Miles, Slime, and Ray: Orchestrating the Post-Training Loop
A practical architecture guide to the post-training loop, Ray as the control plane, and engineering weight synchronization.
Read More
Hands on DSH
A technical guide to DeepSeek Harness architecture, implementation, deployment, usage, and its differences from OpenClaw.
Read MoreDive into the Agent Loop
A practical breakdown of the agent loop: the roles of models, MCP servers, and clients, plus Skills, multi-agent systems, and context isolation.
Read MoreHow uv Works: Python Environments, Dependency Resolution, and Caching
A practical explanation of uv's internal workflow, from Python selection and PubGrub resolution to lockfiles, caching, virtual environments, and uv run.
Read MoreCross-Attention in Multimodal AI: When One Stream Needs to Read Another
A practical guide to cross-attention, from Q/K/V to Stable Diffusion’s text-conditioned U-Net and cross-modal dimension design.
Read More