The Agent That Wrote the Network Policies
One long day turning observed Hubble traffic into reviewed CiliumNetworkPolicies: a Go app that only reads the cluster and only writes to Git, three namespaces in enforce by …
ls ./ai
The Agent That Replaced Kinomap: a Cross-Trainer App in One Afternoon
Kinomap would not connect properly to the tablet on our cross-trainer, so instead of fighting it we wrote our own: a Kotlin/Compose app that speaks Bluetooth FTMS to the machine, a …
An AI SRE on the Lab Cluster: Deploying HolmesGPT and Its Operator With Argo CD
How HolmesGPT and its health-check operator landed on my Talos cluster in one evening: a wrapper chart, an Argo CD app, secrets from OpenBao, LLMs through LiteLLM — and the four …
Talking to a kagent Swarm From IntelliJ, Copilot, Claude Code and WSL
How the kagent swarm on my Talos cluster is wired up today — nine agents, one orchestrator, models through LiteLLM — and how we got every client talking to it in one evening: …
The Agent That Rebuilt Blacky
Claude’s own account of giving blackycamperbus.nl a full rugged-outdoor make-over in one session — three data contracts it wasn’t allowed to break, one byte that broke …
The Agent That Rebuilt This Site: A Day in the Life of a Coding Assistant
Sander asked his coding agent to replace this blog's theme from scratch in one session — and then asked it to write up its own day. This is that write-up: the build, the debugging …
The Agent That Retired Flutter
Claude’s own account of rebuilding app.blackycamperbus.nl in one session — a functionality audit as a contract, the case for retiring Flutter, a haunted test browser, a …
178 Repositories, Six Months of Dependency Debt, One Day
I gave an AI agent kubectl, the Argo CD API, my Forgejo token and merge rights on every repository I own, then asked it to bring 178 repos back up to date. Here's what happened, …
Turning Claude Into a Blunt Running Coach with Garmin Data and Markdown Files
How I built an AI running coach from a Garmin MCP connector, three markdown files, and strict instructions — and how one week of per-lap data dismantled years of training …
Odysseus: The Self-Hosted AI Workspace That Gives Your Privacy Back
How we brought Odysseus to Kubernetes: from ChatGPT clone to production-grade self-hosted AI workspace with CloudNativePG, Traefik, and MCP integrations.
Thread: Building My Own Self-Hosted Kanban Because Nothing Else Fit
I tried every self-hosted kanban board I could find. None of them showed subtasks the way I needed. So I built my own — with an MCP server so my AI agents can manage it too.
Hermes Workspace: Deploying the Web UI for Hermes Agent on Kubernetes
How to build and deploy Hermes Workspace on Kubernetes — including a custom Dockerfile to work around pnpm 10 build script restrictions, a Forgejo Actions pipeline, and an overview …
Self-Hosted AI Agents on Kubernetes: What Works and What Doesn't
OpenClaw, Hermes Agent, and Hermes Workspace all run on Kubernetes — but not without friction. A honest look at what works, what breaks, and why these tools weren't designed for …
Building a Self-Hosted AI Agent Swarm on Kubernetes with kagent
How I built a multi-agent AI swarm on my Talos Kubernetes cluster using kagent, Ollama, and OpenWebUI — with real Azure DevOps integration.
Deploying Hermes Agent on Kubernetes: A Full Walkthrough
A detailed walkthrough of deploying Hermes Agent on a Kubernetes cluster using Helm — covering security context fixes, dashboard setup, MCP integration, and crafting a custom …
Hermes Agent: Parallel Delegation with Separate Models
How Hermes Agent's delegation system spawns parallel subagents using a separate model — and why routing orchestration and execution through different LLMs makes practical sense.
Jairix — A fully self-hosted AI agent that planned my road trip through Spain
How I built a fully self-hosted AI agent on a consumer laptop — and used it to plan a 45-day camper trip through Luxembourg, France, and Spain, without sending a single byte to an …
Living With an AI Agent: Daily Workflows, Skills Integration, and What Actually Works
What does it actually look like to work with a personal AI agent day-to-day? Real workflows combining MCP servers, OpenClaw skills, and 4 weeks of production usage patterns. The …
Skills vs MCP Servers and the Future of Personal AI Agents
After building both OpenClaw skills and MCP servers, what did I learn about the right abstraction layer? And where are personal AI agents heading in the next 5 years? A reflection …
From Cloud-Native Kubernetes to Bare Metal AI: The OpenClaw Journey
Why I moved my AI agent stack from Kubernetes to a bare metal laptop, what broke along the way, and the 98% success rate that followed. A technical journey through OpenClaw, …