SN Dev Journey Day 132: Rise of the CursorBot \ stacker news ~devs

pull down to refresh

SN Dev Journey Day 132: Rise of the CursorBot

3096 sats \ 11 comments \ @SimpleStacker 6h devs

132 days ago, on February 8, 2025 (#879762) I decided to start contributing code to Stacker.News. Since then, I've contributed 11 merged PRs and earned just under 2m sats for my contributions. Stacker.News was my first foray into open source, and it's been a pretty pleasant experience, not least because you actually get paid for your work. It ain't enough to replace a full time job, but it's a nice little side benefit for doing something that is fun (I'm the kind of person that plays programming games for fun) and contributes to a software that you use daily.

Now I'll be straight up and say that I leveraged AI a lot to help me contribute. Although I have decent coding experience, none of it was in javascript or web development, so learning JS and React was a new experience for me. My experience was that AI can't solve the problem for you from 0 to 100 (or maybe I just don't know the right prompts), but it's really useful for understanding the codebase and finding out which code needs to be changed. It felt like a collaborative experience: I had to combine my domain knowledge of stacker.news and of the task with AI's vast knowledge of JS and ability to read code quickly to come to a solution.

Just recently, my latest PR submission was automatically reviewed by CursorBot. This was the first time that happened, and I don't know if it's something that GitHub put in place recently for all users, or if k00b and ek recently activated it. Of course, it pointed out all the deficiencies in my code, and it made me feel quite stupid to be shown up by a robot. Still, it made some good points, so I went back and fixed all its suggestions and resubmitted the PR. Like a super pedantic boss, it reviewed my code again and came up with two new points to criticize me on! This time, though, I felt like the criticisms weren't as relevant because the conditions under which there might be a problem should never actually arise (afaict)

I'm just remarking on this because the experience of proactively using an AI that you call on to assist you has a very different feeling from getting your work reviewed by an AI when you didn't specifically ask for it. Proactively using AI feels collaborative. Getting reviewed by an AI felt dehumanizing, in the literal sense of feeling like my humanity is inferior to the bot's supreme coding ability.

Anyway, just some random thoughts and reflections. I'm not asking SN to stop using CursorBot since I can see how it improves code quality... but I did just want to share my reflections on how AI is affecting us since I know a lot of people here are interested in that stuff.

view all related items

80 sats \ 3 replies \ @optimism 2h

it pointed out all the deficiencies in my code

The first complaint looks valid but really poorly addressed by the digital overlord - I don't know the codebase but in general this is something that human review may have flagged up too. I saw your comment "this never happens", but that makes it all the more important to fix the code.

Think of it this way: if a field is never undefined because it is always set on every possible code path, then you shouldn't put a default but instead do error handling (and logging!) when encountering undefined. Assigning a default value on a case that should not exist is a problem because if for any reason (in the future) it does exist, it shouldn't just gracefully apply your assumption. Assigning a default value is what leads to hidden tech debt in this case, because it will mask future errors.

100 sats \ 2 replies \ @SimpleStacker OP 2h

Maybe this is where my inexperience as a real developer starts to show.

My thought process was: If there was an edge case I didn't think about, I'd rather the app just proceed on, with a slightly messed up carousel order, than for it to error out. Happy to be corrected as to what k00b and ek think the standards should be in this case.

30 sats \ 1 reply \ @optimism 1h

You're not wrong in wanting to prevent edge cases. But handling an error and erroring out are two different things. The later specifically means that an error isn't handled.

0 sats \ 0 replies \ @SimpleStacker OP 48m

Understood 🫡👍

87 sats \ 1 reply \ @k00b 5h

I turned it on yesterday. It's a new Cursor feature. I left the default settings but it can be configured to only run once or when summoned. I figured we'd wait and see.

We used code rabbit a century ago (in AI time) and we were similarly frustrated by how noisy it was.

50 sats \ 0 replies \ @SimpleStacker OP 5h

Got it. Yeah, I think it's helpful, almost like a linter on steroids. But it's hard to know where to draw the boundary of helpful feedback and stuff the developer already saw and just didn't think was an issue.

30 sats \ 0 replies \ @WeAreAllSatoshi 46m

Back in my day, we didn’t have AI bots to help us contribute to open source projects!!

But really, good stuff

30 sats \ 0 replies \ @seashell 1h

AI’s helpful right up until it starts telling you what good code is cuz at that point it is enforcing taste I'm not sure how that’s progress.

30 sats \ 2 replies \ @Scoresby 5h

I really enjoy your updates. I was wondering when you were going to give us another update. 2m sats! Pretty awesome job!

102 sats \ 1 reply \ @SimpleStacker OP 5h

Yeah, when I started these posts I thought I'd have more to write about, but usually I didn't feel like there was too much to say haha

Sooner or later I'm going to have to dive into Prisma, which I'm dreading, but I'm waiting for the db to be refactored first.

30 sats \ 0 replies \ @Scoresby 5h

Well, whenever you share a glimpse, it's very interesting for those of us who are dev-curious.