craigr retweeted
Everyone will get better models. The most interesting consumer AI companies will also get better every time someone new joins.
15.9ms → 4.9ms 👀
Another project using PixiJS 3D, and it's up to 3x faster 🎉
Go check it out!
pixi-effects 0.52.0 をリリースしました。
3D の描画エンジンを、サードパーティの three.js から PixiJS 公式の 3D エンジン「PixiJS 3D」(@pixi/3d) に全面移行しました @PixiJS
pixi-effects は、動画を「データ」として書く PixiJS ベースのライブラリです。テキスト・図形・画像・音・カメラを普通のオブジェクトで並べるだけで、ブラウザで再生でき、そのまま MP4 に書き出せます。AI に動画を書かせることを前提に設計しています。
■ 何が変わったか
3D シーンが PixiJS の描画の中で直接描かれるようになりました。2 つ目のレンダラーも、キャンバスのコピーもありません。
そのため、2D レイヤーと同じことが 3D にもそのまま効きます。
・マスク(文字の形に 3D を切り抜く、など)
・フィルタ(ぼかし、ゆがみ…)
・合成モード(乗算、スクリーン…)
・キーフレーム(scene3d.knot.rotation.y のように 3D の中の物も動かせる)
・2D と 3D の重ね順(3D のおもちゃの間に 2D の文字を挟む、など)
■ 速くなりました
同じシーンを各エンジンで描いて比べました(素のメッシュ、three.js は WebGL レンダラー、同じマシン・headless Chrome、2026-11-13 測定)。
・立方体 2,000 個:1 フレーム 6.5ms → 3.2ms
・3D レイヤー 20 枚:15.9ms → 4.9ms
・3D レイヤーごとの WebGL コンテキスト:1 → 0(ブラウザの上限 約 16 を気にしなくてよくなりました)
■ できること
・glTF モデルを読み込み、アニメーションを時刻で再生(loadModel / fitCamera)
・映像側のカメラ・ライト・霧に 3D シーンを合わせる(followCamera / followLights / followFog)
・2D レイヤーを 3D の画面に貼る(スマホやモニターの画面に UI を映す)
・3D の物をクリックして、ドラッグで動かす(movie.pick / grab)
・3D の物の横に 2D のラベルを付けて追従させる(follow)
・書き出した動画が毎回同じ絵になるかを check コマンドが確かめる
■ 破壊的変更
three.js のレイヤー(pixi-effects/three)は廃止しました。3D は scene3d レイヤーで書きます。three.js 自体を使いたい場合は、canvas レイヤーの中で描けます(recipe three-in-canvas)。
npm i pixi-effects @pixi/3d
サイト:yjmtmtk.github.io/pixi-effec…
リリースノート:github.com/yjmtmtk/pixi-effe…
この動画も pixi-effects 0.52 と PixiJS 3D だけで作り、BGM もライブラリの機能(楽譜をテキストで書く)で鳴らしています。
AI is making games faster and cheaper to build, giving developers more room to take creative risks, revive overlooked genres, and let players reshape the worlds they play in.
In this conversation, a16z's Jon Lai and Anish Acharya join Elena Burger to discuss why human taste still determines what makes a game fun, how modding created some of gaming's biggest franchises, the opportunity for new distribution platforms, and games that adapt to whatever players choose to do.
0:00 Intro
0:57 How AI can now build a game
6:21 What shorter dev cycles unlock
10:08 Can AI make a game fun?
12:27 When creating is the game
14:01 Building a Habbo Hotel record shop
16:16 His son rebuilt Ghost of Tsushima in Minecraft
17:54 Mods: where billion-dollar franchises come from
20:36 Decompiling and reviving dead games
22:45 The renaissance of niche genres
25:51 Managing AI agents like an RTS game
28:02 The business of AI games
33:53 What excites and scares studios
35:16 The never-ending game
37:56 The games they'd finally buQild
YouTube: youtu.be/HDRy6XmE2RQ
@VirtualElena @illscience @Tocelot
wrote up some thoughts on ai x gaming with @illscience and @Tocelot.
read more below!
Another new demo on goodstuff 🕹️
@Doormat23 built this with PixiJS 3D, inspired by the Sonic the Hedgehog 2 special stages, using only Opus 5.5
Give it a go 👉 goodstuff.pixijs.io/mach-2/
Not bad from just a few prompts to Opus 5.5...
The visual effects AI can pull off are incredible compared to a few months ago, and it understands PixiJS 3D and WebGPU really well too 👀
Sign up to the closed beta and build something yourself!
pixijs.com/3d
PixiJS 3D closed beta signup is now open 🎉
2D and 3D in one place, the PixiJS way!
Fast by default.
WebGPU first with WebGL fallback.
Drop in 3D like it was always there.
Closed beta for the next couple of weeks, then open source!
Register 👇
pixijs.com/3d
dude GPT-6 Astra is some kind of turbo-AGI machine god for 3D games. It one-shot this in 45 minutes for hardly a couple % of my quota.
I figured out how to get great graphics out of it. The trick is image gen. I'll share the process below.
craigr retweeted
If anyone believes that AI coding + Astra's capabilities with Blender isn't going to radically impact game development they are a fool.
I have been using Codex + Unreal to develop a game by myself all summer, just to learn the capabilities. The state I've reached would have taken a team of 20 people six months. I have characters, maps, music, story- fully playable.
It's very clear that what a team of programmers will be replaced by one programmer, and a team of artists will be replaced by one artist.
You can pretend that this genie is going to go back into the bottle, but you would be wrong.
Yes, there are studios that will choose not to use AI. Good luck fundraising. You will either have to bootstrap or moonlight. Meanwhile the studios using it will ship.
craigr retweeted
Game development is the next frontier. GameDevBench is the first step, but I'm very confident progress will be measured in some way related to game dev
We're starting to leave the territory where you'd test an LLM by e.g. "create an svg of pelican on a bicycle". As one idea to generalize it, I was interested what Opus 5 would do if I gave it the first paragraph of the Lord of the Rings, a 1M token budget (~$10) and asked for three js render of it. Opus went off for ~2 hours and wrote 5500 lines of code that (procedurally) rendered the story. It's kind of janky but fun. But it's a bit mindboggling that the LLM has to place and orchestrate various polygon assets in (x,y,z) coordinates and write code that animates it all, and that it even does anything at all.
I also like this kind of examples because no one in their right mind would ever spend the time to write something this custom but LLMs have all the stamina and patience in the world, so it's an example where we go from "no one would ever do this" to "sure, why not, it's ~free". There might be a lot more. But I'm excited about creating hyper custom worlds that you can imagine dropping players into, e.g. here to participate in the LoTR story as a spectator NPC, or one of the characters, or etc. Something like an ephemeral GTA of X on demand.
Last thought is that the domain of worlds/games exposes a weakness in LLMs: they can't easily audit their work because they aren't able to efficiently and natively perceive videos or play games within them. Here, Opus 5 had to very slowly and painstakingly take screenshots at different points, and it messed up a few times and created a bunch of jank. An example of raw capability (multimodal, gameplay) that I think is still quite lacking.
craigr retweeted
Fable 5 is the new best model on GameDevBench.
At a staggering 67.3% task success rate, it outclasses every model except for GPT-5.6 Sol.
I think we all expected this result, but it's still shocking to see the numbers. Just half a year ago, models couldn't even break 50%!
Neural motion synthesis in Threejs/WebGPU,
even runs on my 7y old Iphone!
try it here: motionsynth.sweriko.com
sauce: github.com/sweriko/ai4anim-w…
PixiJS Skills are live!
Give your AI coding agent better context for WebGL/WebGPU builds (games, interactive apps, data viz, and more)
Perfect for #vibejam if you want to vibe code a game this weekend
Install:
npx skills add github.com/pixijs/pixijs-ski…
craigr retweeted
Hiya! I'm actively looking for my next role. I created a portfolio site to show off things I've made plus other bits of information about me.
sketchpunk.github.io/portfol…
If willing, please share this post. Big thanks to all who does! It's much appreciated.















