
Creative technologist on the AI frontier. Deep in ML and deep learning research, quietly building something new. Co-Founder @thewavexr. Advisor at Qoral
Joined March 2012
- Tweets1.3K
- Following1.3K
- Followers1.7K
- Likes1.1K
Pinned Tweet
Using Fable 5 to add fur and a new voice to my RL creature. The voice is based on the neural network activation (pictured top left). Made with @threejs @webgl_webgpu @runpod trained with JAX and MuJoCo from @GoogleDeepMind
Very impressed with Gemini 4 Argon's modeling, rigging and animating skills. All done headlessly in @Blender
via @antigravity based on some Moebius reference I assembled
@GeminiApp @Google
Aaron Lemke retweeted
Fully onboard with vibe manufacturing. I've been designing this little answering machine for my kids + fam. Had Astra source parts and design the enclosure around their specs and build this model site. 3d print + ESP32 + Astra is so fun.
mycelium.aaronglemke.workers…
Mycelial network compute shader sketch with Opus 5.5
The music also generated by Opus
Built with
@threejs @WebGL_WebGPU @claudeai
sisyphus.aaronlemke.com/
Karl Sims style evolved creatures.
This time they are trained to push a boulder up a massive mountain.
@threejs @webgl_webgpu
AAA cinematic depth of field with proper bokeh in three.js has been a white whale of mine for a while. I've thrown every new model release at it over the years.
Astra has gotten closer than any model before.
Still a few artifacts left but the model is working through them methodically.
@threejs @webgl_webgpu @OpenAI
Having fun teaching some RL agents to play soccer.
It's all about falling with style
@threejs @webgl_webgpu @runpod
Starting a new practice: treating the art I encounter in museums as prompts for making games. This first one was inspired by Wardha Shabbir’s beautiful work at the Venice Biennale. #artasgames
garden.aaronglemke.workers.d…
Aaron Lemke retweeted
Introducing Fluoddity.com!
Hundreds of thousands of particles self organize into a stunning variety of structures. Explore a world of possibilities with mutation and artificial selection!
This is my personal tool for exploring Fluoddity, rebuilt for the web by Opus 5
Inspired by this travel field notes trend I asked Claude to make my own version - picking a well documented day from my life and building a handmade scrapbook collage of that day.
‘The Last Day in Tokyo’
This is actually somewhat emotional and quite nostalgic to look at for me.
Built with
@claudeai
很喜欢这种把旅行照片收进田野笔记里的方式。
照片留住现场,旁边那枚略微缺墨的彩色橡皮章,只刻下最值得记住的轮廓。
树、穹顶、海岸都没有被讲满,反而像旅途中随手盖下的一页,安静,又很有地方感。
prompt:
请将我上传的每一张照片分别制作成一张独立的「橡皮章旅行田野笔记海报」,每张照片单独输出,不要多图拼贴。
整体采用4:3横版构图,将画面划分为左右两个区域,但不要绘制明显的分隔线。
左侧约占画面58%,忠实保留原始照片。准确保持主体身份、地形、建筑、植物、人物、空间关系、自然光影、真实质感和原有色彩氛围,仅进行克制的艺术出版物级摄影调色,并加入极轻微的细腻胶片颗粒。为适配版式可以自然裁切,但不得拉伸、扭曲、移动、替换或重新绘制主体。
右侧约占画面42%,使用温暖的米白色旧纸作为背景。纸张具有细微纤维、天然颗粒、轻微使用痕迹和哑光触感,同时保留大面积未经印刷的纸张留白,让空白成为版式的重要组成部分。
分析原始照片,从中提取最具地点辨识度的主体轮廓、建筑结构、地形走势、植物姿态、道路、水岸线或其他关键视觉关系,将其压缩成一枚小型多色橡皮章图像。
不要逐项复制照片中的全部内容。只保留能够让人一眼认出原始地点、主体和场景关系所必需的最少信息。删除人群、车辆、密集窗户、重复建筑、细碎植被、装饰构件和无关背景。
章印位于右侧纸张区域的中下部,整体只占右侧区域高度的约30%—38%,周围必须保留充足留白。章印不能放大成普通插画、完整风景画或品牌Logo。
根据原图构图决定章印的组织方式:
- 标志性建筑:保留最有辨识度的外轮廓、屋顶、穹顶、拱门、塔楼或主要结构。
- 山地聚落:将建筑压缩成沿地形排列的少量阶梯状色块。
- 海岸风景:保留山体走势、聚落层级、岸线和少量断续水纹。
- 城市远景:保留主要天际线、一个标志性建筑及一两层远山。
- 自然景观:保留主要山体、树木、水岸或道路的方向关系。
- 近景遮挡物:如果其对原图叙事重要,可作为前景章印轮廓保留。
从原始照片中提取2—4种专色油墨。优先使用炭黑、深绿、砖红、赭黄、灰蓝、灰褐等经过降低饱和度的颜色,但不得强行套用固定色盘。保留原照片最有辨识度的色彩性格,仅允许一个小面积颜色作为视觉强调。
每一种颜色都要呈现为分别手工盖印的效果:
真实橡皮章雕刻纹理、手工刻痕、粗细不均的排线、轮廓缺口、断裂边缘、干燥缺墨、纸张透底、颗粒状油墨、压力不均、局部重影,以及约1—2毫米的轻微套色偏移。
不同色层之间允许出现自然错位,边缘不能数字化平滑。印迹应像真正刻制后压在旧纸上的橡皮章,而不是套用滤镜的照片、平滑矢量插画或线稿Logo。
根据照片中的地点、主题和视觉意象生成文字:
地点英文名称
No. 编号
三个简短英文关键词
西元年份
文字放在章印下方或邻近留白处,使用小型、克制、略带机械误差的打字机字体。排版应像旅行者的田野记录,而不是广告标题。确保所有文字拼写准确,不添加无关标语、品牌或装饰性文案。
整体气质像建筑师、旅行作家或自然观察者保存的田野笔记:安静、克制、触感真实、地域明确、带有手工误差和收藏感。照片负责记录现场,章印负责留下记忆中最值得辨认的部分。
Avoid:明显的中央分隔线、圆形印章、中文红色印章、邮票齿孔、蜡封、贴纸拼贴、旅游纪念品模板、平滑矢量Logo、通用城市图标、完整复制所有建筑、密集刻画、儿童手工感、卡通风格、3D渲染、塑料质感、光滑数字渐变、过度饱和、过多文字、装饰堆积,以及重绘或改变左侧原始照片。
SOUND ON! I've always been inspired by Karl Sims's virtual creatures and decided to make my own. This uses a genetic algorithm with MAP Elites to encourage diverse morphologies. Of course I had to go ahead and give each one a unique synthesized voice based on their neural activation.
Built with @threejs and @webgl_webgpu along with MuJoCo Warp from @GoogleDeepMind and @nvidia - trained on a @runpod A100
I recently fed Sol a bunch of photos I took at a shell museum and asked for a parameterized 3d shell generator so I could explore the different types. After about an hour it gave me this.
Built with:
@threejs @webgl_webgpu @OpenAI
The sea urchin (cidaroid) looks especially convincing
Im pretty impressed with the shaders it created here.
Aaron Lemke retweeted
Robots with tentacles! 🐙
SpiRobs are soft robots inspired by nature, designed in the shape of a logarithmic spiral.
They mimic the flexible movements of octopus arms and elephant trunks to grasp objects. 🐘
SpiRobs use simple cables to move, making them easy to control and operate.
They can scale from small to large sizes, ranging from centimeters to meters.
The robots can adapt to different object shapes, improving their grasping ability. A mini version can act as a tiny gripper, while a large version can be mounted on drones.
Interestingly, multiple robots can work together to wrap around and hold large objects.
That's just out of this world! 😮💨
~~
♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
Aaron Lemke retweeted
One brain. For any robot. 🤖
We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
Also went down a rabbit hole of teaching a synthesizer to mimic animal sounds. Inspired by the VST plugin SynPlant 2
Sound on!
built with Fable 5. this involved building out a synth and a harness that could run entirely on the GPU to parallelize training search through evolutionary algorithms. trained on a @runpod RTX 5090
Built this at my first SF hackathon yesterday: a browser-based robot builder and RL trainer using @threejs and @webgl_webgpu, inspired by @mrdoob’s toy builder.
Design a robot, hit Train, and it gets sent to my GPU back in Austin to begin an RL curriculum—learning how to move from scratch.
Great write up for those looking to get in to Reinforcement Learning!
If you want to read more, I just finished a quick write-up on going from 0 to a moving monster: writing.lindaxue.com/rl-mon
Anddd if you want to mess around with it, the code is all here: github.com/linder0/rl-mon
Aaron Lemke retweeted
now an obsidian plugin - grow your notes as a living garden 🌺
🌿 every tag (or folder) grows into a plant: notes = leaves
⏳ scrub the timeline, or hit ▶ GROW to replay note growth sequence
✍️ alive while you write
🖱 click any leaf to open its note
🌗 two themes: dark observatory & light field notebook
available for my x subscribers

















