Same for me lately. Though I still get anxious about it generating large files and not isolating logic enough. Do you run extra passes specifically to alleviate this or do you just accept it?
The point about disconnecting makes a lot of sense. This is a problem in society that surpasses just open source development and I'm glad to see you all take steps to alleviate that as much as you can. Can't wait to see the end result of this ridiculously cool initiative you've all been a part of!
Out of pure curiosity: was there a real, looming threat of burn out among npmx contributors that lead to the “vacation”? I haven’t been involved but this interests me. I can absolutely imagine that going super hard at something in such short time could lead to burn outs on the project.
You’re totally missing the point. The PR and suggested merge changes were good in this case. It’s a matter of principle because it was fully automated without intervention, intent or responsibility by a human contributor. The suggested changes will likely still find their way into the codebase.
You know you made it when there’s impostors running around! 😅 Thanks for the heads up!
Congratulations!
Work’s not paying for it anymore unfortunately and it’s too much of a personal expense. Government budget
cuts strike again! Have fun there everyone, I’ll be watching online!
I hope you looked at my name and realized all was still good in the world? 😅
It’s saying you’re in the wrong because you don’t care about localization you caveman English superiority freak!! 😇
What kind of voodoo magic is this?! I feel like putting that in a password at work to mess with people…
Whoever put up a website on contributing.md and agents.md, please reconsider 😕
Exactly what I'm suggesting. Putting all the eggs in the CONTRIBUTING.md basket like the article proposes wouldn't cut it, so keep the AGENTS.md around. Put more effort and a clear human touch into the former, focused on nurturing trust, and have the latter evolve only according to technical needs.
I understand that and definitely see merit in that approach! But when I look at what I ask of my agents to adhere to... Let's just say it would scare away any potential contributor 😅 On the other hand, having the agent use CONTRIBUTING.md written for humans would undercut its strict requirements.
That's an interesting way to go about it, but I feel like we can be a lot stricter on the demands we ask of an agent compared to a contributor. For example, my AGENTS.md file puts a lot of emphasis on using the latest Vue syntax, but I'm sure a casual contributor could be forgiven when they don't.
That IS insulting and precisely why these zealots shouldn’t be given the time of day. Even if you wanted to defend their fervent hate for AI on a superficial level, their witch hunt shenanigans target indiscriminately and hurt their cause and everyone at the whims of it even more. Block and move on!
That was fast! Looks good and minimal. Pretty close to how Drizzle and Zod work together, plus this opens up event handler body validation with the same validators for when you need more control over route handling. Top notch, thank you!
I looked into it and am loving what I see! Is there a good way to connect Zod validation to Convex fns/schemas? Or have Zod schemas as the original source definition? That would help a lot with sharing types for form states for example. First time touching Convex so I might be missing something too!
You nailed it in the conclusion! One thing I’d like to add that has been true for me is that LLMs have given me the opportunity to direct my efforts to the parts I love to work on myself (like front end) and rely more on AI for what I consider “boring” work like writing long, detailed SQL queries.
Thanks for this! It's inspired me to take this tech stack seriously in a side project I'll be tackling soon. Will be keeping an eye on any further insights and improvements you put into the repo further down the line!
Not a chance! It literally can't do anything without the power of open source libraries behind it to actually make use of.
I can’t even do the base drawing. Guess I’m out of luck 😭
I don’t think anyone is any less appreciative of your (or others’) hard work, trust me 🙂 Your target audience hasn’t changed (we’re still here), it’s just grown with the rise of AI. You’re under no obligation to welcome those of kind of contributors or consumers of your open source work, you do you!
Still, it would be much better to handle this on a human level, with PR templates that allow for full disclosure (or simply not accept) rather than trying to trick vibe coders with a workaround.
What’s to stop someone from deleting it before prompting and restoring it after?
Was literally just about to play around with posthog’s nuxt module, thanks for the heads up!
Yes, at that point just start automating PRs the moment an issue is created. Maybe one in X will be good (and that percentage could go up as the tech improves), but it will take vastly more time to check than to write it manually (for people like Daniel at least).
Exactly. The main issue isn't that AI is involved as an assisting technology, but the base line should be for a contributor to at least understand the code they deliver and be able to vet its accuracy and be able to react to feedback for possible changes. Still, brave and honest disclaimer!
I don't think anyone could be against a prompt that asks for information on how a code base works or is organized. The problem is that fully LLM-generated PRs don't really have an actual human contributor behind them. There is no OSS community aspect at play at that point and that is harmful.
Totally my GOTY!
That hasn’t been my experience, even without extra instructions. What do you feel are the shortcomings?
