I’m a software architect with many years of hands-on experience in the software industry. These days, I find a lot of joy in building and open-sourcing AI coding workflows and agent skills.
I share what I learn from building and using them, and I’d love to hear your thoughts too.
Announcements
- When AI Edits Your Words, Keep Your Meaning
Keep your meaning and voice when asking AI to edit emails and announcements. Everyday WQE examples, practical requests, and what still needs a human read.
- Putting SessionCue on Hold, and Trying iTerm2’s Claude Code Integration
After pausing SessionCue, a menu bar app for tracking agent sessions, I tried iTerm2’s Claude Code integration and reconsidered what was still worth building and maintaining myself.
- How I Rebuilt svg-infographic So Broken Diagrams Don't Count as Success
How I rebuilt svg-infographic around standalone SVGs, embedded Korean fonts, and receipts that bind inputs, runtime, and artifacts so defective output is not quietly counted as success.
- Why Relationships and Spacing Come Before Drawing a Diagram
The design principles behind svg-infographic: not just producing a plausible technical diagram once, but keeping an editable SVG and a verified PNG that can be revised and reused.
- What to Design Before Pressing the Release Button
The safety design of github-release-guide: check a repository for credentials, personal-environment traces, and internal information before publication, then treat a release as a sequence of decisions, approvals, changes, and verification.
- Clearer Writing, the Same Meaning
How WQE preserves meaning, voice, and the answers readers need, why new Korean drafts use a separate path, and where observed results still fall short.
- What I Noticed Too Late While Making Two Agents Collaborate
What I learned while running two agents as Driver and Reviewer — widening the Reviewer's role from pointing out problems to falsifying independently, then solving collaboratively.