Migrating TailSense to a new tech stack from WordPress took weeks, migrating hvitravnur.com took 2 hours. Why?
As I mention in my article about LLMOps and the TailSense case study, I now have a toolbox and way of working when it comes to using LLMs for development work. I don't have to make the same mistakes again, and development becomes a much more streamlined process.
Why did I migrate the company page?
I migrated TailSense to a new tech stack because WordPress was slow and very manual for that use case. For the company website, the main issue was the difficulty with using the free themes; it took an obscene amount of time to edit the pages the way I wanted them. Even if I had paid for some fancy extra features, the process was still slow. With Claude Code, I have 4-5 testable drafts in seconds. Though it might sound difficult and time consuming to use .js files to post articles, its not; my work consists of writing the article and later pushing the code the web host's FTP server, Claude Code basically does the rest for me (from inserting images, links to pushing to building the page).
Ok, how did I do it?
I wrote a PLAN.md file which was basically a copy pasted version of the TailSense one, of course edited to suit the company page needs. The plan laid out the goals, what counts as a success and how testing is to be carried out. As for TailSense, the plan included the creation of drafts in locally accessible HTML files that I could preview; I asked for 5, and finally settled for one.
I then let Claude Code create a plan in plan mode based for the specific tasks per goal/plan phase.
I already had MCP instructions in place for pushing to GitHub (version control matters people!), and of course re-used that to incrementally test and deploy (yes, even during a 2 hours dev shift, that is important).
I put Claude Code on auto edit mode and … done. Pushed to FTP, changed the web host settings to point to the FTP where I had the Next.js build instead of the WordPress stuff and that was it.
Key take-aways
Is the site perfect? Oh heck no. The chosen tech stack is not optimal and I will probably either re-migrate or adapt it; Next.js and my chosen web host Loopia do not go hand in hand, and I feel its too much overkill to host this site on Vercel at the moment. I have a new page in the pipeline that will be created for course content, which in itself might warrant a new tech stack.
The main take-away is to start, do it quick, brake s***, learn and optimize. Though obvious, this is an "obvious" practice I have to remind myself of at times; spending too much time ideating or researching will kill the idea; an idea that this time and age might be obsolete the next month.