vunderba

38 ms·

931.
by vunderba 9mo ago
Masters of Doom for about the 100th time. It's such an engaging read about the rise of id Software.
932.
by vunderba 9mo ago
It's not a perfect analogy but I often compare the advent of LLM development to the Battle of Crécy [1], where the introduction of the longbow meant that a common peasant with minimal training could potentially bring down a knight who
933.
by vunderba 9mo ago
Nice job. I think the OG of Win98 JS recreation from about a decade ago (which even has a working version of Space Cadet in it) is still around as well! https://github.com/1j01/98 https://98.js.org One int
934.
by vunderba 9mo ago
Well, one of the most fundamental algorithms for building a chess AI is minimax [1] (or variants like negamax), and that’s been around for close to a century. The key difference is that as compute power and available RAM have grown, it’s be
935.
by vunderba 9mo ago
Well I can’t speak to ASCII in particular, but I create a lot of mermaid UML diagrams specifically because unlike an image, they are: - Text searchable - Easy to adjust - Supported by a surprising number of markdown viewers.
936.
by vunderba 9mo ago
Agreed - the landing page specifically claims “Over 700 expert-crafted courses for any language pair imaginable.” I’m a trust-but-verify kind of person, and I can’t find a single mention of any language-learning facilities, academies, lin
937.
by vunderba 9mo ago
This is why penpal language exchange sites were so valuable like Lang-8 back in the day. You'd get corrections from actual native speakers and could dump all those corrected sentences into a database to link standalone words to actual
938.
by vunderba 9mo ago
Yeah understandable. If you've got the hardware for it, consider looking into Flux.2 Dev with the Turbo LoRA. It's probably the best-in-class for sheer prompt adherence right now, though Qwen-Image is still very respectable. And o
939.
by vunderba 9mo ago
Nice job. Unfortunately I'd say (just like with AI generated crosswords) we're not quite at the state where we can automate this without any manual supervision. Case in point: The image clue from "Alexandrian who found area
940.
by vunderba 9mo ago
Nice job. FWIW, I honestly think this would have been more useful back in the days of SD 1.5 and SDXL, when prompt witchery was far more important. Nowadays, SOTA models like NB Pro and gpt-image-1 are really good at adhering to even modera
941.
by vunderba 9mo ago
Obama seemed to manage it (around 3 million people) without high-profile incidents of American citizens being killed in the streets by ICE like Renée Good and Alex Pretti. Admittedly while a non-significant number of those deportees were tu
942.
by vunderba 9mo ago
I was also initially confused by the scale A3 to A4 until I realized that the musical interval trainer starts you on "A Minor Pentatonic" . This might be a bit confusing for non-musicians so I'd suggest starting with a more
943.
by vunderba 9mo ago
I never said it was.
944.
by vunderba 9mo ago
That's definitely true, and the medium also really makes a big difference as well (photorealism, digital painting, watercolor, etc.). Though in some cases, it is a bit easier to fix visual artifacts (using second-pass refiners, Img2Img
945.
by vunderba 9mo ago
Nice Try Open AI (Hey it rhymes!)
946.
by vunderba 9mo ago
Agreed! A lot of people were also using ZiT as a refiner downstream to help with some of the more problematic visual aspects of the original Qwen-Image. I'm really looking forward to running the unified model through its paces.
947.
by vunderba 9mo ago
Which is pretty amusing - because it's the exact opposite problem that BFL had with the original Flux model - every single image looked like it was taken with a 200mm f/4.
948.
by vunderba 9mo ago
Exactly - they did the same thing with the original version of Qwen-Image. It was API only for a while before being made available for local hosting.
949.
by vunderba 9mo ago
Couple of thoughts: 1. I’d wager that given their previous release history, this will be open‑weight within 3-4 weeks. 2. It looks like they’re following suit with other models like Z-Image Turbo (6B parameters) and Flux.2 Klein (9B paramet
950.
by vunderba 9mo ago
That's not what they did with Qwen-Image v1 - they announced it and it was available via API, but then they released the weights a few weeks after with an Apache 2.0 license. Let's at least give them the benefit of the doubt here.
951.
by vunderba 9mo ago
I encourage everyone to at least try ComfyUI. It's come a long way in terms of user-friendliness particularly with all of the built-in Templates you can use.
952.
by vunderba 9mo ago
A lot of people started realizing that it didn’t really matter how pretty the resulting image was if it completely failed to adhere to the prompt. Even something like Flux.1 Dev which can be run entirely locally and was released back in Aug
953.
by vunderba 9mo ago
This. I found ACE-Studio 1.5's cover ability to be incredibly lackluster. What actual musicians want is the ability to send in a unfinished song and then say something like "Add a percussive layer". Or let me take upload two
954.
by vunderba 9mo ago
Same for the Supreme Court. 20 years. A lifetime appointment is no different than a king.
955.
by vunderba 9mo ago
We actually made use of Studio Ghibli films with subtitles in our house to help improve younger sibling's reading skills. Kids still get to watch cartoons but get some reading practice in for free.
956.
by vunderba 9mo ago
Regarding turning hand-drawn diagrams into professional ones using GenAI, this was actually one of my earliest use cases with ChatGPT about a year ago. You can take a very rough hand sketch of a workflow or graph and ask it to generate Merm
957.
by vunderba 9mo ago
I know you've gone and stated your purported age (presumably to increase engagement which I'm not going to fault you for), but that being said, I'm still going to treat you like an adult. Vibe coding is not a substitute for b
958.
by vunderba 9mo ago
Nice job! Here's some feedback: - The exercises in focus mode feel too short, maybe they could get longer over time? - Disable text selection within the lesson container (using `user-select: none`). Not strictly necessary, but right no
959.
by vunderba 9mo ago
Shameless plug - but I just built a word game based on an old children's game my family used to play on long car rides. The game generates three words at random: light, blade, kill The player has to type in a word that matches all th
960.
Baldur's Gate to be turned into TV series – without the game's developers(bbc.com)
4 points by vunderba 9mo ago|0 comments

More ›