The Building Blocks of Agentic AI.. I'm super proud of the work our team has done over the past year - rethinking and reimagining our AI stack from scratch has not been easy but we wanted to focus on three things: 1. Being PyTorch Native - PyTorch after all is the open language of AI! 2. Building for scale - it take scale on all axes to achieve super intelligence.. 3. Hardware heterogeneity - it's increasingly important from cloud to device to support hardware from many different places. ExecuTorch 1.0:  End-to-end solution for on-device AI inference 👉 Blog: https://lnkd.in/gmTZKYzw Torchforge: PyTorch-native agentic RL library that lets you focus on algorithms—not infra 👉 Blog: https://lnkd.in/g3G4W_N6 Monarch: Distributed programming framework that reimagines cluster-scale execution and orchestration with a focus on developer simplicity 👉 Blog: https://lnkd.in/gWF4ERNp Helion: PyTorch native DSL with autotuning that dramatically reduces the amount of code developers need to write to build advanced kernels for heterogeneous hardware. 👉 Blog: https://lnkd.in/g3_GsxU7 Torchcomms:  Unified API and PyTorch-native library designed to provide robust, fault-tolerant distributed communication at near-infinite scale. 👉 Blog: https://lnkd.in/gD7yz_RU It's super early and a lot of the projects are at an early stage - but it's exciting that we can share all of this in the open with you all and collaborate. We look forward to building this stack together with you!! P.S. For the folks at the AI infra workshop yesterday that made some wishes - I hope they've now come true.. Learn more at meta-pytorch.org.. Cheers!