Start Here
A guided path through zemna.net's core software architecture notes on AI agents, automation proof, tool choices, and maintainable systems.
zemna.net is not a news feed. It is a working notebook for software systems that need to survive real maintenance.
If you are new here, read these paths in order.
1. Agent boundaries
Start with the contract before the agent touches a repository.
- The Agent Edit Contract I Use Before a Coding Agent Touches a Repo
- Your Coding Agent Needs a Map, Not a Bigger Context Window
The core idea: context is not the same as permission. A useful agent needs a map, boundaries, verification, and rollback.
2. Automation proof
Automation is only useful when the output can be trusted.
- Your Cron Job Is Not Healthy Until the Artifact Proves It
- Why AI Cron Jobs Lie to You: The Exit 0 With Empty Output Pattern
- Zero-Cost Observability for Agent Crons
The core idea: process status is not proof of work. Check fresh artifacts, useful output, and downstream delivery.
3. Tool judgment
Tools are useful until they become load-bearing without an exit path.
- Before You Adopt a Beta Library, Prove the Exit Path
- I Swapped My LLM Backend: The API Call Worked on the First Try
- Open Weights Just Ate the API Margin
The core idea: the best tool is the one you can verify, afford, and replace when the assumptions change.
4. Frontend systems
Generated UI still needs engineering discipline.
- Code That Renders Is Not Code You Can Trust
- The JavaScript I Deleted With CSS: A 2026 Survival Guide
- Vue 3.6 Vapor Mode: No Virtual DOM, No Rewrite
The core idea: correctness includes behavior, accessibility, performance, and future edits.
5. Regional context
I write from Indonesia as a Korean programmer and software architect. That context matters because tooling, pricing, hiring, infrastructure, and business constraints are not identical everywhere.
- The Indonesian developer scene is having its Linux moment
- The quiet consolidation in Korean dev tooling