add vdcore open repo

Author: jiange91

content/posts/vdcores.md

@@ -107,15 +107,15 @@ Recall that VDCores aim to achieve pipelining without programmers explictly defi
 
 ## Decoupled Cores: In Live Action and in the Wild
 
-> We are working to bring VDCores to the open-source community and to a wider range of cores and hardware platforms. Stay tuned!
+> VDCores is now open source! Check out the repository at [github.com/vdcores/vdcores](https://github.com/vdcores/vdcores) and give it a spin. We are continuing to expand support to a wider range of cores and hardware platforms. Stay tuned!
 
 VDCores's decoupled model goes beyond cleaner way to write one kernel, it is {{< highlight-text >}}**a substrate for systematic overlap and composition**{{< /highlight-text >}}. Once memory, compute, and control communicate only through explicit dependencies, we can schedule and explore pipelines at a higher level and let the runtime safely exploit concurrency and performance. Here's exciting directions we are exploring and may reveal the full potential of VDCores system:
 
 - **Communication Virtual Cores:** VDCores are all about isolating and decomposing work into separate, decoupled components, and communication/networking should be one of them. Whether it is inter-rack communication over InfiniBand or communication between GPUs, it composes naturally with existing memory and compute decoupled cores.
 - **Adapting to Tiered Memory and new Arch:** By decoupling issue/completion and separating memory/compute concerns, the same kernel structure can adapt to evolving GPU mechanisms (e.g., new [async memory](https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/#streaming_multiprocessors_compute_engines_for_the_ai_factory) types on Blackwell), and wider range of memory locations and operations (e.g., ) without changing the VDCores application.
 - **Composition at Runtime:** VDCores is designed to be a runtime substrate that makes it easier to compose kernels into larger pipelines, coordinate resource allocation, and reason about end-to-end overlap beyond traditional kernel boundaries. Given the power to compose any memory/compute operations dynamically at runtime, fine-grained, runtime-aware policies could be further explored.
 
-We’ll cover these topics in future posts in this series. Before that, we’re excited to bring the runtime to the public and let you all give it a spin very soon. Stay tuned!
+We’ll cover these topics in future posts in this series. In the meantime, VDCores is now open source — head over to [github.com/vdcores/vdcores](https://github.com/vdcores/vdcores) to explore the code, try it out, and contribute. Stay tuned!
 
 
 <hr style="border-top: 1px solid var(--text-color, #ccc); opacity: 0.3; margin: 2rem 0;">