notice on AI contributions

we basically have AI everywhere nowadays

it has an amazing sales pitch: "we 10x your productivity" (and they give it as free LLMs because they are very nice)

it's crazy the amount of reliance people are developing on chatbots that are just curated products

well no AI in this repository. this is not a web app you can hack your way around

we need to ensure that code in this repository does not cause undefined behavior and is maintainable, that means we can't accept things like:

• simple PRs adding hundreds of lines of code
• AI code slop for something that could be done much easier
• AI defining API boundaries and behavior
• AI making mistakes and no one really being responsible for it (because then it's our responsibility as maintainers)
• AI introducing technical debt

for the other case, AI documentation, that doesn't even begin to make sense

documentation is meant to be written by humans for humans so you understand what smallvec is about better

would this readme be better if AI had written it with perfect syntax and punctuation?? no, what matters is that you receive the point I'm trying to get across

AI isn't even substantially faster on that. it has taken me ten minutes to write everything above, is that an excessive amount of time?? no. I mean, if you were writing docs for four hours per day, I'll tell you to reconsider your approach, but this is normally not the case

but there's a reason why people use it. and many times I've seen it doesn't come out because "they are lazy" (though there are cases). many times it's just normal people that need validation about what they are doing:

• is my PR good enough??
• how should I write the PR description??
• etc etc etc

and there's nothing wrong with you needing it. approaching a codebase like smallvec isn't particularly easy. and it's also normal to want to perform, there's nothing wrong with that. but it simply is exactly what you don't want to do: make you look generic and that you have no idea, because you're offloading what you should be doing to a chatbot

if you're interested on exactly why AI is pretty terrible for you, here are some resources:

• https://www.youtube.com/watch?v=6sJ50Ybp44I amazing video about cognitive offloading
• https://www.media.mit.edu/projects/your-brain-on-chatgpt/overview/ what exactly happens to your productivity and output when you use LLMs to accomplish tasks
• read some books on how your dopamine control/reward system works and why you are a human being that needs confirmation of its actions being correct (nothing wrong with that). recommendations: "the molecule of more" by Daniel Lieberman, "dopamine nation" by Anna lembke, and "stolen focus" by Johann Hari

does this mean that you can simply do shitty PRs and barely put any effort but it is okay because it's you doing it and not an LLM?? fuck no, that doesn't make sense

it does mean that if you're new and you don't have much experience and you're not sure of what exactly to do or how to do it, it's okay to go and try and make the best PR you manually can and maybe it'll be shit and that's fine because you don't know, you are learning

it means:

• bad English is better than telling AI to write a description
• bad code is better than AI slop
• nonsense code that you genuinely thought was better actually isn't after review is better than AI making its own stuff

as a maintainer, I'm here to review code and make the codebase better. a part of it is also knowing that I'm not going to be here forever so of course I'm teaching people on the way with patience how to actually manage a codebase like this (and I'm not any genius, many of the guys at servo know way more than I do)

the reason why we as human beings tend to want validation and corrections from AI is literally the same mechanism that wants you to learn from someone who's better at a skill, except that people come to AI because it has infinite patience and is nonjudgmental and validating. except AI doesn't actually know shit. that's literally how humanity has worked for 180k+ years, and if we got this sort of validation and confirmation of our skills from elsewhere we wouldn't need to ask chatgpt at 1am whether this PR description is good and it telling you that you're never enough (that's the business model, it keeps you returning) and you should do it in some other way instead

so once this stops being a life priority for me there's someone that will replace my role and do it effectively and properly without me having to leave responsibility to a void and hope someone arises like happened to Matt