rsync Two Large Directories: Resume Interrupted Copies and Verify Them
Don’t stuff two trees into one rsync. Dry-run first, copy serially with partial files and logs, then verify. A 100% progress bar is not permission to wipe the source disk.
OpenAI Terraform Provider 1.0: Manage the API Platform as Code
The official provider talks to the Administration API, not model inference. Get the Admin key right, import existing resources, and treat terraform plan as the review gate.
Codex on Windows: Native PowerShell vs WSL, Sandbox, and Path Failures
Most Codex failures on Windows are environment, not the model. Pick native PowerShell or WSL, then keep Node, Git, paths, and the sandbox on the same side.
I Found an Open-Source Project That “Sees” People with Wi-Fi—Then Decided Not to Install It at Home
RuView claims Wi-Fi CSI alone can do presence, respiration, heart rate, even pose. I pulled the code and ran it: demos work, but it’s not ready to trust in my house.
Product Ready to Ship—Who Makes the Launch Video? I Tried Huashu Design
I ran my own product through Huashu Design—from brand analysis and visual directions to a 15-second promo. It turns Coding Agents into a design SOP that ships editable, exportable launch animation.
I Turned Word, PDF, and Ebooks into Custom-Voice Audio
I chained Pandoc, Audiblez, and RVC into a local pipeline: ordinary documents become EPUB, then spoken audio in a target voice. The stack runs—but how far is it from a real product?
I Turned Chinese Article Illustration into a Reusable Agent Skill
Illustrating an article isn’t hard because of pixels—it’s hard because of selection, format, and character drift. I packed that workflow into an open-source skill: Zuohai IP Studio.
Building Agent Memory That Survives the Next Session
A context window isn’t memory. I use a three-layer Markdown setup, write-time checks, and regular archival so agents can pick up where yesterday left off.
Guardrails are more than swear filters. Prompt injection, poisoned knowledge, risky tools, and compute exhaustion—walk the stack so you know what to defend.
Clear tasks beat fancy adjectives. Conditionals, constraints, and examples lock the output—plus a few patterns I’ve used in automation and local models.