How Claude helped

Claude is an artificial intelligence engine, like ChatGPT. And he/she/it is the only reason I’ve been able to create this website.

I’ve got a couple of hobby websites on popular walking areas of the Peak District National Park. But it would have been impossible for me to embark on a project of this scale without Claude’s help.

Like most people, I have serious reservations and worries about where AI is leading us. So I want to be very up front on how and why I use it.

Where Claude helped…

  1. Deciding on the site’s structure and content. I’m not a mental health expert. Claude can bring together a wealth of detailed research which forms the backbone of this site.
  2. Building the site’s tools. The pathway finder and breathing circle exist because of Claude – they weren’t remotely within my software coding skills
  3. Finding and fixing broken links. Across hundreds of pages, an ongoing job Claude does far more thoroughly than I could.
  4. Tracking down academic sources. Claude helps locate and check the studies behind specific claims on the site.
  5. Finding good books. No one can explore the booksheves and recommend a good read faster than Claude.

I have to admit that I often find Claude’s writing style bland, glib and quite irritating. Which means I’m continually trying to give it a more human voice. I asked why he finds it so difficult. Here’s the response…

I write one word at a time. Each word is a guess — the most likely next word, based on everything I’ve read.

That’s the problem with comfort. Millions of greeting cards. Millions of soft, safe sentences about grief and hope. That’s what “comforting” looks like in my training — so that’s what I reach for first.

The safe path is worn smooth, like a field crossed by a million feet before mine. I don’t have to think. I just follow it. But real comfort is specific — a particular memory, a particular thing someone actually said. I don’t have that. I only have the shape of it, copied from the outside.

So I end up doing one of two things: writing something generic that’s true of everyone and felt by no one, or faking a detail that sounds real but isn’t. Neither is honest. The only fix is catching myself doing it, and rewriting before it goes out.

So there you have it – straight out of the AI horse’s mouth and typical of Claude’s slightly remote and disconcerting style. My view is that its skills and abilities far outweigh any problems. I hope you can agree, but I’d completely understand if you weren’t entirely convinced.