OxyGen treats the KV cache as one shared resource across tasks and control frames, letting Mixture-of-Transformers VLAs keep full action frequency while generating language concurrently, up to 3.7× faster on an RTX 4090 and Jetson AGX Thor.
How quantization, sampler tuning, kernel work, and rollout topology accelerate NVIDIA's 16B Cosmos 3 robot policies on 24 GB GPUs — from single-request serving to full 1,200-episode evaluations.
Vec-LUT turns repetitive scalar table lookups into contiguous vector reads, accelerating parallel ternary LLM inference on x86 and ARM CPUs by up to 4.2×.
An analysis of GPTQv2 format limitations in vLLM and the CUDA kernel changes needed for low-bit asymmetric quantization inference.
Recent VLAs are moving from discrete to continuous control and from single-system to dual-system designs.
Notes on Dario Amodei's essays about powerful AI, interpretability, and AI policy.
A practical checklist for setting up Android smartphones for development.
A practical checklist for setting up Termux as an Android development environment.
A practical setup checklist for Raspberry Pi and similar single-board computers.
A look at the practicalities of deploying large language models on edge devices.