Engineering Notes

What I built, configured, debugged, and learned.

Practical notes from independent DevOps work, self-hosted infrastructure, application platforms, and local AI experiments.

A Windows development workstation connected through a private network to an Apple Silicon local LLM server

AI Infrastructure / 10 min read / Sep 2, 2026

Running Coding LLMs on a 16 GB Mac mini M4

An engineering log of running coding-focused language models on a 16 GB Apple Silicon machine using Ollama, MLX, llama.cpp, GGUF quantization, and remote inference.

A deployment flow from Docker build through K3s and a Kubernetes Service to an application

DevOps / Kubernetes / 8 min read / Sep 2, 2026

Deploying Next.js to K3s Without a Paid Container Registry

A lightweight workflow for building a Next.js Docker image, importing it into a self-hosted K3s node, and performing a rolling deployment without a paid registry.

A Proxmox virtual disk flowing through a partition and LVM layers to a constrained Linux root filesystem and K3s node

Kubernetes / Linux / 8 min read / Sep 2, 2026

Debugging K3s DiskPressure on a Proxmox VM

A troubleshooting story where pods stopped scheduling because the Linux root filesystem was full even though the Proxmox virtual disk had much more capacity available.

An AI coding agent connected through an MCP control layer to scoped PostgreSQL access

AI Infrastructure / Backend / 8 min read / Sep 2, 2026

Giving AI Coding Agents Controlled PostgreSQL Access with MCP

An experiment connecting AI development tools to PostgreSQL through MCP while keeping database privileges deliberately scoped.

A Headscale-controlled private network branching to Windows, macOS, K3s, Caddy, and AI infrastructure

Networking / 8 min read / Sep 2, 2026

Self-Hosting Headscale for a Private Developer Network

Building a private network connecting development machines, application infrastructure, K3s, and local AI services using Headscale and Tailscale clients.

An Apple Silicon inference server with three local LLM runtime paths leading to coding tools

AI Infrastructure / 8 min read / Sep 2, 2026

Ollama vs MLX vs llama.cpp on Apple Silicon

A practical comparison of three local LLM runtimes based on using them for coding workloads on a Mac mini M4.

A K3s application pod passing through controlled network and authentication layers to PostgreSQL

Kubernetes / Backend / 7 min read / Sep 2, 2026

Connecting K3s Pods to PostgreSQL Safely

Connecting applications inside K3s to PostgreSQL while handling pod networks, listening interfaces, authentication rules, SSL, and database roles deliberately.

A Windows laptop connected through a private Tailscale network to a Mac mini local LLM server

AI Infrastructure / Networking / 7 min read / Sep 2, 2026

Building a Private AI Coding Server with Tailscale

How I separated my development workstation from my AI inference machine and connected them through a private Tailnet.

A Flutter mobile application and Next.js administration dashboard connected to Supabase and PostgreSQL

Application Architecture / 7 min read / Sep 2, 2026

Building a Flutter App and Next.js Admin on One Supabase Backend

How a mobile application and administration dashboard can share one PostgreSQL-backed content platform while keeping their responsibilities separate.

Two structured Hadith data sources being compared through a REST API and provenance review

Data Engineering / 7 min read / Sep 2, 2026

When an API Is Correct but the Data Is Not

A data-quality investigation into why two Hadith APIs can return valid responses while disagreeing about Hadith numbers.