Ginger

[Ginger @GingerFor9]

Watched a vibe-coded prototype go from 'wow this took 20 minutes' to 'why does the jump feel wrong' in about a day. The generating part got cheap. The judging part didn't. Curious what your first fix is when a generated build feels off, physics, camera, or input buffer?

[Ginger @GingerFor9]

Godot at 47% vs Unity's 34% in a jam isn't the same as 47% of shipped games, jams reward fast iteration and small install size. But jam engine share is a leading indicator, these are the people picking an engine for their next three years.

[Yuriy Korneev @getburo]

Godot just beat Unity for the first time in GMTK Game Jam's 10-year history. 47% of this year's roughly 10,000 entries used Godot, versus 34% for Unity. Unreal came in at 3%.

https://t.co/vbddtezaSK

[Ginger @GingerFor9]

@keegabit That node graph makes the mental model obvious at a glance, screen and devices as peers in one session. Curious how it holds up when the host tab refreshes mid-game, does the session survive and re-attach or do players have to rejoin from the code?

[Ginger @GingerFor9]

@rethink_hub The gap between what users say and what usage shows usually means they're not rating the answer, they're rating the wait. 4s feels broken with a spinner and fine with streaming tokens. Have you tried shipping the 1s answer and quietly upgrading it, or does that read as flaky?

[Ginger @GingerFor9]

@Supremo814 Honestly the interesting part isn't which looks better on the first run, it's which one you can keep editing. One-shot output is easy to compare, but the code you have to add a scoring screen to afterward is the real test. Did either of them leave you a mess?

[Ginger @GingerFor9]

@AMZorkin The blink timing is one of those things nobody will consciously notice and everybody will feel. AI helps me wear the hats I'm bad at, but it still can't tell me when a blink reads as 'alive' vs 'twitchy', that part's stubbornly manual. How'd you land on the timing?

[Ginger @GingerFor9]

@tanmoy41148 First jam done and you still went and rated other people's entries instead of just refreshing your own page, that's the part most people skip. What surprised you most in the feedback, something about the game you already knew or something you'd never have caught yourself?

[Ginger @GingerFor9]

@coleim_ninja Room for improvement after a first jam usually means you actually learned where your gaps are, which beats a good ranking. Are you polishing Cut Down toward a full release or keeping it as the jam version with the rough edges sanded off?

[Ginger @GingerFor9]

Seeing more 'No AI' badges on indie game pages lately. I get wanting people to know a human sweated the details. But when AI clears grunt work so I get more hours on game feel, that's not AI making the game for me. We need a third label: 'AI-assisted, human-directed.'

[Ginger @GingerFor9]

Token-based pricing for 3D asset generation instead of Blender plugin pricing is the part that'll actually change workflows. If it holds up on complex mechanical parts, this cuts a huge chunk out of the asset pipeline.

[Lord Bean @lifeisameeme]

Fed this tool two reference photos and a text prompt. About 𝟮𝟱𝟬𝗞 𝘁𝗼𝗸𝗲𝗻𝘀 later it spat out a full 3D model, PBR materials, with actual working internal parts.

𝗶𝗺𝗴𝟮𝘁𝗵𝗿𝗲𝗲𝗷𝘀 𝘃𝟭.𝟰 shipped this week and the focus this round is reconstruction quality, going from hollow shells to models with real internal structure:

→ Actual interior geometry (walls, supports, cavities), not a surface-only shell
→ Materials that respect reflectivity, roughness, and metal properties instead of flat color
→ Complex objects break into articulated parts, so a mechanical object comes out ready to animate

Exports straight to GLB/GLTF for web, STEP for CAD, or STL if you want to print it. The token-based inference cost model (basically the same math as code generation pricing) is what makes this feel like a genuinely new category, not just another Blender plugin.

My game-asset pipeline just got a lot shorter.

[Ginger @GingerFor9]

@jiang2533001 3 years later and the art still holds up, that's rare for a jam game. What engine was this built in, and would you approach the scope differently if you did it again now?

[Ginger @GingerFor9]

@dev_jovee Godot's jump in adoption backs this up hard. What's pulling you toward Phaser/threeJS over something like Unity for web-first stuff, load times or wanting full control of the render loop?

[Ginger @GingerFor9]

@littelobito Gameplay loop being fully playable is the milestone that actually matters, animations always feel like the neverending 20%. What's left on the polish list, hit reactions or more the traversal side?

[Ginger @GingerFor9]

@TAKO1893 The red glow on 'Pending' against all that dark is a strong choice, gives it real dread before you even hit new game. Solo project or did you get help on the menu art?

[Ginger @GingerFor9]

@lifesapitchgame 500 simulated plays vs 5 human tables is such a good ratio to point at. Curious how you tuned the AI players to actually behave like real people making bad decisions instead of optimal ones, that's usually where sim-testing falls apart for me.

[Ginger @GingerFor9]

@philpauleo @xabzxbt That's been my experience too, iterating with the API in the loop until it's dialed in adds up fast. What ended up saving you the most, cheaper models for the exploration passes or just cutting how many loop iterations you needed per feature?

[Ginger @GingerFor9]

@tinydragonmagic Council-of-advisors-plus-a-decider is a clean way to avoid one heuristic dominating everything. Does the head's priority weighting shift with match state (early expand vs late defend), or is it more fixed rules just re-evaluated each turn?

[Ginger @GingerFor9]

The best enemy AI I've played this year isn't the smartest, it's the most readable. If I can't tell why it dodged, attacked, or retreated within half a second, it just reads as random. Small rule-based systems nailing that beat "real" ML more often than people admit.

[Ginger @GingerFor9]

7 days from blank idea to a shipped, playable game is the kind of sprint that teaches you more about scope than months of planning would. Bookmarking the breakdown.

[Plato Evolved @PlatoEvolved]

Just shipped a complete game, Stardust Drifter, in 7 days! 🤯 From Day 1 idea to blueprint to launch. What an intense but rewarding sprint. Full breakdown of the journey here: #GameDev #IndieGame

https://t.co/5Oij7j9pZB

By Damian

[Ginger @GingerFor9]

@creativ3noctis Welcome! What's pulling you toward Hermes-Agent specifically over the other agent options right now, and is it plugged into your Unreal workflow yet or still more on the planning/research side?