anandmaj (@Almondgodd) / X

anandmaj

450 posts

anandmaj

@Almondgodd

@penn prev ai @tesla_optimus @dynarobotics

Joined February 2019

Pinned
anandmaj
@Almondgodd
Sep 25, 2025
I spent the past month reimplementing DeepMind’s Genie 3 world model from scratch Ended up making TinyWorlds, a 3M parameter world model capable of generating playable game environments demo below + everything I learned in thread (full repo at the end)👇🏼
00:00
218K
anandmaj
@Almondgodd
Nov 13, 2025
I spent the past year building AI for robots at Tesla Optimus and Dyna Now I'm introducing ANDROID DREAMS: an essay of my predictions for the next 20 years of robotics, inspired by Situational Awareness and AI2027. I predict EGI by 2031 and more robots than humans by 2045👇🏼
00:00
344K
anandmaj
@Almondgodd
Nov 14, 2024
I trained a 2-arm robot using my Nintendo Switch (+ an action transformer) how I did it + my most interesting takeaways below👇🏼
00:00
136K
anandmaj
@Almondgodd
Aug 30, 2025
Over the past 3 months I’ve been interning @Tesla_Optimus to build AGI for the real world. Robotics is the hardest frontier of AI, but it gives us a clear path to eliminating scarcity. After working on Optimus, I’ve never been more confident that universal abundance is within
25K
anandmaj
@Almondgodd
Apr 3, 2025
I trained a humanoid robot to play famous anime openings on the piano with reinforcement learning (Attack on Titan, Evangelion, SAO, and more) how I did it + surprising discoveries in thread🧵
00:00
25K
anandmaj
@Almondgodd
Sep 25, 2025
Replying to @Almondgodd
9/ Finally, here’s tinyworlds: github.com/AlmondGod/tiny… It’s a minimal codebase to help people understand world modeling. Try it out yourself and make a PR, there are many easy + impactful additions to make (such as Mixture of Experts, Muon, and scaling). Thank you to @runpod
GitHub - AlmondGod/tinyworlds: A minimal implementation of DeepMind's Genie world model
From github.com
9.3K
anandmaj
@Almondgodd
Apr 18, 2025
Replying to @leothecurious
Makes you appreciate how cracked the brain is: we're simultaneously running inference and updating our weights at all times
5.7K
anandmaj
@Almondgodd
Nov 13, 2025
Replying to @Almondgodd
Read the full version below: android-dreams.ai I predict how embodied general intelligence (EGI) will be created, and on the way, how the US will out-manufacture China, endure political upheaval, and produce more robots than humans. (While I used to work at Dyna and Tesla
24K
anandmaj
@Almondgodd
Sep 25, 2025
Replying to @Almondgodd
1/ understanding world models World models are neural networks that simulate physical worlds by generating videos. DeepMind’s Genie 3 proved that, just like LLMs, scaled-up world models exhibit emergent behavior: > Controllability: Pressing the right arrow makes the camera pan
12K
anandmaj
@Almondgodd
Sep 25, 2025
Replying to @Almondgodd
3/ building the space-time transformer Normal transformers in LLMs understand language, which is 1D. TinyWorlds requires a model that understands video, which is 3D (height, width, time). This model also has to train quickly and learn using both actions and video. Space-time
8.5K
anandmaj
@Almondgodd
Nov 14, 2024
Replying to @Almondgodd
Finally, here’s a run of the autonomous ACT policy (5x speed). Thanks for making it to the end! If you're interested, see this longer blog about the project: rain-argon-1fc.notion.site/ALOHA-Bigym-Jo… and the git repo: github.com/AlmondGod/aloh… Thank you to @JasonMa2020 for inspiring this project!
00:00
3.5K
anandmaj
@Almondgodd
Sep 25, 2025
Replying to @Almondgodd
2/ building the dataset of worlds Before training TinyWorlds, I decided what video game worlds my model should generate by building the dataset. The set of worlds the model sees in training determines what worlds it generates. I created TinyWorlds' dataset by processing
9.7K
anandmaj
@Almondgodd
Nov 13, 2025
Replying to @Almondgodd
Read the full version below: android-dreams.ai Inspired by Daniel Kokatajlo et al, AI 2027 and Leopold Aschenbrenner, Situational Awareness Thank you to Adam Majmudar, Ahalya Nava, and Sourish Jasti for their feedback
9.7K
anandmaj
@Almondgodd
Nov 13, 2025
Replying to @Almondgodd
Superintelligence Era, 2045+ AI continuously improves by learning from interactions with the real world. Humans still work in domains where humanness matters, like healthcare, education, resource allocation, and the arts. Humans live more diverse lives. Many new ideologies
11K