Looks beautiful, and UI-less seems very noble.

I agree this looks really great and looks like fun, too. The rocket fists are cool.

Stress tests are cool, too. What I meant by compiling to JS was sort of a loop unrolling thing. So, say, instead of this:

Code:

output = dot(weights, inputs);

You'd have something like this (inventing weights):

Code:

output = 0.7 * a + 0.3 * b - 0.5 * c ... and so on.

That could end up being a bunch of ugly code for your whole network but automatically generated.

I'm not sure what the difference would be. Just seemed like one potential trick to try if you want to squeeze performance. Running AI in a web worker background thread might also have value. Not sure. Anyway, just thinking random thoughts. As I said, I think your stuff's pretty cool.

And yeah, I really enjoy this community, too.

Also, you might compile your learned networks to JavaScript and let them be jitted for speed. Anyway, this game is definitely slick looking, and it's nice to be aware of all your technical effort, too.

I think 3D is a great idea. What kind of rendering are you doing here?

100 cars is a brave goal. Good eye to scalability. Simple heuristics are likely enough. However, offline learning should still allow building an efficient model. How big are your neural nets? How big/slow is your feature set, and how much lookahead do you expect to need for your manual AI?

A score could also just include performance at high load cases, too, instead of trying to design a metric.

Does look beautiful. As an aside, if you want a realistic timeline, I personally think 2081 is at least a few centuries too soon for widespread solar system colonization.

Quick tip: For SDL on Android, use the latest from hg, not the 2.0.3 release (especially if you want SDL_GetPrefPath to work).

Android was simpler with OpenFL/Lime, I'd have to say.

Anyway, latest code base nominally working now on Windows, Linux, and Android. Next chance I get, back to doing things (minimal UI then character animation).

Because I'm crazy, I've started my second rewrite. Now in C++ instead of Haxe. (I originally started in TypeScript.) My reason is that I switched to wanting pixel-specific rendering, and tile-copying in OpenFL was going to be a mess for that. And I'm seriously considering procedural animation on the characters and more.

Haxe can compile to C++, but it's sort of slow at that. And Neko was sometimes substantially slower at runtime for serious loops. And I was still trying to learn what was efficient or not in Haxe -> C++, too. I'd sometimes do this by looking at the C++. Just coding in C++ (with faster compiles now) is more straightforward.

So, instead of moving forward still, I'm catching up again. But I do have something new to show. I now have edge highlight/shade for bricks chopped off at edges.

Old ("enlarged to show texture"):



New:



And I'm using std::normal_distribution for Gaussian noise (still bounded) instead of uniform random sampling. And sadly, it seems that the different sampler causes a different set of colors to be chosen, making this slightly harder to compare. But the left and right edges of cut bricks should be visibly different, and there's overall more concentration toward mean values.

For details, I'm on raw SDL using C++11.

Going forward, instead of going into energy block rendering now (which I'd intended before starting over again), I should probably spend some time on UI and then figuring out how to procedurally animate the little dudes.

(Note: Post updated to include old version for comparison, along with a bit of discussion on the matter.)

Sorry to keep yammering, but your work is so cool. I think the scenery you plan so far is also fantastic.

To keep scope down, might it be possible to release each animal as a different game? The first "game" would still be huge scope (having to put all the groundwork in place), but it would make sequels very natural and allow sweet focus on making each creature as amazing an experience as possible.

Makes sense and thanks for the info. I sadly ditched my Amiga 1200 in the early 2000s. Maybe for the better (in terms of declutter), but I do miss the old days sometimes.

Amiga forever! Might want to be careful with side view (flying saucer) vs top view on things?

What do you code in?

Overall sounds like a fun project. (I'd likely do this with arbitrary angles instead of fixed 8 myself, but that's personal preference, I suppose.)

Tigers don't live in Africa naturally, by the way. Maybe lions and cheetahs interact some? Or maybe that's outside your concern. Figured I'd mention, just in case.

It's hard to be sure where you're going with this as a game. As for the look, it seems like you are aiming for realistic, but in that case, I think this is too sparse in its current form. That might or might not matter, I guess.

Looking cool, as before. Curious to continue to hear how the story details end up going. If where a creature sees from matters, then you must be expecting that to play into the AI. Got a lot of work cut out for you (and/or whomever you might manage to get on board for coding). (Edit: Not that this one thing is the big deal. Just a lot of things you have planned suggest a big project overall.)

I think the evolution stuff looks good, by the way. Nice introduction of details that were missing in earlier stages, but it still looks fitting.

I like the Gollum idea, too, however that turns out to be. From what I understand of SotC, you'd be copying it even more if you (or your mount) got eviler as you go. Random other plot idea: Some local warlord wants to kill the great beasts to prove he's awesome or to extract some chemical from their brains or ..., and your protection of them isn't just about that but more specifically about opposing the warlord. And maybe you win some and you lose some of the battles, just based on how you did. And maybe control strategy with AI teammates before going into battle? Anyway, I'll shut up now. Just was pondering this and figured I'd brain dump. I don't mean to be imposing.

Thanks much for the encouragement and constructive feedback.

And hmm, looks like I've written a text wall coming up. Sorry about that. Notes for myself, at least.

So far, no one claims to actually like the new graphics anymore than anyone claimed to like the old. Neither has anyone claimed to find the game plan itself engaging, although I still haven't gotten most of the new features done yet, either. So, it's hard for folks to know the full plan. I am glad people appreciate the development story of the game. Still, I need to assess my goals. These are among the current big ones:

• Actually finish a very old project.
• Maintain some commitment to the old plans.
• Try to make it fun and see if I can get people playing it.
• Have fun with my kids.

I want to promote the game, if I can make it good, but I don't expect to make much if any money on it. If I can finish it, here are my current deployment plans:

• Give it away for free (without ads or any other junk) on Android.
• Depending on the success there, either give it away or try to sell it on Windows and Linux.
• Depending on the success there, maybe get around to Mac and iOS.

I won't be hiring a pixel artist for two reasons: (1) I don't expect to make much money on this anyway even if I can get some people interested, and (2) I'd rather get personal experience. And I have looked up some pixel tutorials by now, including having previously looked at Derek Yu's. (Still thanks for the recommendation as it gives me context and reminders.) But it's pretty clear I'm not winning any visual awards right now, and I want to move forward anyway. So, here are my graphics plans:

• I'm not really a pixel person. I used pixels before, but only because that's how things were. Still, I'm sticking to them here.
• Also stick to the current new graphics for now, but I'm not committing to it or anything else.
• Don't necessarily try to conform to mainstream pixel art, but still looking good is nice, if I can figure it out at some point.
• Do stick to current tiny pixel counts (320x200 with 8x10 grid).
• Get more pixel-specific on rendering, including procedural animation. If I have just a few pixels, I might as well have fun with them.
• Back-fill support for the old 90s tiles as a display option at a later point, if I make the time for it.

I'd love to learn pixel shaders, but I'll defer that to the future. I can manage pixel-specific CPU rendering at 320x200 quite easily at 60 fps, if I stick to native code.

Pixel-specific rendering doesn't mean the gameplay changes, and as I introduce new features, I'll show where the feature came from in the old tiles and concept level layout. (I have sometimes fine-tuned ideas on how to use the old plans, but again, I currently still only plan to use features that were approximately intended in olden days.)

To clarify, I have nothing against ethical hunting, but I fear the "That's big amazing wonderful cool unique and special so I want to kill it" syndrome. Fun could be to keep the big animals dangerous, though, so maybe you'd have to avoid being killed by them while also protecting them? Anyway, just some thoughts.

Big monsters are awesome (a la SotC), and I agree your visual style is great. Animals are cool. Have you considered maybe not being about killing them, though? Maybe saving a la Princess Mononoke?