Blog ✍️

Notes from building AI systems and the infrastructure they run on.

2026

September

• Harness Engineering 101: How Coding Agents Actually Work

  Everything around the model that turns an LLM into a coding agent: the loop, tools, context, guardrails and checks. Explained with code, diagrams and data.

  AIarticleai-agentsharness-engineeringcontext-engineeringclaude-codecodexcoding-agentslinux
• Nine ways an AI agent fails in production, and how to catch them

  The failures that matter are not the funny screenshots. Six of the nine ways a support agent breaks look exactly like an agent working normally. Here is each one and the check for it.

  AIarticleai-agentsragprompt-injectionevalsobservabilitycustomer-support
• The cloud stack has seven levels now, it's not just IaaS, PaaS, SaaS

  Every course still teaches three boxes. The market has seven levels for compute, a second stack for data, and a category that only exists because AI agents write code that has to run somewhere.

  Cloudarticleserverlesspostgreskubernetesai-agentsvercelcloud-runneon

August

• Set up Prometheus, Node Exporter, and Grafana on Debian or Ubuntu

  A step by step guide to a working monitoring stack on a single Debian or Ubuntu server: Prometheus for storage, Node Exporter for host metrics, Grafana for dashboards, all running under systemd.

  DevOpsguideprometheusgrafananode-exportermonitoringsystemd
• Self-hosting WordPress with Docker: Apache or PHP-FPM, plus a Redis object cache

  A production WordPress stack in Docker Compose: pinned images, a database and Redis that never publish a port, an object cache with the drop-in that actually makes it work, and both the Apache and PHP-FPM routes behind nginx.

  Self Hostingguidewordpressdockernginxredismariadbphp