@Vipitis profile picture
Vipitis @Vipitis

thermal photography enthusiast, CS enjoyer, language model researcher

Joined February 2012
  • Tweets 3,760
  • Following 2,366
  • Followers 71
588 Photos and videos
Vipitis@Vipitis
9 Jun 2025
Has Shadercoding changed over the decade? Maybe not so much based on some fun queries today:
1
63
Vipitis@Vipitis
10 Nov 2024
Adding 3:39:54 to my resume
ThePrimeagen
@ThePrimeagen
6 Nov 2024
ITS HAPPENING, ITS YAPPENING x.com/i/broadcasts/1vAGRONDp…
1
244
Vipitis@Vipitis
20 Aug 2024
Is there a Dell A770, and what is codenamed window maker?
110
Vipitis@Vipitis
18 Aug 2024
Does finetuning on arXiv dramatically increase benchmark contamination? Surely all these ablation studies and examples in the appendix are a source for leakage
82
Vipitis@Vipitis
16 Jul 2024
New institution LLM benchmark: blind debugging. Model can request certain inputs and gets outputs. Based on that information alone: can the model 1. predict the algorithm and 2. predict a Bugfix for said algorithm?
90
Vipitis@Vipitis
2 Jul 2024
"see less often" should instead be "show me never again"
1
68
Vipitis@Vipitis
17 Jun 2024
Battlemage before December 👀
80
Vipitis@Vipitis
16 Jun 2024
Shadertoy buffers are RGBA and 32bit floats. so they could easily hold the position and velocity of particles in a 2D simulaiton. experiments with 1 particle: shadertoy.com/view/lXK3WV
75
Vipitis@Vipitis
15 Jun 2024
print(f"{s=}") > printf(s)
1
61
Vipitis@Vipitis
8 Jun 2024
Stochastic binary search where you sample against a normal distribution?
54
Vipitis@Vipitis
27 Apr 2024
The worst solution is always the one you aren't using.
2
66
Vipitis@Vipitis
22 Apr 2024
Did you know that Intel ExtraSS implementation is open source? This seemingly contains the scripts to generate the dataset, model architecture and training code... 👀 github.com/ltkong218/IFRNet/…
81
Vipitis@Vipitis
21 Apr 2024
It works on my other machine.
57
Vipitis@Vipitis
9 Apr 2024
In times where everything is an "LLM" where is the love for lLM or the little language model?
1
67
Vipitis@Vipitis
8 Apr 2024
When is Melville Sound?
63
Vipitis@Vipitis
5 Mar 2024
Any alarmist blog post or sentient machines fiction will end up in training data and cause language models to act in this way. If you want models to be more friendly - write about "helpful AI" than a system prompt telling a model 'you are a helpful AI assistant' makes it better.
1
85
Vipitis@Vipitis
2 Mar 2024
Intel seems to be dropping some Gaudi (3?) News in the coming weeks. Just in time for GDC 🤔
4
123
Vipitis@Vipitis
27 Feb 2024
where does the dependency spiral begin?
64
Vipitis@Vipitis
22 Feb 2024
We need an evaluation benchmark against safety or overdone diversity. Targeting API products specifically, were you pay for it and don't know what preprocessing is done on your inputs.
1
63
Vipitis@Vipitis
16 Feb 2024
Will replacing the multi billion dollar movie industry help Sam get the $7B
1
47
Load more