vunderba

18 ms·

361.
by vunderba 5mo ago
Nice job. When I was a kid, we tried to merge our two favorite things: Magic: The Gathering and chess, combined into one ridiculous hybrid game. It takes a game that’s normally one of perfect information and adds in the indeterminism of the
362.
by vunderba 5mo ago
You need to build out a baseline style using text-to-image and then always add those images as "style reference" images in future prompts e.g. "Create a four-panel comic of bla bla bla. Use the attached images strictly as st
363.
by vunderba 5mo ago
It's surprisingly decent for being kinda budget! I'm a big fan of 8bitDo because all their controllers are usu. easy to pair with all my stuff (Switch, PC, etc). I’ve actually got my arcade stick temporarily modded with a Sanwa ga
364.
by vunderba 5mo ago
Nice! I'm a HUGE DoDonPachi fan, Mushihimesama is also great fun as well. I have the Nintendo Switch port coupled with an 8BitDo arcade stick and have sunk tons of hours into it.
365.
by vunderba 5mo ago
In my experience, LLMs are pretty good at writing Go code. In fact, that’s one of my principal use cases: I have a number of bespoke Node programs from many years ago that usually run in the background on my computer. Each process (with the
366.
by vunderba 5mo ago
Just to put it out there, since some people might not know, Koshiro who is probably best known for Streets of Rage , is widely regarded as a master video game composer, right up there with Nobuo Uematsu ( Final Fantasy ) and Koji Kondo ( M
367.
by vunderba 5mo ago
That sucks to hear - hopefully it'll be addressed! I'm a huge shmup fan particularly danmakus. Shmups from Cave in particular do a fantastic job: even when the screen is absolutely saturated with bullets like having a bag of dried
368.
by vunderba 5mo ago
And Rocket Knight Adventures - which is quite colorful. https://en.wikipedia.org/wiki/Rocket_Knight_Adventures
369.
by vunderba 5mo ago
Anyone remember the book "Absolute Beginner's Guide to QBASIC"? https://www.thriftbooks.com/w/absolute-beginners-guide-to-qb... Like most people, I also learned how to program in BASIC from a mustachioed
370.
by vunderba 5mo ago
Awesome, you're one of the few people who played nearly the same set of games I did. My favorite was probably The Island of Dr. Brain, though all of the Dr. Brain games were great as a kid. One particularly hilarious part was right at
371.
by vunderba 5mo ago
Much of this comes from the fact that, as is true for almost everything, an LLM (generative model etc) presents itself as an expert. It'll very confidently produce results that, to a layperson, look quite good. But the more of an expe
372.
by vunderba 5mo ago
> When they had a game over, they could spend the wealth accumulated to unlock more valuable crops that would start spawning in their next attempt. This is what Hades and other "persistent rogues" for lack of a better term do
373.
by vunderba 5mo ago
> Chess is a lot trickier than it looks. It has so many rules: castling, en passant, pawn promotion, pinning, the discovered check, and the deadlock case of stalemate. As a kid playing chess with other neighborhood kids back in the day
374.
by vunderba 5mo ago
Nice job. The mazes aesthetically remind me heavily of the pseudo-mazes you could generate with a C64 one liner: 10 PRINT CHR$(205.5 + RND(1)); : GOTO 10 Related video: https://www.youtube.com/watch?v=m9joBLOZVEo
375.
by vunderba 5mo ago
It's a toss-up - NB Pro (and NB2) were in the lead for a long time, but gpt-image-1.5 and then particularly gpt-image-2 pretty much closed the gap on my GenAI Image Showdown benchmark site. NB still generates better looking images thou
376.
by vunderba 5mo ago
I do have to kind of chuckle a bit at no land destruction but Sol/Mox is fine. I guess it's fine if you're playing with proxies so the playing field can be levelled otherwise... RIP for the less wealthy opponent.
377.
by vunderba 5mo ago
I had a friend back in 4th Edition who ran a blue/black deck stuffed with counters and discard stuff like Counterspell and that damn Hypnotic Specter, a flying creature that made you discard a card every time it hit you. And not just a
378.
by vunderba 5mo ago
Nice job, it kind of reminds me of College Bowl trivia. The idea is that each card contains a lot of information, and players can “buzz in” as early as they want as the card is read out loud. The clue on the card starts out extremely esoter
379.
by vunderba 5mo ago
I wouldn't worry too much about it. I think the big AI standout (not your fault) is the classic GenAI lightning that behaves more like a weird viscous liquid. Another one I see a lot is the number of avian creatures flapping their wing
380.
by vunderba 5mo ago
Apparently this site is by the same person who created js-dos [1], which is an absolutely fantastic emulator for running and hosting DOS games in the browser. I used it quite successfully for an official sequel to an old DOS game a few week
381.
by vunderba 5mo ago
yeah, I ca see that. A lot of that (particularly for RPGs) is more that you’re trying to simulate a real-world experience. How do you represent how dextrous a player character is, or how hard someone can bludgeon you over the head with a mo
382.
by vunderba 5mo ago
> Pretty much any game does (even the ones I enjoy) The idea of video games being the equivalent of coin pusher arcade garbage is a relatively new one. If games were just non-deterministic randomized RNG generators than the entire spee
383.
by vunderba 5mo ago
So I use a combination of Neo Forge and ComfyUI. Forge has an easier learning curve but ComfyUI gets all the new "hotness" almost immediately since there's so much custom nodes for it. If you're on a Mac, I've heard
384.
by vunderba 5mo ago
Yeah - if that's true then it's even worse because the output price says $0.067 per 1K image*, $0.101 per 2K image*, and $0.151 per 4K image*. But if all the "compute time" is spent on a 1K image and they're
385.
by vunderba 5mo ago
With the number of fine-tuned LoRAs and checkpoints - from a realism standpoint, yes SDXL is still very viable. From a prompt adherency perspective, absolutely not. Qwen-Image-2512 / Z-Image / Flux.2 absolutely crush SDXL if you&#
386.
by vunderba 5mo ago
Where did you get that info from? According to Google's own docs as well as my own image generation tests via the API, it supports up to 4K natively for gemini-3.1-flash-image-preview (aka NB2). It just defaults to 1K. But I didn&#x2
387.
by vunderba 5mo ago
Definitely. I run an entire site built around a series of benchmarks that focus on prompts of increasingly difficult complexity with a focus on adherence , and even the state-of-the-art local models are probably only about thirty percent a
388.
by vunderba 5mo ago
Right I think that’s why you probably need to start with very low levels of denoising and experiment with different approaches. Set up as a ComfyUI workflow that does a few things: it tries SDXL, Flux, and a couple of different denoising me
389.
by vunderba 5mo ago
Yup, OOC a while back I put together a ComfyUI node that took in a NB image and start with the smallest amount of denoise strength using Flux.1 (but works with any model), then run img2img with a synthid check incrementing denoise in a loop
390.
by vunderba 5mo ago
Nice job. I think it would be even better if there were some kind of visual reward for getting an answer correct or not. I tried Quick Play, and I had to fill in every single cell and then hit Submit before I could see what I got right or w

More ›