AI practice
I have been running agentic systems in production daily since February 2026. Here is what that means in practice, and what I think it changes about the job.
I build with it daily, and I run what I build
The system I run every day, the tools I have published, and how the job changes.
The system I run
Frankenverse: A Self-Hosted Agentic System
A production agentic system I built and run every day. My own knowledge base wired to Claude through a self-hosted MCP server, with custom skills and workflows that do the everyday work and run unattended on a schedule.
- Self-hosted an MCP (Model Context Protocol) server on my own VPS - encrypted at rest, reachable remotely over HTTPS and WebDAV - so agents can read and write a live knowledge base from anywhere.
- Built a library of custom Claude Code skills and sub-agents that run autonomously: routing inputs, holding state across sessions, and taking real actions (scheduling, reminders, external API calls) rather than one-shot prompting.
- Owned the agent architecture end-to-end - orchestration, tool use, guardrails, model selection, and change control on the agent definitions themselves - the same production concerns an AI-product team faces.
- I operate it daily as a real dependency, not a demo. The difference between having used agents and having lived with them in production.
How do you actually work with it?
Claude Code every day, with Cursor alongside it. I use it to argue out a design before any code exists, to document a codebase I have never seen, to stand up an MVP (see the "Create MVP" tool below), and to write and refactor the code itself.
What it has not changed is who is accountable for the result. I still have to know whether the answer is right.
Will it replace developers?
No, but it moves the job. Less of the week goes on writing the code and more of it goes on deciding what should exist and directing the work.
That shift is the part I care about, and it is a people problem before it is a tooling one. Coaching was my first profession, so getting a team to actually take up a new way of working is the thing I have most practice at. I do it first myself, then bring people with me.
Tools I have published
AI Skill: Create MVP
Productized my consulting methodology as a Claude Code skill - discovery → planning → implementation, resumable across sessions.
Lovable Output Analyser
A web tool created with Lovable that allows users to analyse a GitHub repository's codebase and rank it for maintainability and complexity.
The full record, one page at a time
Dive deeper into my work, my skills, my experience, and my certifications.
Work
Every project with the technologies, the constraints and what it produced.
Built and operatedAI practice
The agentic tooling I run daily, published tools I've authored, and what I actually use them for.
9 roles · 2008 → PresentExperience
The full timeline, every role, and the achievements attached to each of them.
TemperamentAbout me
Who I am, how I behave under pressure, what I'm still learning, and what the personality tests make of me.
16 certificationsCredentials
What I have, what I don't, and every credential linked back to the issuer.
PrintableOne-pager
The scannable printable version, for a fast skim.