Wow this explains so much arxiv.org/abs/2603.07267
1
84
Just released a new open source library for profiling and agent-driven optimization of GPU kernels github.com/michaelalbada/War…
1
3
112
I’m delighted to share that I just received my first hard copy of Building Applications with AI Agents. The book will be available on October 21. Building agents is easier than you think, and this shows you how. Huge thanks to all of those who helped make this happen!
"Less than 0.1% of the code that humanity needs has been written. There are so many problems that we can solve, and these new tools will help us create the software that humanity needs." -Tim Bozarth
The pace of improvement for world models has been incredible. Great work by DeepMind deepmind.google/discover/blo…
4
264
Very exciting launch of GPT-OSS by OpenAI. It comes in two sizes, for both maximum reasoning capabilities & on-device, cheaper, faster option, all Apache 2.0. The price of intelligence keeps falling, this is a big step in the right direction 🔥
openai.com/index/introducing…
1
2
247
What an incredible time to be alive. The pace of technical innovation in so many domains these days is inspiring. designboom.com/technology/je…
1
3
242
OpenAI just launched the Agents SDK. We've known that they have been working on this behind the scenes, and it's great to finally get to see it. This is going to be a game changer for building agents.
openai.com/index/new-tools-f…
2
220
With appropriate model size and computational capacity, reinforcement learning is proving to be incredibly effective at challenging tasks. Most real-world tasks don’t come with rigorous evaluation sets like the IOI. Building rigorous evaluation sets is the next bottleneck.
207
These latest numbers from o3 represent that this is a big step forward. Very curious to see where this new path of training takes us substack.com/home/post/p-153…
137



