next year we will be able to run this year's largest models on last year's hardware. :) Maybe we won't need as many data centers and as much power as we thought. Maybe we can run more powerful models locally.
Just look at deepseek V4, this preview model uses only 8 GB for 1M token KV cache(the context). It's insanely efficient already. It's just that most models that are coming out are barely catching up with technical breakthroughs.
Deepseek are pioneers.
Unfortunately V4 is not trained for most real world usage, it is mainly for world general knowledge.