LightSeek Foundation

@lightseekorg

LightSeek is the creator of TokenSpeed, TorchSpec, and SMG — building next-generation AI infra systems tokenspeed.app

San Francisco Bay Area

lightseek.org/blog

Replies Reposts Media Articles

Pinned
LightSeek Foundation
@lightseekorg
Aug 7
Over the past two years, inference providers competed on engine performance. I think that era is ending.
Article
LLM Inference: Past, Now, and Future — Where Does the Value Move?
LLM Inference — Past, Now, and Future: Where Does the Value Move? For the past two years, LLM inference has been one of the most contested layers in AI infrastructure. Dozens of inference providers,...
16
LightSeek Foundation
@lightseekorg
Sep 8
TokenSpeed × Kimi K3, Part I: TP8 optimization on GB300. Now stable 3+ weeks in production across ~1,000 B300 and ~2,000 H200 GPUs. Thanks @NVIDIAAI for engineering support and GB300 NVL72 access, and community teams for their support and contributions. 👇 lightseek.org/blog/kimi-k3-o…
1
LightSeek Foundation
@lightseekorg
Sep 3
Grateful for the collaboration and support from @PyTorch, @vllm_project, and the whole community. Let's keep building a stronger PyTorch ecosystem — together. Come build with us 🚀
PyTorch
@PyTorch
Sep 3
TorchSpec is a PyTorch-native framework for training speculative decoding draft models. This release, a collaboration with @vllm_project, shows it in action on Kimi K3 @lightseekorg
1
LightSeek Foundation
@lightseekorg
Sep 3
Releasing the @Kimi_Moonshot K3 Draft Collection — 3 draft models (EAGLE-3, DFlash2, DSpark) trained with TorchSpec and @vllm_project on @NVIDIAAI GB200. 🚀 We also shared the data recipes. More on the blog → lightseek.org/blog/kimi-k3-d…
2
LightSeek Foundation
@lightseekorg
Aug 31
Adoption ≠ Contribution: Some Thoughts on Open-Source Inference Engines A few thoughts on how open-source LLM inference engines get adopted. Users usually fall into a few groups: model providers, neo labs, inference clouds, and neo clouds. There’s some overlap between them.