I have been thinking a lot about the current state of tech lately. AI is everywhere, and it kind of starts getting on my nerves. I have tried Claude Code, and it didn’t work for me. I know I could write a bunch of markdown files to guide the thing better, but I’d rather spend the time coding than writing instructions for a tool. So, with that out of the way, I thought I could try a different approach.

Instead of coding with AI, I conducted a little experiment. What would happen if I used Claude for other tasks? I had two things in mind that I would call hobbies of mine. In the first experiment, I asked Claude to be my editor. Very occasionally, I write on something that could one day become a short story, or even a novel. The important part here is how it handled the task. To give you a little more context: I have a very rough first draft of what could become the beginning of the first chapter. It’s around 5000 words, and I basically just wrote down what came to my mind. So, the prompt I gave it sounded something like: “You are an editor. Read the following text and tell me if you can find anything that needs changing.” The result was actually really good. It noted that the story sounds rushed, and that there are massive plot holes. I already knew that, so I responded with: “This is just a very rough draft, and at the moment, I’m just writing down what comes to my mind. The plan is to polish the text afterwards.” And now comes the interesting part: It immediately shifted tone; instead of being critical, it all of a sudden started giving me hints on how to continue with the story. Claude stated that the story so far is very intriguing and fits perfectly into the genre. It was so enthusiastic and validating that I almost felt like a real author. We had a little bit of a back-and-forth conversation, and after a while, I caught myself having a bunch of fun talking to the thing. We were bouncing ideas off of each other, adding some to a list, replacing others we thought valid before, and after a while, I had a complete list of things to add to the story to finish the first chapter. To be clear, I won’t use any of that, but it gave some interesting ideas I will definitely consider adapting to fit the story. As someone who has no idea how to write a novel, the entire process was really fun, and also very productive.

The other thing I wanted to try is solo role-playing. I have two TTRPGs I like to play from time to time, but I don’t have a group, so I always play solo using the Mythic Game Master Emulator as a guide to advance the story and help with ruling. So the next experiment was using Claude as a game master for a round of role-playing. Now here comes the tricky part. When I asked it if it could do that at all, it told me it could. So we started setting up the game from where I last left off, clarifying some details about my character and what part of the adventure we’re in. The start of the game felt actually very good. We were running in a comfortable pace, the adventure progressed naturally, and I almost felt as if playing with another human being. I was totally in the zone, combat flowed, skill checks flew by, the chaos factor jumped up and down, random encounters were added to the game, and in the end, I had finished three scenes. And then I made a big “mistake”. I asked Claude if it actually knew all the rules of both the game and Mythic. Of course, it doesn’t. It’s not part of the training data, which makes sense of course, but I hadn’t considered that when starting the game round. It told me, though, that I could upload PDFs of the full rules so it can look them up when necessary. Mythic I could actually upload, but I don’t have the rule book for the game as a PDF, so I would have to scan around 450 pages, convert them into a PDF, and then hand that off to Claude for reference. Of course, I didn’t do that. So, in the end, I ended up with a round that felt really good, although it could be that we had a couple of errors in it. In the context of a TTRPG, that isn’t so bad at all. When playing in a group, the GM also has the main job of keeping the flow of the game running. And if that means improvising from time to time, that’s totally okay. But again, I don’t think I will be using any AI for the foreseeable future to run my role-playing sessions for me anymore. Not because I didn’t like the result, but setting up the thing to have all the context it needs, telling it how to set up the scene, and generally just having to type all kinds of things into a prompt while also trying to stay in the flow of the game is exhausting. And I also keep a game journal, so that’s already two kinds of text to keep track of while also playing a game solely in your imagination. My mind actually needed a rest after just playing for an hour.

So, here’s my conclusion about where and how to use AI: I have absolutely no idea 🤷🏻‍♂️. For coding, I don’t want to use it since I have too much fun coding myself. For editing, it was too enthusiastic and validating while it should have been critical, giving me the wrong impression about the current state of my ability to author a book. For gaming, there’s too much stuff to keep track of, and too many things to remember to actually enjoy using it. The good, and also bad, thing is that it’s well-trained to mimic human interactions. After a while, I almost forgot that I was talking to a machine. It’s so easy to lose yourself in conversation with AI, and it’s even easier to let it do the thinking for you. I’m not really comfortable with that, so I guess I will delete the apps from my iPad and iPhone for now. I’ll keep my Claude account should I ever need it again, but I’d rather have conversations with real people.