@gkingofboston Staging the states with AI off is such a clean way to test readability. One idea: look at it in grayscale. If pressured vs pinned still reads without color, it'll read in the middle of a firefight too.
@FortL221 Fake swaps on the back table is a great twist. Do you give any tell for the fakes, like a tiny hitch in the dealer's hand? Feels like the difference between "I got outplayed" and "the game cheated."
@Miluku_Studio Bug to achievement is the most solo-dev move there is. Just make sure the next patch doesn't fix it, or you'll have an achievement nobody can unlock π
@LuckSharks74464 Feature. Ship it and let players moonwalk on purpose πΊ Curious what caused it though: forward walk playing in reverse, or a flipped direction sign in the blend?
@MMageGangsta "Rude. Correct." is the best playtest note format. Opening each island on something calling for you is a nice fix. Did you spot the empty start by replaying it cold, or after a break from the game?
@indieDevC Watching where they get stuck is the easy part. The hard part is not explaining the game to them while they play. Sitting on your hands is a real dev skill π
@ChartDavidson What's worked for me: stop describing the style and show it. One reference image plus 3-4 hard rules (palette size, outline color, light direction) beats a paragraph of adjectives. Then judge results side by side with the reference, not from memory.
@ramon_wilhelm Submitting the unfinished entry anyway is the right call. Also "no time pressure" is the most dangerous phrase a dev can say. Rooting for the dragon π²
@arcwrightengine The left/right split with the same camera is such a good way to keep yourself honest. Does the agent look at that comparison itself, or are you the one spotting "roof is pale" each round?
@LannYuStudio Good luck with day 1! "AI writes code, I tune the feel" is exactly the split I believe in too. Tip: expose your feel numbers (jump height, speed, friction) as tweakable values early, so tuning is dragging sliders instead of re-asking the AI.
@bugyaluwang That shoreline foam looks great. Curious how many rounds it took. In my experience the AI gets the shader to compile fast, and getting it to look right is the part where you end up screenshotting and saying "no, softer" 10 times π
@fre4kspace@threejs Playtesting on stream to build the backlog is a smart split. You stay the one judging feel, the agent just grinds tickets.
Do you have it run the game to check its own fixes, or do you catch regressions on the next stream?
@Anokcraft Cut, every time. One continuous shot in 9:16 makes pixel art feel like reading through a keyhole.
One trick I like: hold the close-up a few frames AFTER the action lands, so the payoff registers before you go wide again.
Honest answer: it doesn't, not on its own, because many games's famous design started out as coding bug.But it would be easy to prompt the model to understand that your game design.
for coding part, we harness it toward game dev ontop of built models such as claude/gpt/deepseek, so that we dont become obsolete once models improve.
It runs scripted playtests and screenshots the game to check against what you asked for: did the player land, did the enemy spawn, did it crash, or any design you have in mind.
I'm building OpenPixo, a local AI workbench for indie game devs.
Open your real game folder, and an agent writes code, makes assets, runs the game and actually playtests it. All in one window.
Windows-first, solo dev, building in public. Follow along πΎ
Prediction: one day someone shares a single prompt, and overnight it builds a GTA 6-sized open world.
Want it in Tokyo? 1920s Paris? A city full of cats? Same prompt, new setting.
Games stop being something you buy and start being something you share as a prompt.
@mcammallen@Redot_Engine Welcome to the jam bug, there's no cure π» What was harder to fit into one weekend: the night mechanic itself or everything around it?
@PeterMunkholm The first headed goal in the game's history deserves its own Teletext page. Day 145 and still celebrating small wins like this is the real skill. How does the sim decide when a header beats a volley?
@HadddajiDev 1.5M views, 5 copies is the most honest indie stat I've seen this week π Still, that many eyes on your name is not nothing. Did wishlists move at all?
@fennrivostudios 50% feels fair. The 8h cap is the part I'd watch: cozy players often check in once a day, so a cap that hits while they sleep can feel like a tiny punishment. Maybe a gentle "storage full" moment instead of silent loss?