Re: No cost, no value
This is a response to: https://jola.dev/posts/no-cost-no-value
I agree with… the first half of this post. You can mass produce software using llms, and a lot of this software does get abandoned the moment it’s finished because the creator moves on to the next thing.
First I just want to quickly rebut something. The reason why “vibe-coded twitter” has no value is not because anyone could make their own. It’s because everyone is already on original twitter, and the people who are interesting in leaving twitter have already migrated to bluesky. Making a social network have value is incredibly difficult because you need to convince people to join it, and no one has any incentive to join your vibe coded twitter replacement, and that would still be the case even if you had written it by hand.
Now, on to my main point. Jola’s claim (as I have interpreted it) is: “putting time and effort into making software makes it inherently more valuable than llm generated software, which is why so many of these abandoned softwares are generating and not handmade”. I have an alternative explanation.
Back in ye olden dayes (2019), software was a team effort. Yes, there were some people who made entire, large, functional, good pieces of software completely alone, but we generally looked up to this people as mythical geniuses. For us mortals, making a webapp meant you needed a web guy and a backend guy, and probably also a graphics guy to be safe. If the software was sufficiently big you would even have more than one of these goals. And then the three or five of you would work together, you would bond, and when you finished this thing it would be your baby.
The difference between this software and an identical one someone vibe coded in an hour isn’t in the software. The two pieces could be line for line identical, and one would still have a far higher chance of abandonment than the other. The difference is that the creators of the handmade (artisanal?) software are emotionally invested in it and each other.
When you build something in an hour by yourself, you can say “well, that was cool, on to the next”. If you spend a month building something by hand on your own, it’s still vulnerable to being left behind whenever you get sick of it. There’s nothing keeping you involved but sunk cost fallacy. But building something as part of a team is different. You’ve befriended these people, and if you abandon the project, you feel like you’re betraying them. You know if you’re not working on this together, the friendship will probably slowly wither away, especially if you only know them online. So you have a real incentive to not abandon the software.
Jola writes: “But I believe that sometimes LLMs take something away from you, something valuable. They take away toil, but we lose meaning. They make it too easy to not engage, to not learn, to not struggle.” But I don’t think that toil actually has value. I don’t think that struggling is something that anyone should have to do if they don’t want to, even if it does promote learning. I think this entire genre of argument against ai comes from the cultural Calvinism that the western world is absolutely steeped in.
When llms let one person do in a day what five people used to do in a month, the value that’s lost isn’t the “month”, it’s the “five people”.