zataomm
1d ago
using AI for a prolonged period will lower your overall ability to code as a developer
This strongly depends on what your ability to code was before you started using AI. The reality is a lot of people can make websites now despite not having a professional approach to coding. If you're a classically trained developer, computer scientist, and you only use Vim, then yeah, you're probably right. That set of developers is going to be a smaller and smaller proportion of the total as time goes by.
On the one hand, I'm sure you're right that it's not as good as a human developer. On the other hand, it's kind of like the web in general. Websites all kind of suck now because React gives developers the ability to be lazy and not worry about resource usage or anything else because everything's kind of "good enough." But the cost and speed advantage of using AI is so overwhelming that I don't think your concerns are going to hold up. Developers are just going to have to add more bad code on top of bad code until it works.
Full disclosure: I really think that a lot of concerns from sophisticated developers about how AI makes mistakes are just snobbery/posing. But that's what I, an unsophisticated developer, would think, right?
Like, I'm not saying I've never had an experience where AI is going down a path that I think is bad and needs to change course. But in those cases, I usually hit the stop button and say, "Hey, I think you need to do this in a totally different way," or I can just say, "Hey, your last change was terrible. Can you just revert it?" There are a few instances where I feel kind of dumb because I'm using three prompts to say like hey I need you to change the font size, when I probably could've done it more easily myself in that specific case. But overall, I'd rather be talking to my agent than looking at code.
This is fairly impressive. Did you make this with assistance of generative AI?
Oh, very much so. Github Copilot + Claude Sonnet 4, if you care to know. Technically I am a web developer but at this point I am basically all-in on using AI wherever possible. Nice to get done in a couple hours what would have taken me much, much longer, and much more frustration, using the old method of typing code by hand.
zataomm
2d ago
Thanks! My inability to find such a summary was what caused me to embark upon this folly.
zataomm
2d ago
I felt kind of annoyed by the claim that "most" Supreme Court cases go 6-3 along ideological lines, although I guess the more defensible version would be that the controversial cases all go 6-3 along ideological lines. Be that as it may, I created a website this morning to help understand data from the most recent term. Spent more time than I intended on this so I'm hoping someone else finds it interesting: https://wbruntra.github.io/scotus/?tab=dashboard
I know what you mean. Any suggestions on how to break down the data this way? I mean, at some point, if you just define important cases as cases that go 6-3 along ideological lines, then by the way that you've defined it, 100% of important cases are going to be decided along ideological lines. I'm trying to think about what would be a good middle ground that's still data analysis-based while giving insight into these potentially controversial cases.
On the Agreement Matrix tab, I've added a checkbox that excludes unanimous cases from the analysis. Which I think is an interesting way to look at it, because you're seeing that if there's any disagreement at all in a case, then, for example, Thomas and Jackson are most likely to find themselves on opposite sides, while Thomas and Alito are most likely to be on the same side. You didn't really need data analysis to come up with this insight, but I guess it's good to confirm at least that data analysis confirms what everybody knows.
https://wbruntra.github.io/scotus/?tab=matrix
More options
Context Copy link