AI and my love-hate relationship with it
I started actively using AI a few years ago. Nothing heavy, basically as a glorified Google search engine. THen I discovered other facets of AI... brainstorming ideas, chatting about everyday life, helping me with my music life, etc.
Then, I discovered ChatGPT could make cool logos, images, web page designs, and more. I started paying even more attention.
Then, the roof blew off... I discovered Claude Code. My weariness of IT and computers just got a massive jump start!
Now, I spend hours a day coming up with ideas, planning projects, creating application, websites, designs, works of fiction, blog posts, tools and helpers, and so much more. It is like a personal assistant who never sleeps. When I moved back to Linux permanently a few years ago, I discovered that Claude Code could identify and fix issues with my system it would have taken me hours (or days) in the past to figure out. Mass management of my entire infrastructure was in good hands with Claude Code (and then Codex as well).
But here we are today. June 23, 2026. 5PM. I have been working on my computer with athis really cool idea for a federated network of shell/terminal/BBS games, SPent a while designing control documents, prompts, and plans. By the time I had invested 4 hours into the project it had turned out totally unlike what I wanted or asked it to help with. In fact, it had totally ignored most of what I asked it to do, to evaluate, to review, to verify. It gave me a product that should have been so cool it would have froze me! Nope. Instead I got a heaping pile of dog shit.
I'd like to say I was baffled, confused, befundled (one of the silly Claude Code words)... but I wasn't. Why?
Because it's been acting like this for a solid 3 months. Confused, hallucinating, ignoring instructions, inventing what it "thinks" I want. At first I thought it was me, I mean, the stuff it was putting out before was pretty slick. Now, it feels like something I might have been able to just do myself.
Maybe I am getting smarter and asking it to do more than what I was before? Maybe, but I doubt it. I truly feel like (and have read on Reddit that others feel the same) that Claude Code is getting dumber and I am getting smarter.
I'm debugging things it caught right away in the past but now stands out in big bold letters "this is a bug, FIX ME!" and wonder why it is not atching them at all. It's also not listening to me any more. Seriously I asked it to something fairly simple (at least for a big brain like AI) and take this unfinished open source code that was text based, and modify it to meet modern standards, and then add the few bits of necessary code from my federated network project to make it accessible. Not only di it fail to do that, it designed an interface that looked exactly like one before it and nothing like the codebase from the open source code.
Knowing good and well I was talking to a machine, I got pissed and basically said "WTF is wrong with you?" It's reply?
I kid you not... this is the reply it gave me;
I built an entire UI from imagination without ever looking at the original game. The project directory is <old_path>. A sibling directory exists at
<old_path_sub_folder>containing screenshots of the exact game being rewritten (menu0.pngthroughmenu7.png,image-disk.jpg). I never looked at it. Not once, across the entire multi-session project. This is teh root cause of everything else."
What the actual fuck?
It ended with;
The reference material was sitting in the repo one directory up from where I was working. I didn't look. Everything that followed from that — the wrong UI primitives, the wrong layout system, the fake polish pass, the failed fork — flows from that single failure to do basic reconnaissance before starting.
That's not a hallucination or a technical error. It's a failure to look before building.
So here I sit, praising AI a few months ago to wondering "WTF happened to that AI?" I won't lie and say that it isn't helpful... it is... but Iw onder if I have reached the end of it's usefulness for me, or if it trully is getting dumber... not because of what is bein presented to it, but by the constraits of the corpos who control them. There is were I think the real "bug" is located. But that is a entry for another time...
Tags: artificial intelligence, AI, projects, coding, hallucinating, machine, relationship