I want to be honest about something. For years, I hated maintenance.
Not the idea of it. The reality of it. Logging into a CMS, navigating a dashboard that had not changed since 2014, clicking through menus to find the one page that needed a date updated or a link fixed. Doing that across hundreds of pages. Doing it again next month. The work was not hard. It was heavy. It sat on my shoulders like a thing I should be doing but never wanted to start.
So I would put it off. For weeks. Sometimes months. And the sites would drift — stale content, broken references, schema that never got added, metadata that never got written. Not because I did not care. Because the friction between caring and executing was too high.
That is over now.
What Changed
Two things happened in sequence. First, I killed the CMS. Moved everything to static HTML, flat files, Git. No more dashboards. No more login screens. No more clicking through seven menus to change a sentence. The content lives in files I can see, search, and version.
Second, I started directing AI through the work instead of doing it myself.
That second part is the one people misunderstand. They hear "AI does the work" and picture someone typing a prompt and walking away. That is not what happens. What happens is I sit down with a clear picture of what needs to change, describe it in plain language, and Claude executes it across every file that needs touching — while I watch, course-correct, and make judgment calls the model cannot make on its own.
It is conducting. Not playing every instrument.
The Old Way
In the CMS era, maintenance was a solo performance. I was the researcher, the editor, the QA tester, and the deployment pipeline. Every hat, every time.
Need to update operating hours for 40 attractions across six islands? That is me, opening each page in Expression Engine, cross-referencing Google, making the edit, saving, previewing, publishing. One at a time. For hours. For days.
Need to add structured data to a site that has never had it? That is me, learning the schema spec, writing the JSON-LD by hand, pasting it into each template, testing it in Google's validator, fixing the errors, doing it again. Multiply by every page type on the site.
Need to audit whether content from 2019 is still accurate? That is me, reading every page, checking every claim, deciding what stays and what goes. With no tooling to help beyond a browser and a spreadsheet.
Each task was doable. None of them were hard. But stacked together, they formed a wall of tedium that made me avoid the work entirely. And the longer I avoided it, the more the wall grew.
The New Way
Now I describe the wall. Claude knocks it down.
I do not mean that flippantly. The description matters. The specificity matters. When I sat down to do the five-hour audit of Hawaii-Guide.com, the reason it worked was not that the AI was fast. It was that I knew exactly what needed to happen across those 193 pages, and I could articulate it clearly enough for Claude to execute without ambiguity.
That is the conductor's job. You do not play the violin. You know what the violin section needs to do, when they need to do it, and how it fits with the rest of the orchestra. The value is not in the execution. The value is in the vision of the whole piece.
Post-CMS, that is my entire workflow:
- I assess. What is stale? What is missing? What would make this better?
- I direct. Here is the scope, here are the files, here is the standard.
- I review. Did it land? Does it match my intent? What needs adjusting?
- I ship. Git commit. Git push. Done.
The research, the bulk editing, the schema generation, the cross-referencing — that is Claude's job. And it does it in minutes instead of weeks.
Why This Matters Beyond Productivity
The productivity gains are real and I have documented them. But that is not the part that changed my relationship with my own work.
The part that changed everything is that I stopped dreading maintenance.
When a task takes weeks, you avoid it. When it takes an afternoon, you just do it. The emotional weight disappears. The guilt disappears. The backlog disappears. You stop being a person who knows their site needs work and start being a person whose site is current, because the cost of keeping it current dropped by an order of magnitude.
I am writing more. Publishing more. Shipping more. Not because I found more hours in the day — because the hours I had stopped being consumed by work that felt like punishment.
The Altitude Shift
There is a concept in management theory about operating at the right altitude. Individual contributors work at ground level — they execute tasks. Managers work at mid-altitude — they coordinate execution. Executives work at high altitude — they set direction.
For most of my career as a solo operator, I was stuck at ground level by necessity. I was the only one who could do the work, so I did the work. Every instrument, every note, every rehearsal.
AI gave me altitude. Not by replacing my judgment — by replacing the execution loop that kept me pinned to the ground. I still make every decision about what gets built, what gets written, what gets shipped. But I make those decisions from the podium instead of from inside the string section.
That is what post-CMS actually means. It is not a technology choice. It is an operating model. Static files plus agentic AI plus domain expertise equals a solo operator who functions like a team.
The Protocol: The CMS kept me in the weeds. Static plus AI put me on the podium. The job did not get easier — it got higher. I stopped tuning instruments and started conducting the orchestra. The music is better for it.
