Top all-time · #2656
Sander Dieleman
Personal blog · sander.ai · on HN as benanneaiscience
Sander Dieleman's personal blog at sander.ai covers ai and science. Since 2023, 28 different Hacker News readers have linked it 41 times, most often “Generative Modelling in Latent Space”. It is still writing: the last post was on 2026-08-23, and you can follow it by feed.
Describe this nonsense.
Most-linked posts?
- Generative Modelling in Latent Space9 people · 9 mentions · first linked 2025-04-15
- Diffusion Is Spectral Autoregression5 people · 5 mentions · first linked 2024-09-02
- Perspectives on Diffusion4 people · 5 mentions · first linked 2023-07-20
- The paradox of diffusion distillation (2024)3 people · 3 mentions · first linked 2024-02-29
- Diffusion Language Models3 people · 3 mentions · first linked 2023-04-05
- Guidance: A cheat code for diffusion models3 people · 3 mentions · first linked 2023-05-08
- Musings on Typicality2 people · 2 mentions · first linked 2023-03-05
- Continuous Diffusion Language Models (CDLM's)2 people · 2 mentions · first linked 2026-08-24
What Hacker News says?
Here's a blog post I liked that explains a connection: https://sander.ai/2024/09/02/spectral-autoregression.html They call diffusion a form of "spectral autoregression", because it tends to first predict lower frequency features, and later predict higher frequency features.
Comment linking /2024/09/02/spectral-autoregression.html
on “BERT is just a single text diffusion step”Submitted Diffusion Is Spectral Autoregression as a story · 223 points
HN discussionThere's a great blog post from Sander Dieleman about exactly this - why do we need a two step pipeline, in particular for images and audio? https://sander.ai/2025/04/15/latents.html For text, there are a few papers that train the tokenization and language model end-to-end, see: https://arxiv.org/abs/2305.07185
Comment linking /2025/04/15/latents.html
on “Neural audio codecs: how to get audio into LLMs”Submitted Continuous Diffusion Language Models (CDLM's) as a story · 135 points
HN discussionI actually wrote down some thoughts about audio phase in a previous blog post: https://sander.ai/2020/03/24/audio-generation.html#motivatio... I have an example audio clip in there where the phase information has been replaced with random noise, so you can perceive the effect. It certainly does matter perceptually, but it is tricky to model, and small "vocoder" models do a decent job of filling it in post-hoc.
Comment linking /2020/03/24/audio-generation.html
on “Diffusion Is Spectral Autoregression”Submitted Generating music in the waveform domain (2020) as a story · 92 points
HN discussionOof, you're not going to like this other blog post I wrote then :D https://sander.ai/2023/07/20/perspectives.html
Comment linking /2023/07/20/perspectives.html
on “Diffusion Is Spectral Autoregression”Submitted The paradox of diffusion distillation (2024) as a story · 18 points
HN discussionI’m trying to understand what the ‘spectral’ interpretation of IMM is: but perhaps I shouldn’t https://sander.ai/2024/09/02/spectral-autoregression.html
Comment linking /2024/09/02/spectral-autoregression.html
on “Beyond Diffusion: Inductive Moment Matching”Submitted Generative Modelling in Latent Space as a story · 15 points
HN discussionThere are so many similar diffusion model distillation techniques these days that it's become hard to tell the difference between them, this is probably the fourth example I've seen that uses adversarial loss to distill the model, the others being UFOgen, some other work by StabilityAI on SDXL-turbo and similar, SDXL-Lightning by ByteDance. I found this blog page that explains some of the differences: https://sander…
Comment linking /2024/02/28/paradox.html
on “MIT Unveils Gen AI Tool That Generates High Res Images 30 Times Faster”Submitted Guidance: A cheat code for diffusion models as a story · 8 points
HN discussionhttps://sander.ai/2024/09/02/spectral-autoregression.html
Comment linking /2024/09/02/spectral-autoregression.html
on “Faster convergence for diffusion models”Submitted Continuous Diffusion Language Models as a story · 7 points
HN discussionI've since moved on to work primarily on diffusion models, so I have a series of blog posts about that topic as well! - https://sander.ai/2022/01/31/diffusion.html is about the link between diffusion models and denoising autoencoders, IMO the easiest to understand out of all interpretations; - https://sander.ai/2023/07/20/perspectives.html covers a slew of different perspectives on diffusion models (including the "a…
Comment linking /2023/07/20/perspectives.html
on “Generating music in the waveform domain (2020)”Submitted Generative Modelling in Latent Space as a story · 3 points
HN discussionHey! I’m the post author. Highly recommend Sander Dieleman’s blog for alternative interpretations https://sander.ai/2023/07/20/perspectives.html I personally find the SDEs the most intuitive, and the deterministic ODE / consistency models / rectified flow stuff as ideas that are easier to understand after the SDEs. But not everyone agrees!
Comment linking /2023/07/20/perspectives.html
on “Diffusion Without Tears”Submitted Generative Modelling in Latent Space as a story · 3 points
HN discussion- just a headsup - your links to the slides for deeplearning you did here https://sander.ai/2014/05/29/slides-meetup.html are broken
Comment linking /2014/05/29/slides-meetup.html
on “Learning the Integral of a Diffusion Model”Submitted Let's Talk about Latents as a story · 3 points
HN discussionI briefly covered that connection in an earlier blog post: https://sander.ai/2023/07/20/perspectives.html#flow ... but it's definitely something that might deserve a longer-form treatment at some point :)
Comment linking /2023/07/20/perspectives.html
on “Learning the Integral of a Diffusion Model”Submitted Guidance: A cheat code for diffusion models as a story · 3 points
HN discussionhttps://sander.ai/2022/01/31/diffusion.html The differences are thoroughly discussed here. From my point of view, while there is this surface-level similarity, the training process for diffusion models is directly a result of the mathematical framework on which they are defined which is distinct from denoising autoencoders. I'd suggest reading the above blog if you are interested, it is thorough!
Comment linking /2022/01/31/diffusion.html
on “Physics Advanced Generative AI”Submitted Diffusion Language Models as a story · 3 points
HN discussionSubmitted Generative Modelling in Latent Space as a story · 2 points
HN discussionSubmitted Generative Modelling in Latent Space as a story · 2 points
HN discussionSubmitted Generative Modelling in Latent Space as a story · 2 points
HN discussionSubmitted Diffusion Is Spectral Autoregression as a story · 2 points
HN discussionSubmitted Guidance: A cheat code for diffusion models as a story · 2 points
HN discussionSubmitted Musings on Typicality as a story · 2 points
HN discussion
