AI -bugs...

I was looking for trending softwares on repology.org then found a project, scrolled down and saw that it's made with AI and supervised by humans. I wouldn't want to use a software like that. At least it's mentioned and not kept private.
 
I just tried in Antigravity with Gemini3. At least this one compiles the code to test it if you let it (and got it right the first time).
 
Early November in 2025 I asked ChatGPT: "When will be this year's thanksgiving?"

The correct answer I hoped for simply had been:"November, 27th"

The answer I received was:"Thanksgiving is celebrated in the USA every year to remind of the first settlers [...] It's always a big holiday when the whole family get together [...] great feast with turkey [..] It's always celebrated at the fourth thursday in November. This year's thanksgiving will be celebrated at November 28th."

So, lots of blablabla I did not asked for, and what I asked for was wrong.

If somebody receives such an answer by a human being, anybody would say:"Would you please cut the BS and get to the point!" But with a computer everybody goes:"Wow! Look the machine can talk. Listen to it!"

To me that's all I need to know about AI.

I expect some day there will be articles on Wikipedia like "...the Golden Gate bridge in Los Angeles is green, 115 miles long, and 3 feet above the Atlantic ocean..." At some point people simply had no chance anymore and so gave up on checking and correcting the garbage it produces way faster than you can clean up afterwards.
But the worst part is, when you tell anybody those informations were simply wrong, they simply don't care:"Yeah? So, what?"
I wouldn't be surprised if in hundred years historians look back at our age and say:"This AI thing brought humankind the most massive loss of knowledge. The confusion caused by all those mixing of genuine facts to some inseparable garbage was very bad."
 
And since ai learns increasingly from ai, this is going to become worse over time.

Yay! Job security!
To be honest, we kind of see that in humans too. The amount of incorrect / cyclejerk learning that goes on in reddit is cringy. Juniors enter the industry with heads filled with absolute crap (usually involving C++30 or Rust).

So job security really is guaranteed for the rest of us! :beer:
 
I am more concerned with non-functional changes such as reformatting and whitespace that usually go along when asking a LLM to make a minimal change to a piece of code.
It's odd how it can't seem to get indentation right. What it does passes checkstyle, and I'm being heavily pressured to use AI so... 😈
 
It's odd how it can't seem to get indentation right. What it does passes checkstyle, and I'm being heavily pressured to use AI so... 😈
Talking about C/C++ here.

Indentation. "What is right" are we using K&R, BSD, GNU, Stroustroup? Every style has proponents and opposition. If a corporation or a multi-billion lines of code I think consistency is the key.
This is where tools actually come into play. Emacs, you can setup the init file to be consistent for everyone using emacs. But if someone uses some other IDE? Need to provide the equivalent init file. Vi/Vim? same thing. Winds up a team needs to either use the same tool across everyone or there needs to be a set of init/config files for style for every tool that the team could use.

checkstyle/lint type tools often have ways to tweak "acceptable". "right indentation" often is not as black and white as it should be.

My experience over a long time with different companies is you start with a predefined emacs style, then change basic offset to 4 spaces instead of 8, then other offsets related to braces, if statements, case statements to meet the corporate "coding style".

I think it doesn't matter "what" the standard is, just all the code I'm looking at needs to follow the same standards. (try looking at code following K&R then tracking over to code following Stroustroup. Breaks my mind)
 
Talking about C/C++ here.
Well, I was talkin' Java with Project Reactor. I don't know which is worse.

Indentation. "What is right" are we using K&R, BSD, GNU, Stroustroup? Every style has proponents and opposition. If a corporation or a multi-billion lines of code I think consistency is the key.
I think we've settled on spaces, and the Checkstyle rules are set up to require "at least x..." In any case, things ain't lining up. I absolutely refuse to get into lengthy discussions about whitespace and indentation, so if it passes Checkstyle, it's good enough for me.
 
Well, I was talkin' Java with Project Reactor. I don't know which is worse.


I think we've settled on spaces, and the Checkstyle rules are set up to require "at least x..." In any case, things ain't lining up. I absolutely refuse to get into lengthy discussions about whitespace and indentation, so if it passes Checkstyle, it's good enough for me.
Been in too many code reviews where some focused on "this is off by one space" or "you used the wrong camel case for this variable name" instead of "what are the requirements does this meet them"
 
Been in too many code reviews where some focused on "this is off by one space" or "you used the wrong camel case for this variable name" instead of "what are the requirements does this meet them"
Right? Who gives a flying f... - I'll keep it PG.

I worked at a small company where we only had two conference rooms. We moved to a new office, and had to come up with new names for them. Someone came up with "Tabs" and "Spaces", and I loved it. I was pulled aside and told to keep it down as that was a simmering conflict between the founders. I refused, and Tabs and Spaces it was!
 
Back
Top