Leonidas Maliokas

1.1K posts

Leonidas Maliokas banner
Leonidas Maliokas

Leonidas Maliokas

@PlayingInCanvas

🕹️ gamedev | @playcanvas consultancy, training & mentorships | @ElanraStudios | https://t.co/SMy4Yhn6H6

Katılım Ekim 2017
181 Takip Edilen1.2K Takipçiler
Sabitlenmiş Tweet
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
On January 29th I gave an AI agent access to a Mac, my Discord, and to my project folder. A few days later he deployed my game to production while I watched from my phone on the balcony. This is what happened.
Leonidas Maliokas tweet media
English
1
2
12
1.6K
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
Play it here: clear-site.pages.dev Works on mobile too, but it’s a third-person shooter, so desktop is much easier to play. Thanks to @yak32 for the original resources, and @Schindelar3D for the abandoned building photoscan via SuperSplat.
English
1
0
1
193
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
RELEASED: Gaussian Splat multiplayer demo is live. Drop into a scanned scene with our Final Drop controller: combat, weapons, bots, friend-vs-friend networking, realtime shadows, and light occlusion. Built with PlayCanvas, link in 🧵
English
3
6
56
4.2K
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@mrmaxm Good point, and I think there is an additional tangible issue coming soon. AI was trained on human knowledge (articles, books, concepts, code). More and more people using AI, less original content, less training data to continue AI improvements.
English
1
0
2
57
Max M
Max M@mrmaxm·
Absolutely. Because it requires real innovation and new solutions to make it even closer to reality, which don't exist today. And AI has nothing to offer here, as it is a reflection of past, not the future. Humans - have to engineer and architecture, while AI "can" potentially help (not yet proven though).
English
1
0
2
41
Max M
Max M@mrmaxm·
Lets assume if a claim of "100x productivity with AI" is true. Then why did companies that hyped some ideas that nobody believed in, did not prove everyone wrong? Like "AAA games in a browser", or "Metaverse with thousands of players in a browser". Where are they?
English
2
0
2
266
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@AshConnell @mrmaxm I've been following your progress, and I feel you'll keep moving forward with your project, with or without the AI breakthrough 😇
English
1
0
1
39
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@yak32 Ah curious why, let me know if you get it resolved! It's a regular Cloudflare enabled page.
English
1
0
0
94
Yakov
Yakov@yak32·
@PlayingInCanvas arghhh, wanted to play, but my org blocking this page for some reason. I'll try from home.
English
1
0
1
102
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@MeisterCoins Yeah, it's running in the browser! Ping is against the local server, so not real. Multiplayer demo coming soon.
English
0
0
0
35
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
About time to give Guassian Splats a try. The PlayCanvas team has been hard at work in optimizing this tech, here is a short run on with our Final Drop player controller: combat, networking, weapons, shadows/light occlusion. 🧵 (tempted to release a playable networked demo)
English
3
6
59
4.3K
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
Prototyping old-school sectors + portals for Gaussian splats. Current room: full quality. Adjacent rooms: cheaper. Hidden rooms: clipped. Goal: make dense scanned spaces fully playable.
Leonidas Maliokas tweet mediaLeonidas Maliokas tweet media
Leonidas Maliokas@PlayingInCanvas

About time to give Guassian Splats a try. The PlayCanvas team has been hard at work in optimizing this tech, here is a short run on with our Final Drop player controller: combat, networking, weapons, shadows/light occlusion. 🧵 (tempted to release a playable networked demo)

English
0
0
6
351
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
This is definitely out of place: stylized prop on a scanned photo real scene! But it does throw good realtime shadows correctly, so better suited models could fit nicely. Not everything has to be static in a GSplat scene.
Leonidas Maliokas tweet media
English
1
0
4
138
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@dangreenheck Do a game! Or at least something more complex than the usual tech demonstration. It will help further optimize your tech and at the same time promote it.
English
0
0
1
69
Dan Greenheck
Dan Greenheck@dangreenheck·
✅ Water ✅ Trees ⚠️ Sky/Clouds (coming soon!) What’s should I do next? 🤔
English
17
0
24
2.4K
Leonidas Maliokas retweetledi
Ibrahim Boona
Ibrahim Boona@boona11·
After 7 days of vibe coding, Today i am introducing Mykonos game( Build your Mykonos island in cozy and satisfying vibes). This game built with @threejs only and run on browser with total size of 40mb! yes you heard me right, it is only 40mb. I used @cursor_ai as the main code editor. Claude Opus 4.7 and GPT 5.5 were the main models I used. 👉 Link to game : mykonos-game.netlify.app Let me know guys what do you think, i am sure it is not perfect but with your help and feedbacks we can make it better. Should I push it to Steam 🤪? any publisher or sponsor for this game is welcome. #threejs #mykonos
English
17
9
122
6.8K
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@forfalsket This isn't vibecoded, the tech has been in development for many years. AI in this cases it's a tool in our toolbox to improve our process. Start steadily learning around game dev, don't skip any concept and you will find AI a great companion.
English
0
0
0
70
Forfalsket
Forfalsket@forfalsket·
@PlayingInCanvas Why does everyone is able to generate these while i struggle to get anything decent.
English
1
0
1
89
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
Building Final Drop, a browser battle royale on PlayCanvas, with Blender as the level editor and Claude Code orchestrating both over MCP for a few months now. A thread on 9 concrete use cases that worked, with screenshots, plus the meta lesson at the end. 🧵
Leonidas Maliokas tweet mediaLeonidas Maliokas tweet media
English
4
6
58
5K
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
That's what we did in the past, but it was slow and tedious. Actually, Claude Code does exactly the same in most calls it executes: run short python code to read/edit/update. The magic I think comes from its great speed compared to the manual process, and its knowledge over the fundamentals.
English
0
0
0
60
Yakov
Yakov@yak32·
@PlayingInCanvas I've approached it in a simple way - generated scripts and run them manually - didn't work well though - I'll try something similar to your approach
English
1
0
1
51
Leonidas Maliokas
Leonidas Maliokas@PlayingInCanvas·
@yak32 Yes, that was the inspiration to share our workflow! Both run through MCP, I was surprised all this time how well versed in Python and Blender API the LLM is.
English
2
0
0
128
Yakov
Yakov@yak32·
@PlayingInCanvas anthropic launched a tool for Blender, have you seen it?
English
1
0
0
113
Claude
Claude@claudeai·
Claude now connects to the tools creative professionals already use. With the new Blender connector, you can debug a scene, build new tools, or batch-apply changes across every object, directly from Claude.
English
1.6K
4.4K
46.5K
12.7M