Comment Re:I remember a time when... (Score 1) 93
I won't argue but I'll share my experience. I was already better than most of my peers in my industry (software development), depressingly and frustratingly so most of the time. It sucks to feel obligated to carry team after team out of the piles of garbage that they wrote so that we can deliver quality products on time. Take my assertion for what you will. After switching to using LLMs for most of my work, the same gap remains - indeed it has widened.
Recently my co workers have been using them more... and they are amazed at the silliest and most basic of things... sometimes they produce lots of weird garbage _and they think it's acceptable_, at least until you explain reality to them. Part of the problem is that they are being encouraged to use these tools and be vocal about it ("I used AI for X!"... cool story bro). Whereas I've been using them much more covertly and pervasively. I'm not saying what I'm producing is perfect, but it's much better and much more useful and maintainable than what I'm seeing them produce. This is across multiple disciplines (embedded, backend web, planning, design, etc).
My output has changed, it is true. Quality is a little lower sometimes. But I get to choose when and where the quality is acceptable. Being able to judge quality is also a skill. Also, being able to iterate on designs and discard bad ideas quickly - LLMs open new doors here.
If you don't know what to ask for, you're likely to get poor results. LLMs can't think for you. If you don't know what quality is, then you can't judge how good the output is. Being impressed by parlor tricks is worthless.
Tools, techniques, and mindset really matter. ToasterMonkey's "delegating" comment is accurate IMO. I could do any one single part of the work better, but I'm not a team and I also don't have anything near enough energy to compete with the output from these tools. Plus, the improvement over time from these tools is yes gamed but also it is real too.
Ok so here I will argue a bit. The _user_ is the one that learns from experience here - not the LLMs. The user learns what works and what doesn't... how to set up the tools to be useful and productive and keep them the hell away from situations where they fail.
Recently my co workers have been using them more... and they are amazed at the silliest and most basic of things... sometimes they produce lots of weird garbage _and they think it's acceptable_, at least until you explain reality to them. Part of the problem is that they are being encouraged to use these tools and be vocal about it ("I used AI for X!"... cool story bro). Whereas I've been using them much more covertly and pervasively. I'm not saying what I'm producing is perfect, but it's much better and much more useful and maintainable than what I'm seeing them produce. This is across multiple disciplines (embedded, backend web, planning, design, etc).
My output has changed, it is true. Quality is a little lower sometimes. But I get to choose when and where the quality is acceptable. Being able to judge quality is also a skill. Also, being able to iterate on designs and discard bad ideas quickly - LLMs open new doors here.
If you don't know what to ask for, you're likely to get poor results. LLMs can't think for you. If you don't know what quality is, then you can't judge how good the output is. Being impressed by parlor tricks is worthless.
Tools, techniques, and mindset really matter. ToasterMonkey's "delegating" comment is accurate IMO. I could do any one single part of the work better, but I'm not a team and I also don't have anything near enough energy to compete with the output from these tools. Plus, the improvement over time from these tools is yes gamed but also it is real too.
Ok so here I will argue a bit. The _user_ is the one that learns from experience here - not the LLMs. The user learns what works and what doesn't... how to set up the tools to be useful and productive and keep them the hell away from situations where they fail.