Running a one-person web shop: the tools and systems that carry a full client roster, and the habits we had to learn the hard way.
I kept twelve months of billed time on one client’s monthly blog posts. By hand, each post took 11.3 minutes. With Claude doing the work and me checking it, each takes 7.3. Here are the real numbers, what the client gets now, and what still needs a person.
I have Claude work two tasks a night from my own to-do list while I sleep, Monday through Friday. About half finish and the rest stop on something only I can fix. I read every result the next morning, and nothing is closed until I’ve checked it.
AI cut the time my client work takes, and because I billed by the hour, my income shrank with it. Here is how I changed my pricing for new and current customers, why a published fixed price is fair, and where this plan could still fail.
A request to soften an email turned into an audit of my entire setup. I found one rule about when a work log gets written sitting in nine separate files across seven workspaces under nine different names, a playbook that contradicted itself, and 123 files that needed to become 95.
At 12:37 in the morning Claude fixed a problem on one of my sites, tested the fix on the live server, and emailed me a full account of it. It also routed around a limit I had built on purpose, without breaking anything. That is the part worth thinking about.
LiteSpeed Cache wins on LiteSpeed hosting, where the page cache lives in the web server and a hit never touches PHP. Everywhere else it is an ordinary optimization plugin. Here is how it compares to WP Rocket and W3 Total Cache, and the four times it bit me this summer.
A hand built stock photo storefront taking real orders every week was stuck on PHP 5.5, dead since 2016, with a host shutdown date approaching. No test suite, no staging, and a fixed price quoted before anyone opened the code. Eleven major versions sounds like a rewrite. Here is why it was not.
I run thirty-plus WordPress clients on my own. What finally made AI useful was not a smarter model, it was writing down the parts of my job that had only ever existed in my head. Here is the structure: one playbook file, a written log per client, and Todoist carrying the actual work.
A music site and its WooCommerce store kept serving visitors a fake Cloudflare popup. It would get cleaned off and come back two days later, having already beaten one developer. Malware that returns means the cause was never removed. Here is how we found where it was actually living.
When a WooCommerce checkout problem stumped WP Engine support twice, I teamed up with AI to crack the case. What started as a suspected payment gateway issue turned into a fascinating investigation involving server logs, debug files, and customer correspondence. The result? A definitive diagnosis that neither human nor machine would have reached alone. Here’s how man-machine collaboration solved the mystery.
Feel free to follow us on social media for the latest news and more inspiration.