gsoutz 2025-04-07 09:49
impressive overall, idea code graphics and controls. couldn't get into the game loop though supposed to match left to the right but how? good job pulling this off.
Foon → Ludum Dare Explorer → LD57 → Neuronic Netronics
By kaliuresis
| Category | Rank | Score | Count | |
|---|---|---|---|---|
| Overall | 62 | 3.76 | 25 | |
| Fun | 152 | 3.30 | 25 | |
| Innovation | 4 | 4.52 | 25 | |
| Theme | 223 | 3.17 | 25 | |
| Graphics | 123 | 3.60 | 25 | |
| Audio | 179 | 3.04 | 23 | |
| Humor | 239 | 1.88 | 20 | |
| Mood | 203 | 3.15 | 21 |
impressive overall, idea code graphics and controls. couldn't get into the game loop though supposed to match left to the right but how? good job pulling this off.
Unfortunately I am too dumb to understand the game. But I liked the idea, animations and the mechanics.
Can you please reply to this tomorrow so I try this one again? I'm confused but very intrigued
This is a really cool game, I think that especially the was the sounds and graphics interact really sell what you are going for. Moving the connections around and sliding the sliders was cool. I did have a lot of trouble figuring out how to get connections to form, and the overall goal of the puzzles I did solve was still a mystery, I ended up messing with things until it worked. I think that with just a little explanation on what the individual pieces did it would have been a lot more approachable. Aside from that though this is a very impressive piece of work, well done!
Oh my gosh, I don't understand! Or at least that's what I feel like, though I managed to fumble through all of the levels, and even get the special ending, however it felt like I had to cheat for that one - I modified my graph half-way through and for some reason sometimes it didn't reset the progress and it gave me the win. Was it possible to do somehow without that (I can't imagine how, since there's only one input, but... I might have missed something)?
Screenshot 2025-04-09 172341.png
It took me a while to figure out what my available tools were - in the end I decided it was "add" and "multiply by -1...1" and, with what felt like a hack, "generate a 1.0", is that what they're supposed to be?
Doing this all in WebAssembly (5K lines of it?!) is really impressive and cool. The visuals were wonderful and I love the excessive fire effects (reminds me of old ASM demos on the 486 :laughing:).
Pretty funky puzzle game, I stopped on the 7th round. It could definitely use a bit more of a tutorial. I nearly gave up on the second puzzle just because I had no idea what I was doing.
@gargantuan-janitor Here's your reminder
@jimbly That sounds like a bug, but it is possible to solve without that. Those are the intended tools, they're roughly based on how real neural networks work with linear layers followed by a non-linear activation function (in this case the function is clamping to [-1, +1]). There are also a couple other "hack" tricks that can be useful.
@talif-moor Yeah, I knew it could probably use some explanation but I like non-verbal puzzle games and the thematicness of unsupervised learning too much to give up on that idea. I still think it could work without explanatory text but it would need a lot more fine-tuning and play-testing to make the visuals and feedback more clear. I've added some hints to the description to help explain the basic mechanics.
@kaliuresis thank you for the reminder.
I love this kind of game and I loved figuring this one out. I do think a slightly better tutorial is needed, if only because I would not have figured out that links are sliders if I hadn't opened the hints. I do feel I'm fighting the interface half of the time. Wanting to see how tweaking some values around works for one input i part6icular is super annoying, as you have to wait for the entire cycle and only get a couple seconds to do so. Also, if there's another way of linking a node to the output that isn't at creation, I haven't figured it out, which makes trying stuff near the output very annoying every time you want to re add a link you closed. I'm enjoying it, will probably keep playing a bit before rating it, but these seem like the frustrating reasons I would drop this game over. The sounds and the graphics are really nice and I would love to hear about how you hand write wasm.
@gargantuan-janitor You can create links from non-input nodes by clicking near the edge of the circle, I'll add that to the hints. For the automatic case switching I chose to do that since there are some puzzle exploits from letting you manually switch cases, but maybe there could have been a better solution. It might also have just been fine to let people cheat to skip levels considering how difficult they can be, and the cheating method is not very obvious either.
My setup with wasm is that I write to a wat file (the text format for wasm) in vim, with a build script running the wat2wasm command whenever I save. The javascript side continuously checks the hash of the main.wasm file, and will reload it when it changes. I have the data separated from the update loop, so that it can be reloaded without losing state or having to refresh the page. At least for a project this small everything was nearly instant.
The graphics are a buffer that gets directly copied onto the canvas from the js side, and similarly the audio is just writing samples from a ring buffer into a js audio worklet. For data structures I have hard-coded addresses for everything, which would probably be a problem if the project grows but it worked out fine for a jam.
This was my first time trying this and I just learned things by reading the wasm docs (which are thankfully not too long) about a week before, so there are definitely ways this could be improved. One thing I originally wanted that I didn't have time to setup before the jam was a custom pre-processor that I could use to have macros, binary file embedding, and a more flexible way to layout memory.
I definitely wouldn't recommend doing hand wasm like this except for fun or to optimize specific functions, but I think a similar setup with a higher-level language before wasm would work very well. The normal way web builds are done is with emscripten, which is meant to let you run desktop apps on the web and consequently it has tons of bloat like emulating a file system which isn't really necessary if things are made for the web in the first place. As a comparison the last web game I made for ludum dare used emscripten and was a 2.58 Mb wasm file, while this one has a 15 kb wasm file.
Awesome man, thanks for the thorough reply, it's been illuminating. I'm really enjoying this game and I'd love to see it expanded after the jam is over, it feels like it has potential to be a commercial product.
Very cool interpretation you came up with. I have to admit though that it's kinda beyond my understanding. Probably really cool when you are already into the entire topic. Are you planning on continuing this project and maybe add a tutorial?
This is a really interesting idea but it could do with maybe some more tutorialization, and it would help massively to be able to save certain structures or at the very least copy over stuff from the previous level, since most of them seem to quite naturally follow on from the previous level. I think especially for the not gate-related levels the idea that you have to capture a value in a loop of nodes to get a constant value is not very obvious, especially when you can't connect two nodes directly to each other in a loop. Overall, a really interesting idea though, and very well-designed levels, and of course impressive given you're hand-rolling wasm.
Love this entry! Messing around with the mechanics and figuring out what's possible was really fun. I know the whole "figure it out yourself" design philosophy isn't for everyone, but I personally really enjoy it. Bonus points for writing it in Webasm, of course.
Works very smoothly and I'm sure you would if you had more time but this kind of game urges for an efficient tutorial
Very cool concept!
Lacking a bit of explanation about dragging the weight along the edges.
For me, the -1,1 range is really hard to understand with your visualization. Maybe a bottom/top splitting of the node and edge would make it more visual.
I'd love to know more about asm and the dev setup you have use. Have you per chance use git / github during the dev process ? Would you share the source code that way ?
I am making a list of github projects I have rated [github list of ld57 entries](https://github.com/stars/dhmmasson/lists/ludum-dare-57-compo), and I'd love to have your entry in it. (see my post on tips for a repo if you wish [see my post on preping your repo](https://ldjam.com/events/ludum-dare/57/focus-hunt/preparing-your-github-repository-for-ludum-dare-57)
@dhim Definitely agree that a different visualization method could work better. Having something with more continuity across zero would have helped make it slightly more intuitive.
I had a local git repo which is included in the source zip. I've put it on github just now: https://github.com/kylerNat/ld57-Neuronic-Netronics/tree/main
Really wish I could figure this out but I haven't dug enough into NN's I guess! Loved deep learning as a take on the theme, and with some more tutorialization a game like this could definitely be used to help people understand the fundamentals of NN's. I hope it gets developed out a bit more with the feedback from the comments here. Nice job setting it up with web assembly too!
This game is, interesting - cant quite wrap my head around how it works, but i love the graphiks and sounds, feels very fitting. I thought i kinda got it, but even just to reverse a signal seems to be kinda hard - i didn't quite see how what exactly influenzed what - and in the end i just made a connectionless strobo maschine that wouldn't stop flickering what so ever xd
Way too confusing for a game. Needs a tutorial. I.e. few braindead levels where player is told explicitly what and how he has to do. Only after thenm when they see how thing works, you give them (us :) ) something requiring actual thinking.
As for WASM, I guess coding in wat is educational and flexible, but have you checked out AssemblyScript? It's easy to setup, compiles fast and outputs relatively small wasm.
@baturinsky The first three levels were intended to be the braindead tutorial levels where the only challenge is figuring out the controls and there are only a small number of possible things to click, but I definitely still underestimated how confusing they are.
I hadn't heard of AssemblyScript before, it looks interesting. I was planning on doing C to wasm for next time, but AssemblyScript looks like an potential alternative
I was drawn to this entry because I worked with neural nets in my last few entries. This is a very interesting concept that may help in teaching neural nets, and also well executed. I got stuck at level 5 though because I can't find a way to put a bias on a neuron. That is, you basically have to turn a 0 into an 1, but I can't do this with only multiplication. So, did I miss a way to add bias to a neuron?
@loveapplegames There are a couple ways to solve level 5. The main intended way is to (scroll to reveal) ``` Make a loop of neurons, which will stay on forever. ``` The other method, which is a bit less practical, is to ``` Create a delay line long enough to pass the signal from the on case to the off case. ``` There's also a bug where deleting nodes/links does not count as editing your solution and doesn't clear your progress, but that is very much a bug and not intended or necessary to complete any levels.