Sparse images: why your 14GB image is really 1GB
Yocto ext4 images can be tens of GB yet fail to upload because of a 2GiB per-file limit — because most of the file is holes. How to detect sparse files, compress them, and ship them the right way.
AI 网关实战
5 posts
sstate vs SDK vs eSDK: the three most-confused things in Yocto
sstate is a cache for the build machine, SDK is a toolchain for developers, eSDK packs both for offline system development. Here's how they differ and which one you want.
Why a Jetson Image Build Silently Compiles Rust and LLVM
A build kept stalling on llvm-native and rust-native. Tracing reverse dependencies with bitbake -g led to Tegra's OP-TEE / EKS boot chain needing python3-cryptography — which is written in Rust.
Edge AI Gateway Architecture: OpenWrt + Jetson, Each in Its Lane
Why one router plus one Jetson is the most pragmatic edge AI gateway: OpenWrt owns the forwarding plane, Tegra owns inference, wired together with VLANs and containers.
Containerized TensorRT on Jetson Orin: From Cross-Compile to Flashing
Run TensorRT inference in JetPack containers on NVIDIA Jetson Orin and deploy it as an edge AI gateway, including jetson-flash essentials.
Need something built?
Custom development and consulting for embedded, edge AI and agents.