I use Claude Code, and here is how

I adopted Claude Code about four months ago to help with code and writing. This blog is one of the places it shows up, and I would rather say so out loud than leave readers guessing.

The ideas behind any post on this site are mine. So are the anecdotes, the opinions, the experience the post is built on. The prose comes from me as well. What the model does is edit. I write a draft, hand it to Claude, and ask for the kind of pass a copy editor would do: catch the sentence that loops back on itself, flag the paragraph that says the same thing twice, find the word I meant to use but could not pull out of my head at one in the morning. The model is good at that. It does not get tired and it is not billing me by the hour.

Code is similar with a tighter loop. I describe what I want, sometimes in a sentence, sometimes in a paragraph and a half. The model scaffolds. I read what it produced, change what is wrong, run it, iterate. The architectural decisions stay with me. The reading and judging of the generated code stays with me. The model compresses the mechanical part of the work, which has always been most of the typing anyway.

The force multiplier line

This is the line I keep coming back to when people ask.

Claude Code is a force multiplier. A force multiplier is a double edge. A senior developer with the experience and the ability turns out high-quality code, faster and more well rounded. A junior or vibe coder turns out low-quality code, compounded by the AI into a mess.

I believe that, and I think it is the honest framing. The model amplifies what you bring to it. If you bring taste and the instinct to push back on a wrong suggestion, you get more of that on the way out. If you bring "make it work, ship it, I do not care how", you get more of that too, faster, and the mess is larger because the model can write the wrong thing in a tenth of the time you could.

Prose works the same way. If the underlying argument is muddled, an AI editor will produce a muddled paragraph that reads like it is not muddled, which is worse than the original was. The polish is real and it can be applied to nothing. That is the failure mode.

What I do not let it do

It does not invent the post. It does not decide what I think about WinForms, the Cooper and Geary form designer, or YouTube's funnel. Ask the model and it will happily generate a confident opinion on any of these subjects. Those opinions are not mine, I have no use for them under my name, and the way to keep them out is to write the draft before the model ever sees it.

It also does not get the byline. The work, in the sense that matters, is mine. The model is a tool in the chain, alongside the IDE and the keyboard.

Closing

The question is going to keep coming, and silence on it reads as pretending. So here is the answer. I use the tool. The output is better with it than without it. I am not embarrassed by any of that. Plenty of people in this industry are hiding their use of these tools right now and you can tell, because the prose smells and the code smells. Use the tool, bring something real to it, and the force multiplier works in your favour.