image.png

This is why so many are sabotaging their AI projects...

I wonder where this came from, and when it was taken, or whether someone totally made it up.

Basically if you push AI enough it just agrees, F*cking useless

Gpt i assume?

Yeah, but I've seen the exact same response on others. Curious how it picks out October

Nutso

I just wish there was a way to not have it be so perky. No matter how many times I say just give me the data without being all flowery and it never does.

you can change the tone (at least in some) to be snarky and what not

Once I get my in-house LLM setup that will be my first thing.,

I switched to Claude a while back but haven't paid attention to that stuff, so I'm not sure if it adds its own pizzazz.

I'm still just running CoPilot, for what I use AI for it has worked good so far. But really want to get a centralized locally processing one so that every time I ask it something they don't "kill a polar bear" that's what my daughter equates AI too. 😉

I mean, smart kid. 🙂

And I immediately thought of someone at Microsoft seeing my question, standing up, and racking a shell into a shotgun.

They would load the shell backwards and then blame Citrix.

i spent the better part of an hour earlier arguing with copilot about a simple batch script i asked it to come up with - since it forgot to include an important “goto menu” at the top so if i’d run it as is it would just process the first section in order.. kept trying to argue with me that the encoding on the file was wrong instead of the simple piece i kept asking it about

I generally only use AI for assisting with scripts, and I've found that Copilot is, by far, the worst at it which is actually pretty hilarious. ChatGPT and Claude were far better. Claude has found solutions that the other two didn't for a particular problem, so amongst other reasons, I only use that now.

Cursor is king for code, according to everyone I talk to

Literally have never heard of that until just now

It's supposed to be the dog's bollocks

Interesting. It keeps going back and forth between Codex and Claude Code. 5.4 beat Opus 4.6, but Opus 4.7 beats 5.4... It's a leap frog game. I can confirm that CC is fucking GREAT at code.