Most active commenters

    ←back to thread

    Making games in Go: 3 months without LLMs vs. 3 days with LLMs

    (marianogappa.github.io)
    358 points maloga | 14 comments | 24 Aug 25 15:01 UTC | HN request time: 0s | source | bottom
    Show context
    starchild3001 ◴[24 Aug 25 17:30 UTC] No.45006027[source]▶
    What I like about this post is that it highlights something a lot of devs gloss over: the coding part of game development was never really the bottleneck. A solo developer can crank out mechanics pretty quickly, with or without AI. The real grind is in all the invisible layers on top; balancing the loop, tuning difficulty, creating assets that don’t look uncanny, and building enough polish to hold someone’s attention for more than 5 minutes.

    That’s why we’re not suddenly drowning in brilliant Steam releases post-LLMs. The tech has lowered one wall, but the taller walls remain. It’s like the rise of Unity in the 2010s: the engine democratized making games, but we didn’t see a proportional explosion of good game, just more attempts. LLMs are doing the same thing for code, and image models are starting to do it for art, but neither can tell you if your game is actually fun.

    The interesting question to me is: what happens when AI can not only implement but also playtest -- running thousands of iterations of your loop, surfacing which mechanics keep simulated players engaged? That’s when we start moving beyond "AI as productivity hack" into "AI as collaborator in design." We’re not there yet, but this article feels like an early data point along that trajectory.

    replies(23): >>45006060 #>>45006124 #>>45006239 #>>45006264 #>>45006330 #>>45006386 #>>45006582 #>>45006612 #>>45006690 #>>45006907 #>>45007151 #>>45007178 #>>45007468 #>>45007700 #>>45007758 #>>45007865 #>>45008591 #>>45008752 #>>45010557 #>>45011390 #>>45011766 #>>45012437 #>>45013825 #
    1. zerr ◴[24 Aug 25 17:42 UTC] No.45006124[source]▶
    My litmus test for generative AI: generate a complete spritesheet for a 2D pixel art action game, e.g. only for the battle tank or main hero movements. No success so far.
    replies(7): >>45006295 #>>45007019 #>>45007020 #>>45007314 #>>45008002 #>>45010220 #>>45011955 #
    2. smokel ◴[24 Aug 25 18:02 UTC] No.45006295[source]▶
    Mind sharing what you have tried? Have you considered training a diffusion model on pixel art, and then conditioning it on a 3D model?
    3. davepeck ◴[24 Aug 25 19:35 UTC] No.45007019[source]▶
    Ive never once successfully gotten a usable sprite sheet out of ChatGPT. The concept seems foreign to it and no matter how hard I try to steer it it’ll find a way to do something hopeless (inconsistent frame sizes; incoherent animations; no sense of consistent pixel sizes or what distinguishes (say) 8-bit from 16-bit era sprites; it’ll draw graph paper in the background for some reason; etc etc.). If anyone has a set of magic prompts for this, I’d love to learn about it. But my suspicion is that it’s just fundamentally the wrong tool for the job — you probably need a purpose-built model.
    replies(5): >>45007498 #>>45007632 #>>45008828 #>>45008948 #>>45009184 #
    4. maloga ◴[24 Aug 25 19:35 UTC] No.45007020[source]▶
    Agreed, I got an LLM to build me a Super Mario Bros game with 0 code and it's playable after 2 prompts but I'm stuck with the spritesheet as well.
    5. __loam ◴[24 Aug 25 20:07 UTC] No.45007314[source]▶
    Unless you're using a model that was built for pixel art, you will never get a usable piece of pixel art out of a model using a diffusion based image generator because it does not understand what a pixel is. You'll always get pixels bleeding into the others, shitty outlines, and nonsensical AA. They simply do not understand the medium.
    6. typpilol ◴[24 Aug 25 20:30 UTC] No.45007498[source]▶
    According to all the lazy articles I've read here lately you just need to threaten to beat it up lmao...
    7. mac-mc ◴[24 Aug 25 20:48 UTC] No.45007632[source]▶
    Like a full sprite sheet, one sprite at a time or a sprite animation loop?

    Each one would require a different kind of model and model technique to make, so I wouldn't be surprised that ChatGPT has issues with it. A sprite animation loop would be better done by a potentially specialized video-oriented model, for example, and the current image and video models are barely trained on that kind of video data.

    8. peterashford ◴[24 Aug 25 21:33 UTC] No.45008002[source]▶
    Yeah I tried to get a spritesheet generated. Absolute rubbish.
    9. ◴[24 Aug 25 23:43 UTC] No.45008828[source]▶
    10. frozenlettuce ◴[25 Aug 25 00:02 UTC] No.45008948[source]▶
    that might be possible by asking it to create an 3d model with animations (based on a template) and then capture the sprites. but then again, not sure if building it would be worthwhile because 1) openai might add that as a native product (like what happened with .ppt generation) or 2) the capability to do so might be 6 months away
    11. nkrisc ◴[25 Aug 25 00:49 UTC] No.45009184[source]▶
    Have you tried drawing?
    replies(1): >>45014755 #
    12. sandspar ◴[25 Aug 25 04:31 UTC] No.45010220[source]▶
    Maybe AI games can use AI-first art styles. If I ask a camera to create an André Derain painting then it will fail. But that's my fault for misunderstanding the medium.
    13. jamilton ◴[25 Aug 25 09:30 UTC] No.45011955[source]▶
    I've seen people make finetunes for character turnarounds, there's probably one for 2D pixel art in particular, or perhaps combining such a finetune with a pixel art finetune would work.

    It looks like retrodiffusion.ai in particular has something close.

    14. ◴[25 Aug 25 15:13 UTC] No.45014755{3}[source]▶