Well I think I am at a point where I am going to call it done and move on to other ideas. This game was mostly about testing my pipeline tools and process for the bigger ideas I am working on, but I made it a point to complete this project like it would be a shipped game. If you play it, let me know what you think

I've been experimenting with a different AI game development workflow and I'm curious if anyone else works this way. Instead of building an ugly prototype first and worrying about the UI later, I start by generating the actual game screens. I create a PRD, then use ChatGPT, Figma AI, or similar tools to generate screen concepts and even prompts for the content that should appear on each screen. Because everything is based on the PRD, the screens already reflect the intended gameplay and progression. Next, I have my coding AI recreate those screens as a working prototype using mock data. At this stage I'm not focused on databases, APIs, progression systems, or backend logic. I'm only testing whether the screens feel right, whether the user flow makes sense, and whether the game is actually fun to navigate. Once I'm happy with the experience, I simply replace the mock data with real systems and connect everything together. I'm currently doing this for a diablo-like idle-game. Below are a few of the AI-generated screen concepts, followed by the current prototype, which is still in the mock-data phase. 👉 https://idle-hero-loot.lovable.app/game So far I've found that this approach helps me spot UX and gameplay issues much earlier than when I start with functionality first. Has anyone else tried a screen-first workflow like this? Do you design realistic screens before implementation, or do you still prefer building the systems first and polishing the UI later? Would love to hear how others approach AI-assisted game development.

Game: Stardust Drifter Idea: An arcade game where you ride the pull of stars to collect dust and reach deeper systems before your fuel runs out, with the twist that the star that slings you across the map is the same one that kills you. Core action: Read the live trajectory line and nudge thrust - you never fly straight, every second you choose to fight the pull or ride it. Goal: Clear a system's dust to open the gate, then push deeper. Fuel carries over. Score is depth, not points. Unique hook: No engine worth the name, no upgrades, no unlocks. The same gravity that kills you is the only thing that moves you. You never get stronger - you just get better at reading the pull. Private note - for this community, not for players. I want to build games that leave you knowing how something actually works, without ever telling you that's what's happening. No lessons, no facts on the loading screen, no quiz. If the player notices, I failed. Which is why the game has to be a good game first - that's not a compromise, it's the whole trick. Nobody plays it to learn anything. But you can't fly a mote of stardust without reading gravity, and you can't survive a star system without learning what its layout does to you. A tight cluster of planets leaves you nowhere to run. A dead star is a trap that looks like nothing. You find that out by dying in one, not by reading it. Every level is a real star system - named after it, shaped like it. TRAPPIST-1, Alpha Centauri, Sirius. The scale gets thrown out, and that costs nothing: what's true at 0.06 AU is true at 300 pixels. Real distances are just empty space. What can't be faked is the shape. Invent a system, staple a real name on it, and you haven't smuggled anything - you've lied. Game first, theme second. Every time I led with the theme, I paid for it with the project.

Hi All, Working on a new game similar to the meowdoku games that have been popping up everywhere. The base gameplay is not a new concept, just another form of minesweeper, but it is a great concept to build off of. I will be adding what I hope is a unique twist to the concept...but had to spend some time nailing the core game play. Same stack as before: Game Engine - Godot 4.7.1 Animation Addon - Kadence (custom addon built by me), for Game Feel and Juice AI - Claude 4.8 Opus ( I wrote most of the core code this time, claude was used to fix upsteam bugs in my animation library and a bug in the wave animation) Tools - WARDEN(Custom AI code policies with a custom CLI enforce layer ...ask me if you want to know more...and custom skills for Godot)

Hey, After watching a Godotcon 2026 talk Mathijs Koning on Tiny Games I was inspired to re-create one of his game in Godot as a way to test some of my custom tooling. The first on the list was Duel Dice, you can find his original game here. The plan was to create the game in my vision with the tools I have built and "dogfood" them to find any holes, bugs, or features I needed to add before using them in any real projects. The plan is to recreate all 12 of the games he prototype with my own unique spin on them. For those who may want to know: Game Engine - Godot 4.7.1 Animation Addon - Kadence (custom addon built by me), for Game Feel and Juice AI - Claude 4.8 Opus ( 50% base code and logic were written by me, Claude took over to help me finish and polish the game) Tools - WARDEN(Custom AI code policies with a custom CLI enforce layer ...ask me if you want to know more...and custom skills for Godot), Custom MCP for Elevenlabs Design - Affinity Designer and Claude Design Sound - Elevenlabs Since I am using addons, the game can not be exported to HTML...so I recorded a video of the game play. If anyone is interested in playing the real game ... let me know and I will provide a zip. Feedback on the video is appreciated and please play the original game by Mathijs