Homelab Journey Overview
This first post in the AI + Homelab Meta Series explains the workflow, tools, and messy partnership behind building both the Homelab and the writing process that documented it.
🧰 Tools + Workflow
- Obsidian for note development and version tracking
- Claude (MCP) for structural feedback and linking
- ChatGPT for fact-checking and revision
- WordPress + Grammarly for final polish
- Digital Garden for sharing drafts and rough cuts
💬 Meta Insight
Reflect on how AI helped conceptualize the “meta” layer — the director’s cut — that reframed this project as a study in co-authorship.
Final Posted Version
How I Wrote This With AI
A behind-the-scenes look at building and documenting a homelab, with a little help from my algorithmic co-authors.
When I started writing about my Homelab Project, I thought I was documenting a technical build. Hardware, containers, automation, and workflows. What I didn’t expect was that the writing itself would become an experiment in collaboration, cognition, and creativity, utilizing generative AI.
I’m planning on publishing a series of blog posts about building this homelab from the ground up. I’m leaving the notes, drafts, and outlines here in my digital garden.
But running parallel to that technical narrative is this “director’s cut.” A 4-part meta-series about how AI shaped not just the homelab, but the process of thinking, building, and writing about it.
This is the start of the story behind the story.
Tools + Workflow
I didn’t plan this as a multimodal collaboration, but that’s how it unfolded. Here’s how the pieces fit together:
| Tool | Role | Simple Explanation |
|---|---|---|
| Obsidian 🌳 | ChatGPT is my creative writing partner and brainstorming buddy. I use this generative AI to bounce ideas off of, refine a concept, and polish the final sentences, ensuring the writing still sounds exactly like me. | Obsidian is my digital workshop and memory bank. It’s a notetaking app that forms the backbone of my digital garden, where every idea starts as a connected note, like a detailed project binder. |
| Claude (via MCP) 🧵 | My structural editor and linker. Claude helps me identify patterns across notes, refine structure, and connect threads that I might overlook. | Claude is my structural editor and master librarian. It’s an AI model that, when connected to a system called MCP, can look at all my files. It helps me find patterns across notes and connect hidden threads, like a super-smart assistant that organizes your entire bookshelf. |
| ChatGPT ✍️ | My writing partner. I use it to brainstorm, refine ideas, and polish prose while maintaining my own tone. | ChatGPT is my creative writing partner and brainstorming buddy. I use this generative AI to bounce ideas off of, refine a concept, and polish the final sentences while making sure the writing still sounds exactly like me. |
| Grammarly + WordPress ✨ | The polish layer. Once a post leaves Obsidian, these tools catch clarity issues and ensure the writing is accessible to my audience. | Grammarly and WordPress are the final polish and delivery team. These tools ensure the post is crystal clear and easy for anyone to read. They catch any final errors and guarantee the writing is accessible before the post is published online. |
This workflow isn’t linear; it’s recursive, meaning a process that calls upon itself or is defined in terms of itself, and iterative, meaning a process that involves the repetition of a set of steps, resulting in a looping, layered process where tools talk to each other (and to me) in a kind of asynchronous dialogue.
For example, when drafting the hardware post, I had separate notes about server specs, power consumption, and noise levels scattered across Obsidian. Claude noticed I kept returning to themes of “tradeoffs” and “constraints” and suggested framing the entire post around decision-making under real-world limitations. That reframing became the backbone of the piece.
Why a Director’s Cut?
Most technical documentation hides the mess. You see the finished infrastructure, the clean diagrams, the working commands. It doesn’t show the false starts, the forgotten context, or the conversations that shaped the thinking.
By documenting how AI collaborated in this process, I’m exposing what usually stays invisible:
- How ideas evolve through dialogue (human-AI and human-human)
- How memory and context shape what we can build
- How troubleshooting a thought partner mirrors troubleshooting a system
- What it means to co-author knowledge with a machine
The main series asks: How do you build a homelab?
This meta-series asks: How do you think alongside AI while building one?
The idea of a “director’s cut” in my use of technology is not new or normal. In our early research on online reading comprehension, we conducted think-alouds, recording adolescents as they read online to better understand their patterns and practices. The key idea is that we have human beings discuss what they’re doing and why.
The Moment Everything Shifted
As I was building things in my home lab, I encountered a particularly challenging Proxmox networking issue. The basic idea was that the service I was trying to run wouldn’t connect to the Internet. I was trying to troubleshoot using ChatGPT, and it would get confused, asking me to do things that it had previously instructed me to do weeks earlier. I tried using Gemini, GitHub CoPilot, and other AI tools to solve the problem.
At some point, I’m getting into an argument with one AI model and copying/pasting what the other model told me. ChatGPT then suggested something unexpected:
“You’re documenting what you’re building. Have you considered documenting how you’re documenting?”
That one question reframed the entire project.
It turned this from a simple build log into a meta-level inquiry. A director’s cut of the creative and cognitive process. I started seeing each AI model not just as a tool, but as a collaborator with a distinct cognitive style:
- Claude sees structure, identifies patterns across disconnected notes, and suggests connections I miss
- ChatGPT iterates on ideas, helps me clarify muddy thinking, and catches when I’m overcomplicating
- I curate what matters, provide the context they lack, and decide what story needs telling
Together, we built not just content, but commentary. A running dialogue about human-AI collaboration in practice.
Why Show the Seams?
Writing with AI can make the process look effortless. Vibe coding videos make the process look like magic. It’s not.
In reality, there’s:
- Revision fatigue (when ChatGPT rephrases something five times but still misses the point)
- Prompt misfires (when I realize I explained my thinking poorly)
- Context drift (when Claude “forgets” what we discussed three messages ago)
- Philosophical questions (who’s the author when ideas emerge from dialogue?)
By publishing both the polished technical series and this raw meta-series, I aim to demonstrate that creative work with AI remains deeply human. Iterative, uncertain, and fundamentally about making choices.
The main series shows the homelab. This series shows the thinking.
What’s Next in the Director’s Cut
This meta-series runs parallel to the 10-part homelab build, with 4 posts exploring different facets of AI collaboration:
- 🧠 How I Wrote This With AI (you are here) — The workflow, the tools, and the decision to document the documentation
- 🧩 When the AI Forgets — What happens when your thought partner loses context, and how that mirrors the fragility of any system
- 🤝 Using AI as a Research Partner — Treating models as colleagues in inquiry, not just search engines with better prose
- 🔁 What I Learned Building (and Writing) With AI — Synthesis: lessons about thinking, building, and collaboration at the intersection of human and machine intelligence
Each meta post follows the same pattern as the homelab build itself:
design → test → fail → iterate → reflect
And each will publish between technical posts, creating a rhythm:
Technical post → Meta reflection → Technical post → Meta reflection
A Final Note on the Series Structure
As I write, I will try to provide as many details as possible about what ideas I had and what was created by generative AI. I’ll also try to document the specific tools and their affordances. Affordances are the characteristics or properties of an object that suggest how it can be used. So, ChatGPT is good at this, while Gemini is good at that. Alternatively, spoons are good at scooping peanut butter, while knives are better suited for spreading it.
As I write this series, both the main homelab series, I envision this structure:
- The main homelab series (10 parts) documents what I built and why.
- This director’s cut (4 parts) documents how I thought, struggled, and collaborated while building it.
You can read either series independently, but together they tell a fuller story. Building infrastructure is also building a way of thinking. And increasingly, that thinking happens in dialogue with AI.
I’ll document all of my notes in my digital garden and connect them in this overview.
first draft
🧠 How I Wrote This With AI
A behind-the-scenes look at building and documenting a homelab — with a little help from my algorithmic co-authors.
When I started writing about my Homelab Project, I thought I was documenting a technical build — hardware, containers, automation, and workflows. What I didn’t expect was that the writing itself would become an experiment in collaboration, cognition, and creativity with generative AI.
This post kicks off a short “making-of” series about how I built this homelab — and how I wrote about it with AI. It’s not just a process log; it’s a reflection on learning, co-authorship, and the strange productivity of thinking alongside machines.
🧰 Tools + Workflow
I didn’t plan this as a multi-model collaboration, but that’s how it unfolded. Here’s how the pieces fit together:
- Obsidian — the creative core. Every post starts as a note here. I use it to track drafts, revisions, metadata, and the messy thought trails behind each idea.
- Claude (via MCP) — my structural editor and linker. Claude helps me identify patterns across notes, refine structure, and connect threads that I might overlook.
- ChatGPT — my writing partner. I use it to brainstorm, check sources, and tighten prose while keeping my own tone intact.
- Grammarly + WordPress — the polish layer. Once a post leaves Obsidian, these tools catch clarity issues and ensure the writing is accessible to my audience.
- Digital Garden — my public drafting space. I publish the rough cuts here so others can see the work-in-progress — the thinking, revising, and rethinking in motion.
This workflow isn’t linear. It’s recursive — a looping, layered process where tools talk to each other (and to me) in a kind of asynchronous dialogue.
💬 The Meta Layer
Somewhere along the way, the AI I was using to help write about the homelab suggested that I also write about the writing itself.
That one idea reframed everything.
It turned this project from a simple documentation task into a meta-level inquiry — a “director’s cut” of the creative process. I started to see each model not just as a tool, but as a collaborator with a different cognitive style:
- Claude sees structure.
- ChatGPT iterates and clarifies.
- I curate, contextualize, and narrate.
Together, we built not just content, but commentary — a conversation about how humans and AI might co-author new kinds of knowledge.
🪞 Why Share the Rough Cuts?
Because writing with AI can make the process look seamless — but it’s anything but.
In reality, there’s revision fatigue, prompt misfires, missing context, and the occasional philosophical crisis about authorship. By publishing the rough drafts in my Digital Garden, I want to expose that process — not to demystify it entirely, but to show that creative work with AI is still human work: iterative, uncertain, and deeply personal.
🔗 What’s Next
This post begins a four-part reflective series on the AI side of the Homelab project:
- 🧠 How I Wrote This With AI — The workflow, the tools, and the messy process.
- 🧩 When the AI Forgets — What happens when your thought partner loses the thread.
- 🤝 Using AI as a Research Partner — How to treat models as collaborators in inquiry.
- 🔁 What I Learned Building (and Writing) With AI — Lessons at the intersection of building, thinking, and co-authoring.
Each post mirrors the same pattern I used to build the lab itself: design → test → fail → rebuild → reflect.
✍️ Process Notes (for the curious)
- Drafted and organized in Obsidian
- Structured and cross-linked with Claude (MCP)
- Revised and fact-checked with ChatGPT
- Final edits via Grammarly + WordPress
- Published iteratively in my Digital Garden
AI didn’t write this — we wrote it.
And that, more than the homelab itself, might be the real experiment here.
“Every technical project is also a literacy project — a way of learning to speak in new systems, with new voices.”