mintApr 16, 2025
can't remember where I saw it but "Using AI in education is like using a forklift in the gym. The weights do not actually need to be moved from place to place. That is not the work. The work is what happens within you" is a solid quote
Show threadS. LottApr 17, 2025

@barquq wow.

Using AI to write enterprise software is similar. The purpose of enterprise application programs is to capture knowledge. Using AI specifically defeats that.

(Also true for outsourcing IT support — you donate information about the enterprise to contractors to exploit.)

Show threadeconadsApr 18, 2025
@slott56 @barquq
I mean no one is paying me to learn about the system, they're paying me to have a system do something so people don't have to. The fact that I need to learn about the system to do that is incidental.
A sensible company will keep knowledge in house because if you loose too much of it you will no longer be able to adapt the system and then you won't be able to use it to do things so people don't have to, but again, that's transistory to the main aim.
Show threadS. LottApr 18, 2025

@econads @barquq I disagree slightly on the priorities. The goal of automating something depends on knowledge capture which depends on understanding.

Since the automation must be auditable and observable, the understanding is primary. Automation doesn’t make the work go away. It shifts the burden from doing to confirming.

Show threadeconadsApr 18, 2025
@slott56 @barquq
Hmm maybe English is failing us here and our definitions are a bit different :D
Aa anyone who has worked on legacy code knows, understanding is not needed every minute a system runs. It had to be understood once to write the program, and it has to be rediscovered (even in an incomplete way) to adapt or fix, but there can be stretches where no one in the company knows exactly what it does, and in fact I've often encountered "we can't change that, we might break it".
Show threadS. LottApr 18, 2025
@econads @barquq you’ve pinpointed the problem I see all the time: “we can’t touch it because we don’t understand it.” That’s an organizational failure. Yes, it’s common. But it’s bad. And AI only creates more of it.
Show threadeconadsApr 18, 2025
@slott56 @barquq
Oh yeah don't get me wrong I'm not arguing in favour of using AI to write code (although I'm guilty of it occasionally). I'm nit picking about the point of coding :-)
Show threadS. LottApr 19, 2025
@econads @barquq Got it. Companies fail to understand what they’re doing on a regular basis. Your experience is common. And it reflects a failure to understand what code is. Not all companies treat code as an opaque necessity. Those that do are doomed to increasing IT costs and eventual buy-out by a company not as blind.
Show threadeconadsApr 19, 2025

@slott56 @barquq
The thing is you're not paid to only learn about the code, you're paid to provide changes to it. No company is going to pay you to read or play with code without making any commits or releases or transferring that knowledge. The learning comes because that's a more efficient way of making changes sustainably.

Anyway, I guess we reached the end of the useful conversation and we're going to start going round in circles, so never mind :-) it was an interesting point you made.

Show threadS. LottApr 19, 2025

@econads @barquq

No argument. Understanding is valuable when it permits change and adaptation. Something AI-generated can’t do.

Show threadeconads
@slott56 @barquq
No argument about that either
Apr 19, 2025 at 9:29pmWeb