The Wall of Text Fatigue

I think I have “wall of text fatigue”. When landing on a repo’s README.md or some new project’s landing page and I see the usual markers of AI-generated content (many lists, emoji tables, verbose metaphysical descriptions of simple things, over-”communitiesed” descriptions of one-human shows), my brain goes ssssshhhhhhhhhhhh (you know that sound of white noise?).

Also: using tools like Claude’s research option again generates a wall of text. I can’t read this. My brain again falls into the ssssshhhhhhhhhhhhh mode. I recently started to take this wall of text, pass it to another agent session, and ask for a TL;DR to get a summary and be able to ask some questions.

And: when skimming through repos like https://github.com/lopopolo/harness-engineering (not a comment on the content, more an example of the form), I get the same sssshhhhhhhhhhh in my head. This reads as if it is some metalanguage based on English that is meant for computers (which it actually is) and not for human consumption. In the beginning, the model even quotes its rigger. To understand this document, you might need some graph view.

Is this our modus operandi now?

I myself have one or two (maybe even some more) projects that are pure AI slop. For example, the landing page of https://flarelines.cc/. I did some rewriting of the index, but the subpages are already slipping towards slop.

I guess for some things this is fine, but be honest about it. Tell the reader/user who you are, what this is, what you expect from it, and whether this will still be there or maintained in three months.

For reading content (as I call it, not-code¹), I find it even more infuriating to get something passed off as your own that is written by AI. Be honest about it. My friend Ingo Hinterding introduced the AI meter on his blog. That is okay-ish, but it still does not (at least not for me) communicate to what extent AI was used. I know it is cumbersome, but if you use AI (not you, Ingo, but everyone else), explain to me beforehand what I should expect. Do you use it for spell-checking, outlining, ideation, or formulating? Or did you just have an idea and one-shot that whole article while sitting on the toilet and browsing through Etsy? It’s okay if you find this useful for someone, but please, not for me.

For artistic content, even more so. Make sure that the generative AI image you add to your post, which took you no effort, is also marked as such, please. I see generative image models as a useful tool in creative imagery processes. I think tools like ComfyUI or InvokeAI are very powerful and demand a lot of knowledge and attention to detail. Again, this is no one-shot prompt from the bathroom.

So I put this at the end of this post, but I might add an AI-use disclaimer to this blog.

All words that convey information for other entities (human or not) are written by me. The human behind this blog: call me Fabian. Or Fab (but never Fabi or Fabio). I use AI to spell-check and point out errors, ask for translations, or sometimes ask whether I can say something in English in a certain way. I don’t let AI create my thoughts.

I tried using AI for non-code ideas, but it is dull and only a faint echo of what it already knows about me from saved memories or context.

I use AI for generating code. I use spec-driven development with long ideation back and forth, visual sketches, and test-driven development for many projects. I tend to ignore the code for prototypes or domains where I am not at home (see the Flarelines project). I put a lot of effort (and copy-paste from others) into my AGENTS.md and agent skills so the AI does what I want it to do. I would love to run some AI locally for that, but we are not there yet.