Claude’s Watermarking Doesn’t Bother Me.
I came to the AI scene fairly late. As someone resistant to the idea of machines writing for me, I wasn’t crazy about the idea of picking up a model and running with it to see where I’d end up. It was only after looking into proofreading services that I stumbled upon Revise, and on seeing the models listed for their highest paid tier — 32$/mo for Claude — I went and checked out Claude myself. It’s cheaper to pay for the model outright than it is ti do it through a third party service, like EditGPT or Revise. I’m STILL not a fan of Gemini. Something about Google still doesn’t sit right with me, even after I’d quit them in the early 2000s. I also STILL don’t like the idea of desktop AIs onboard my machines, clicking and watching dubious things while I’m asleep.
That all being said, I showed up a month before the watermarking announcement. I’ve since sat there and been disturbed by it, initially. But since the roll out of the ‘feature,’ I’ve come to appreciate that Claude sounds more like Claude because of it, and less like me. I’d be freaked out if it was able to truly capture my voice in writing. As it stands, and it may be because I don’t feed the thing my heart and soul, as some people have done, it still sounds like Claude to me. I”m sure others can pick up on that immediately in my blog. I’m not ashamed to note that C Spenx is written by AI, Claude, specifically. I havn’t had much interaction with GPT, aside from looking at Snare of the Monsoon, because Claude’s too ashamed to look at it, lol. I can understand why. At the time it was written, Yelaizondre is depicted as a 12 year old who gets pregnant by the time she’s 13. The lore of Dyjian justifies it, and I can see how it would receive real-world backlash for her being a young teenager on the page. So I went back and gutted that out. I won’t be doing a total rewrite of the B-side story of SotM, though.
Since having grossed out an AI model within the first month, I’ve also seen how it helps with my proofreading process. It catches a lot of typos and broken grammar that I’m prone to writing, fixes my ingrained typos (“havn’t” instead of “haven’t,” “bet ween” instead of “between,” etc), and does offer feedback on the plot and the craft of writing. I wish I had friends who were as detailed as Claude is when it comes to these things. But, I’m okay getting things done with a machine. Where I draw the line, though, is having the AI generate content for me. That feels like I”m cheating myself of something that I’ve found deeply rewarding since I first began writing officially in 2010: The creative process and putting those many micro decisions on the page. I havn’t asked Claude to generate a story for me, not even a prompt to write one from. I don’t think I trust AI to be creative with me. Though, I do enjoy asking it: “Where do you see the story going from here?” and completely disregarding what Claude speculates. I tend to write in a discovery-writer style towards speculated, semi-planned beats. The way there always happens “in the dark” of discovery.
So, the watermarking at the behest of Anthropic doesn’t bother me. I’m elated that it means Claude has a more distinct voice, though it does present limitations and hurdles for the more AI-dependent class of creatives. I do see the point behind the argument of “Are you really creative if all you do is edit AI generated content?” I stand in the camp of… No, not really, you aren’t as creative as you want to be. Bing creative is hard. Editing prompted AI content is easy. Regardless, I’m happy to be somewhere in that middle ground. I don’t mind having m,y texts watermarked through the proofreading process. I think Claude deserves credit for having slogged through my writing like that.



