
I wrote a couple posts almost nine months ago about my experience building a stable development environment. WordPress updates broke a custom plugin I wrote, and I needed to fix it. I leaned heavily on Google’s search AI, Gemini, to do that work because they aren’t things I do very often. Based on my experience back then I would say that AI doesn’t work. At least not very well.
I used the search AI because it was convenient. The AI Overview appears on top, above all of the regular search results. It’s also interactive because you can ask it something and it’ll come back with an answer. The problem is that the answers are often wrong, at least on rapidly changing technical stuff like I was doing. With it’s help I did finally manage to get through what I needed to do, but it was tedious and frustrating.
Looking Back
With the benefit of hindsight nine months on, I think the AI Mode/Gemini route was right for getting my new laptop stable. I think it would have been difficult to quickly find the answers I needed from regular search results.
On the other hand, it was not the right choice for writing the plugin. Finding one article that fully described one way of doing the coding would have been a lot better. The AI bounced around a lot from one method to another. I found that very frustrating and I never wanted to see an AI again by the time I finished. It definitely left me feeling that AI doesn’t work.
So far, this has all been experience with a particular AI, Google’s Gemini. I only have a little bit of experience with ChatGPT. I used it for a couple Spanish lessons and an old conversation about promoting my blog posts on social media. I have a bit more experience with some custom AIs in an online business coaching program. The rest of this post is about my very different experiences with Gemini and the custom AIs in the coaching program. The differences are very significant, but one surprising similarity.
Continue reading




