
ai experiments • devrel @nvidia • ex msft/amzn/groq • thoughts are my own
LA
Joined August 2020
- Tweets1.7K
- Following604
- Followers1.2K
- Likes6.6K
Pinned Tweet
I created the perfect Cart Surfer AI.
It's a 223K parameter convolutional neural network that can play Cart Surfer from Club Penguin. It achieves the theoretical maximum score while only looking at the pixels on the screen.
Making small game AI models is fun!
2
97
does anyone have recommendations for a good cloud gpu provider?
i need gpus like 5090s and larger for a few hours at a time, sometimes 8+ hours for training
been using @runpod but it's usually out of stock of what i want
ideally it works well with agents via cli/mcp. on demand, quick to spin up and set everything up with 50GB+ of training data, then tear down when done
3
2
205
Introducing NVIDIA Nemotron 3.5 Lightning⚡
An open 30B MoE model with 3B active parameters, built for always-on agents to complete high-volume, specialized tasks faster.
It delivers up to 4x the output speed of similar-sized models.
Ben Ankiel retweeted
Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.
During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.
That’s why we created the Open Secure AI Alliance.
AI security advances when the industry builds in the open, together.
We're introducing the Open Secure AI Alliance with industry leaders to develop new techniques and tools to safeguard software and agents.
By sharing models, tooling and research in the open, we can broaden the community of defenders.
Learn more about the founding members’ contributions: nvda.ws/4pD8Fc5
Ben Ankiel retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
hey, thanks for this @ashleybchae
i spent the next 30 days trying the 10-step Korean skincare routine because apparently i’m very easy to influence
I really like these kinds of charts they did. Very pleasing.
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
thinkingmachines.ai/news/int…
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
2
164
this is the longest goal I've put codex on so far
GPT-5.6 Sol is currently reverse engineering the Catchin' Waves minigame from Club Penguin into pygame so that I can train neural nets to get a perfect score
we are really living in the future
day 3 of making the perfect cart surfer ai
i implemented double dqn and retrained from scratch. needed a lot less training this time
it almost has the optimal scoring strat! alternates between backflip and spin because doing the same move twice in a row gives half points
1
173
next midjourney hardware project confirmed
163
day 3 of making the perfect cart surfer ai
i implemented double dqn and retrained from scratch. needed a lot less training this time
it almost has the optimal scoring strat! alternates between backflip and spin because doing the same move twice in a row gives half points
new high score!
did some training specifically focused on scoring and it's getting so much better! i never taught it to do the backflip, but it learned that it's the best move to get points
going to do a big training run to see how far i can push it
trained some more and now the neural network never loses lives AND completes the minigame every time
next step is to train it to a bunch of of tricks to get maximum points
neural networks are so cool
i trained one that can now beat cart surfer every time!
it's not good yet and dies sometimes...but it can do it! kind of crazy to watch it in action after just looking at training stats
started training my first deep reinforcement learning network on this yesterday!
it can't beat the game and get a good score yet...but it will!
so far there's been a few runs where it's beaten the teacher policy in score but dies during the run
















