I’m glad I’m not alone!
A new use case occurred to me just today. I was reading a chapter of Middlemarch and there is a scene where the Brookes, uncle and niece, visit a tenant and a long description of his dilapidated but picturesque house. I though it would be interesting to look at an illustration based on the text and ChatGPT produced a rather decent one. Definitely something I will try out more in the future.
I am also (re)reading Middlemarch, and while I have a nicely annotated edition, sometimes I would like to better understand its political elements. I have asked ChatGPT to contextualize some remarks, and it seems to do a really good job.
Hardly worth mentioning but I am using Gemini at work constantly to discover how to do small things in Excel or Windows. I’m sure it’s making me stupider, but this kind of info is so much easier to find in an LLM than it used to be in Google.
Example from just now, I saved a PDF into the wrong folder, on a huge server with a million folders, and Gemini told me how to find it quickly by searching the whole thing with the term “datemodified:today *.pdf”. It came right up. Sure, Google would have gotten me there, but it would have taken a lot longer.
It doesn’t make you stupider to:
- Learn things
- Use different tools
… just FYI. 
I only use it for skills I don’t mind degrading. Like making spreadsheets and flowcharts. Things I am perfectly okay with being kind of crap at. I’ve never been good with spreadsheets.
I think any time I am asking “How do I X?” I am doing okay. I have learned some skills from Claude. I’m slowly learning how to prioritize. I definitely write better federal grant outcomes because of Claude.
Claude is good at knowing how to frame an idea in a way that fits funding priorities. I had one case, where I wrote the whole grant, but we had a significant material weakness, which is that we did not have specific identified partners for completing the project. I asked Claude about it and it gave me a way of saying “we’ll find these partners later, we swear” that was really specific, concrete and intentional, you know, an actual plan. This worked so much better, in fact, the the funder circled back with additional questions (this is a very good sign.)
This is something I have to get better at knowing how to do on my own. But by working with Claude I’m gradually learning how to do it. I’ve been writing grants for 15 years and it has really helped identify the gaps in my knowledge and start to close them. I have enough professional discernment to know when it’s incorrect and so far it’s only really gone wonky a handful of times.