Last night I needed a simple thing for TROM II that I wanted to create with HTML plus CSS. An animation of a progress bar that goes from 0 to 50% and stops for 10 seconds. Then continues to 90% and stops for 10 more seconds, to then continue to 100%. I do not know how to do that. So I used chatgpt to test it.

The fabulous: the moment I told the chatbot to create that, it created one. Like it coded it. That's impressive.
The disappointment: when I tested it it was a simple progress bar that went from 0 to 100 percent. That I could do, mister chat! But I wanted something more complex than that.
The frustration: It took me close to an hour of back and forth to be able to make the chat do the code properly and achieve what I wanted to achieve. In that meantime it told me countless of times things like "I apologize, here's the code that actually works." It was bullshitting me all the time.

And the only reason I could tell how badly it did, was because it was a piece of code and I could easily test that. It was a simple task and it failed so many times. I am sure I could have done that in an hour if I were to read some documentation.

But again this is the main issue with these chatbots that I do not know if it will ever be fixed: they are BIG bullshitters! And you never know when they are bullshitting you. So if I cannot trust a word of what they say, they are useless for everything, except "creative" non-factual stuff like make me a poem and shit like that.

They are mesmerizing tho. I'll give them that. They look "magical". But we are fooled.