zlacker

If there’s a “tangible problem” people solve it with a SaaS subscription. That’s not new.

We developers are hired because our coworkers can’t express what they really want. No one pays six figures to solve glorified advent of code prompts. The prompts are much more complex, ever changing as more information comes in, and in someone’s head to be coaxed out by another human and iterated on together. They are no more going to be prompt engineers than they were backend engineeers.

I say this as someone who used TabNine for over a year before CoPilot came out and now use ChatGPT for architectural explorations and code scaffolding/testing. I’m bullish on AI but I just don’t see the threat.

>>akisel+(OP)
I'm just arguing that its a lot easier for AI to replace something that has objectively or technically correct solutions vs something as subjective as art (where we can just decide we don't like it on a whim).

>>oldstr+54
I’m arguing that there is no objectively or technically correct solutions to the work engineers are hired to do. You don’t “solve” a startup CEO or corp VP who changes their mind about the direction of the business every week. Ditto for consumers and whatever the latest fad they’re chasing is. They are agents of chaos and we are the ones stuck trying to wrangle technology to do their bidding. As long as they are human, we’ll need the general intelligence of humans (or equivalent) to figure out what to code or prompt or install.

>>akisel+P7
In the sense that someone asks "I need a program that takes x and does y" and the AI is able to solve that problem satisfactorily, it's an objectively correct solution. There will be nuance to that problem, and how its solved, but the end results are always objectively correct answers of "it either works, or it doesn't."

>>oldstr+Ns
Case in point, I guess :-)

>>oldstr+Ns
I think in both domains there are parts which are purely technical (wrong or right) and others which are well ... an art.

In art these parts are often overlooked, but they are significant none the less. E.g. getting the proportions right is an objective metric and really off putting if it is wrong.

And in programming the "art" parts are often overlooked and precisely the reason why I feel that most software of today is horrible. It is just made to barely "work" and get the technical parts right up to spec and that's it. Beyond that nobody cares about resource efficiency, performance, security, maintainability or yet alone elegance.