To think of a new article, I open an old one.
It sounds like working backward, but it turns out to be rather interesting.
Before closing my previous blog, I asked Kimiko-chan to write Python code for downloading the posts. Kimiko-chan is my nickname for ChatGPT. Python is a programming language; what I asked for was a tool to help keep the blog’s contents at hand.
This was not a request for new writing.
It was a request to preserve writing I had already done.
Later, as plans to restart my website took shape, I supplied the saved blog materials—this time as a source of ideas for articles.
In the saved text, I can read the titles, dates, and words written at the time. It is not a polished list of achievements. Things I tried and passing observations are still there as part of that day’s writing.
Take the post about making business cards, dated November 27, 2024. It records what I told the AI, the repeated exchanges, and the small adjustments I made myself at the end.
That is much more specific than “I made business cards with AI.” It preserves not only the result, but something of how I got there.
Kimiko-chan’s role this time is to read that writing and organize possible topics. Mine is to compare the old record with my memory and check what is appropriate to publish.
Even in something I wrote myself, there may be points I would now explain more fully. There may also be information I should leave out of a public article. Having something saved is not the same as being ready to publish it unchanged.
Nor does AI remember everything about my past. This time, there is text available to read. We can check what it actually says as we work.
A backup can help restore what has been lost. But the things I kept can also help me find what to write next.
I started discussing a new article, and my past self turned up with supporting documents.
At this rate, we may not run out of material anytime soon.