Show HN: AI Game Playtester
ziva.shHey everyone! Wanted to show off a little side project I've been working on. A lot of people have been using AI to one-shot games, but they're often quite buggy - glitchy graphics, collisions aren't setup, or regressions in earlier levels. With an AI playtester in the loop, I've been able to one-shot games like Candy Crush or Crossy Road that would previously have bugs like cut-off UIs or stuck retries.
The way it works basically goes like this:
1/ The playtester starts by reading the code to understand what its parent AI is asking it to play, and to decide what breakpoints, hooks and logging to add 2/ The playtester sets up the scene with the needed environment 3/ The playtester sends a sequence of inputs, then pauses the game 4/ The playtester then uses a vision model to understand the scene, queries the hooks it initially set up, and reads the logs to decide what to do next
After doing a few input loops, it then determines if the game is working, the game is broken, or if the test failed.
It is a paid offering so I'm not expecting anyone to try it out, but I wanted to share it nonetheless ^^. Lmk if you have other ideas on how you think it can be improved
Comments
this is not a crazy idea honestly. i mean, using bots and classic game AI for testing etc is also there ofc and this seems similar but i think considering the state of released games, often game mechanical issues could be fixed
can you give some idea on if and how your model reason about visual bugs though? this is where i find human testing to be more logical. what if a buton click flickers the button instead of 1 time lightup and darken? can it see all the frames and catch it?
this seems where i think a lot of effort is likely still going to be spent so if in some what it could get a description from the code or a person and then reason correctly about visual behaviors in the game.
another tricky area is mandatory testers for appstores. those you can trick perhaps using your AI into thinking someone is testing it really but i am not sure if it would be against their terms... (likely? maybe could be material for exploring market opportunities)