But then I have to upgrade to iOS 26… The cure is worse than the disease!
For now I am using wireguard+pihole on a cheap VPS for all my devices. It’s not perfect (data center IP so some places block it) but it’s good enough for now. When I’m forced to update to 26 will definitely look at Wipr since I tested that out and it was really good other than the in-app issue.
I've been thinking about this too—how different DDN is from other generative models. The idea of generating multiple outputs at once in a single pass sounds like it could really speed things up, especially for tasks where you need a bunch of samples quickly. I'm curious how this compares to something like GANs, which can also generate multiple samples but often struggle with mode collapse.
The zero-shot conditional generation part is wild. Most methods rely on gradients or fine-tuning, so I wonder what makes DDN tick there. Maybe the tree structure of the latent space helps navigate to specific conditions without needing retraining? Also, I'm intrigued by the 1D discrete representation—how does that even work in practice? Does it make the model more interpretable?
The Split-and-Prune optimizer sounds new—I'd love to see how it performs against Adam or SGD on similar tasks. And the fact that it's fully differentiable end-to-end is a big plus for training stability.
I also wonder about scalability—can this handle high-res images without blowing up computationally? The hierarchical approach seems promising, but I'm not sure how it holds up when moving from simple distributions to something complex like natural images.
Overall though, this feels like one of those papers that could really shift the direction of generative models. Excited to dig into the code and see what kind of results people get with it!
3. Answers or relevant explanations to any other questions can be found directly in the original paper (https://arxiv.org/abs/2401.00036), so I won’t restate them here.
Well, I've asked ChatGPT with your prompt and came out with a very long comment. Asking to shorten it, we get:
"Figma is basically Google Docs for design — a fast, browser-based tool where multiple people can edit the same UI file in real time. No installs, no emailing files around.
Its magic as a business is the frictionless onboarding (just share a link), viral team adoption, and a freemium model that naturally expands into enterprise contracts. Works cross‑platform, so it spreads fast.
That combo — great product + viral growth + strong enterprise lock‑in — is why it became the design platform and why Adobe was ready to pay billions for it."
In the longer form it was also enthusiastic about the WASM part, but that didn't make the cut.
On the contrary, Figma's value proposition is increased by LLMs. Current coding assistants are like savant-idiot junior devs: They have relatively low reasoning capabilities, way too much courage, lack taste and need to be micromanaged to be successful.
But they can be successful if you spell out the exact specifications. And what is Figma if not an exact specification of the design you want? Within a couple of years the Frontend Developer Market might crash pretty hard.
Looking forward to seeing an LLM that can produce good design. Figma is working on this themselves, they have the distribution, so even it came to pass why wouldn’t they own the market? They have the data and the resources to buy more.
That’s not what I mean. Designers can work with any tool to produce a design, and then Claude or Gemini or whatever are quite capable of turning that design into working React code or HTML or whatever.
you can create a lot of wealth for yourself by finding the bigger fool so to speak. And arguably, that's what a lot of tech IPOs are in any case, so why single out Figma for engaging in the practice?
I mean in the future there are probably no PMs, Designers or Engineers. All those roles are going to converge. There will be a bunch of people that build and manage the software that creates software.
The best part about requiring them to use the word “unsubscribe” is I can do this email rule: If an email says “unsubscribe” in it, move it to “says-unsubscribe” folder.
I look at that email once a week for the false positives. Huge QoL increase.
For now I am using wireguard+pihole on a cheap VPS for all my devices. It’s not perfect (data center IP so some places block it) but it’s good enough for now. When I’m forced to update to 26 will definitely look at Wipr since I tested that out and it was really good other than the in-app issue.
reply