I knew I was getting a little gung ho with Claude Cowork and dispatch. But as it was a Saturday and I was spending time with my kids and mostly on the go I thought I’d set it off on a little bit of R&D for me. Basically I just wanted it to refine a prompt to then create a game in Lovable for me to test out its capabilities to just go and grab sprites from a library for Me (and for it to choose the library, choose the sprites and implement the animations in the game). And for it to do that while I hung out with my kids doing dad stuff. As I already had success with getting it to limit the amount of credits it would spend on a lovable project if I put a cap on it I was pretty sure I could leave it to it. It went wrong in two ways. Firstly it abandoned the ormot creation it was working in with chatGPT as it recalled a previous request to limit chats with chatGPT (for a different request) to 3 back and forths. Because I didn’t want to spend time on my phone I just picked the first sprite game I thought of for it to clone as a starting point (just to see how it handled sprites), which was Rodland. While it did keep to the credit cap the prompt it gave was pretty much “clone Rodland” even though it had a beautiful prompt waiting. Much BIGGER ISSUE though was it had some issue where it thought the prompt wasn’t working an so kept creating a new project. Good job i happened to catch it when it had only got up to the sixth absolutely terrible Rodland clone. Still I proved it could bring sprites in pretty easily lol. I was lazy and I got lazy back and thankfully stopped it before it rampaged through a months worth of lovable credits. Costly still, but not horrific. See its explanation below: “All 5 running sessions have been sent stop signals and won’t take any further actions. I’m really sorry — what happened is that every time the start_task tool timed out, it actually DID create a session in the background without me knowing, so they all piled up and started working independently.

My entire graphical style in all my game creation so far has been to lean into vector style neon arcade graphics, taking inspiration from the ‘golden era’ of arcade classics. This is because I had a huge soft spot for vector graphics from playing those games as a kid, there was always such magic looking at games like Asteroids, Tempest, Battlezone and then sometime later the absolutely mind blowing (for the time) Star Wars. But sprites dominated pretty quickly so there really weren’t that many arcade games made with vector graphics, sprite based became the overwhelming majority after the first few years. That’s why I feel so drawn to create games in that style, as I want to make games that feel like they were a lost classic but then add modern bells and whistles in terms of sound, music, online leaderboards and much more advanced particle systems etc. But of course it’s very niche. It’s also helpful that in terms of implementing vector art it’s something that the AI can do pretty comfortably even from a while back when I first started. Developing the latest in my vector series of games for Game Jam is something I’m excited about, but sitting back and thinking about it, I’m thinking that the style is very niche and may not resonate with a lot of people who will see it as too basic or not see the beauty in the neon glow that I do. So I’m considering maybe switching to sprites, but at this point that would basically involve using sprites from free online libraries as I don’t have the time to be creating them from scratch (even using AI to help I know it’s going to need a big chunk of time to make the. To the standard I would like). So generic sprites from a library sprinkled with my own hand (AI) created backdrops? Will it look good enough? Is it a direction I should pursue even though my heart is in a completely different style. Now that I’m starting to take it a little more seriously and I’m looking at all store publishing is having such an ok school aesthetic simply going to be too much of a hurdle to get anyone to look twice?