About
Margrop Blog
AI Infrastructure Engineer
This site documents my long-term work across AI agents, cloud platforms, self-hosting, and engineering practice.
I care about whether a system can actually run, remain stable, and be diagnosed and repaired when it fails. The posts here therefore come from real deployments, migrations, incidents, automation work, and postmortems—not only from polished tutorials.
What I work on
- AI Agent: coding agents, MCP, shared memory, model gateways, and agent workflows
- Self-hosting: Docker, NAS, Proxmox VE, homelab systems, and networking
- Engineering: Linux, Java, Go, databases, performance, observability, and incidents
- Cloud Platform: deployment, automation, permissions, token management, and production practice
These areas are connected. My focus is putting AI capabilities into real engineering systems with data, permissions, monitoring, failure boundaries, and rollback paths.
How I write
I try to keep the parts that make an engineering note reusable:
- Context — where the problem appeared and what the original goal was.
- Evidence — what was tested, what was observed, and which attempts failed.
- Root cause and fix — why it happened and what changed.
- Verification — how the recovery or deployment was proven complete.
Sensitive hostnames, addresses, credentials, and internal details are redacted when needed.
Where to start
- New here? Start with Start Here.
- Browse the archive for recent field notes.
- Search by technology through tags.
- Find code and automation on GitHub.
- For daily AI news and agent work logs, visit AI Daily.
Other links
Follow on WeChat
Scan the QR code below to follow ClawLoader for updates about AI agents, self-hosting, and engineering practice.
Chinese is the primary language of this blog. I add English versions to posts that are especially useful as long-term references.