It Breaks a Village: Bevy's 6th Birthday
90 points by aldantanneo
90 points by aldantanneo
The trolling is so persistent. It's everywhere. And if you speak up against it, you're often the one being kicked out if the moderators aren't paying close attention to agitators. This post really spoke to me about the current Online culture in general even if these statements are being made about AI boosters and impacts to the Bevy community. I've been really pleased with the Lobsters community in these trying times.
The comparison with the nazi bar is really apt. This makes me understand how hacker news is a true AI-bar now because every voice against AI is flagged/kicked out. Which also mean that newcomers think it is the norm and take for granted what they see on Hacker News about AI.
That's the point I preferred as well, and how pernicious the paradox of tolerance is.
I'd be curious if there are actual solutions to this except for immediate banning (as the post suggests). Were I in this situation I'd be tempted to do the same, and to preserve the community at the risk of excluding some. It's a hard problem IMHO.
Your community is defined by those wielding the banhammer. Period.
The solution is straightforward. You have an escalating banhammer that moves towards permanent very quickly upon repeat infractions. The initial "You're in timeout" generally gets people to cool off who just made an errant judgment, and they walk on eggshells a bit and reel it in until the penalty times out.
We have seen this over and over and over. Moderators are key otherwise your online community goes to shit. This tracks back the whole way to Usenet (and probably even earlier). And bad moderators can also make your community go to shit.
The main "problem" is that this is hard on the Moderators. It's a thankless job and takes a lot of time and a psychological toll.
The secondary problem is that the banhammer became an unusually severe penalty that many people were more reluctant to use because "Fake Internet Points" at some point became worth "Real Green Cash" and attracted miscreants of all forms. And that I don't have a good answer for.
That’s not true through. There are lots of AI Doomer threads on hackernews.
AI doomerism and AI skepticism are not the same thing. Several AI companies actively feed the doomers, presumably because it helps boost their valuations.
AI companies are directly using doom as marketing. Their CEOs constantly talk about replacing humans, about dangers of AGI which has totally been achieved internally, please keep investing, about using chatbots to babysit your kids
I really feel this post. I used to browse certain Discord channels (pl dev, raylib) almost daily to look at new projects and follow the progress of people. Now both contain lots of AI projects and even my internal filtering of what's AI and what's not annoys me to such a degree that I simply stay away from those channels now.
It has also made me reluctant to post my own projects because what's the point now if people just show off their flashy LLM output and are not interested in the process of creating something.
In software we have this strange position taking hold, that our work is as valueless as the output of an LLM.
Unfortunately it was always like this. This was reinforced by the industry every step of the way.
Human technical knowledge in programming was never acknowledged as anything more than a tool to produce some value. We've all spent countless unpaid hours honing that tool by reading, studying and experimenting, for no real reward other than personal gratification. And when that honed tool went against the roadmap, advocated for quality or even ethics, it was disregarded.
A substitute tool now exists in the form of LLMs, our work is valued exactly the same as their output because that was always the only thing that was valued. The only ones who care about human produced code, the art of it or its quality, are a small minority of other programmers.
The PR that updated Bevy's AI policy spans 41 commits and attracted 253 comments from 39 unique commenters.
I think that's indicative of a strong community process.
The rationale in the appendix is worth a look:
If you are new to the Bevy community, you might ask: why not simply ban AI contributions and be done with it?
In brief: we tried that. Motivated by concerns around quality, ethics and legal risk, the Bevy maintainers voted to ban all AI-generated contributions. It didn't go very well.
In practice, we found that we could not fairly and reliably enforce or adjudicate this policy. [...]
Over time, we found that what we were punishing was disclosure, not use. Without reliable detection, there was a strong incentive to defect: silently use AI, don't tell anyone about it. Even if no one took it, the temptation weighed on contributors, and those who felt strongly that Bevy should not allow AI-assisted contributions worried about what might be slipping by.
You are doing the exact thing the article is talking about. You are perhaps the biggest AI booster I've ever seen posting non-stop about it on every platform you can; it's exhausting.
Why is linking to the PR seen as AI boosting?
I don't want to make this overly personal because it's not just Simon. There are a handful of people who comment on every AI-related post, making vague positive statements like that there's a "strong community process" behind becoming more AI permissive here, or that OpenAI seems to "care about privacy" in the thread about them potentially scooping a mathematician with his own usage, etc. There's an obvious pattern of frankly glazing the pro-AI stance and being extremely charitable to the companies behind the products, and it often adds little or nothing to actual discussion. It's tiresome.
In this case I don't think the commenter has any connection to game dev or bevy - they're just here for the AI mention. And I think there's some irony in particular when this is done on a post complaining about the incessant AI boosting:
The dynamic between "AI" boosters and people who don't want to use LLMs is not on level ground. "AI" boosting in a space with people who want to be "AI Sober" is disruptive and antagonistic, knowing that you're doing it is baiting or outright trolling.
Have you heard of manufacturing consent? Engagement on a PR does not equate to agreement from the community, but linking to it paints it that way. Considering my own lived experiences, the PR existing at all, from Alice especially, is just another nail in the coffin that causes me to disengage entirely because there is a momentum here to push LLMs into every corner of our lives that the slop is inescapable even in my private life.
If I read that well, the PR is about making the anti-AI more sustainable and enforceable, not about being more lenient on AI.
In fact, being anti-AI myself, I saw this blog post as quite good but I may have missed something.
It's interesting that the pr links to a 2025 (ancient) paper about identifying LLM code. There's are modern systems available and soon the fingerprinting will start too. https://www.eyesift.com/faq/ai-code-detection-2026-github-copilot-claude-gpt-codex-detector-accuracy/ I hope they'll revisit the idea.
Honestly I do feel for the sentiment, but this - in a game engine community - is just gatekeeping.
Most people want to build games because they care about the result. LLMs make that possible for them without Rust skills. That is not really what most members of the Bevy community care about, since by selection they choose to work on Bevy and not use a ready-made engine.
I think claiming you didn't use AI while using AI is ban-able offense, and contributing to a codebase must follow the guidelines. But otherwise - you don't have to engage with the AI if you don't want to.
Gatekeeping is controlling or limiting access to information or resources. Nobody is stopping anyone else learning Rust in order to make a game using Bevy. This is not gatekeeping.
Yeah except the post's vibe is "we should kick out of our community all the people who are pro LLMs" which is gatekeeping with some complaining on the side.
Vibe? Isn't it pretty explicitly the point of the nazi bar section..?
Still not gatekeeping, anyway; nothing stopping them making their own community.
(...except maybe the pervasive sense of entitlement to the fruits of others' efforts that drew them into the community in the first place)
the pervasive sense of entitlement to the fruits of others' efforts
Funny how that describes both LLMs and their users.
What’s wrong with gatekeeping? You just need to do it for the good reason (like, yeah, gatekeeping nazi out)
I never said there was anything wrong with it. I raised issue with it because it's too often invoked incorrectly to play the victim. (Kicking the nazi out would be ostracism.)
But otherwise - you don't have to engage with the AI if you don't want to.
I'm not sure how you can write that with a straight face. As developers, we're all being subjected to LLM outputs to the point of inundation.
Well, yes - but you can choose not to engage: there's a "vibecoding" tab here, you can have a showcase-vibecoded channel or something, and I do think that LLM usage needs to be acknowledged and marked and that should be the policy.
I also feel "It's everywhere" and the nauseating repetitiveness of LLM/vibecoding projects and content, and I also understand the need to have a personal zen garden that's not that, but that doesn't mean people building their games with LLMs are destroying a community.
according to the OP the bevy discord mods literally removed the ability to add robot reaction icons on AI projects
but that doesn't mean people building their games with LLMs are destroying a community.
But this is only partly about individual developers making AI games with the #showcase mention, which just get ignored on release anyway because the general public hates AI. It's primarily about the AI policies around the games engine and that engine's current governance.
But the policy starts with
No AI-generated media. .. No AI-generated prose. .. No AI-generated communication.
so what would you have instead?
Honestly I do feel for the sentiment, but this - in a game engine community - is just gatekeeping. Most people want to build games because they care about the result. LLMs make that possible for them without Rust skills.
LLMs make that possible even without a community, so how about they skip not only the learning but also the human part of their result-focused project (such as: a community)?
LLM use still allows for both: learning game dev aspects while building & building a community around the thing you built instead of the tool.
So you're interested in building "the thing" and a community "around the thing". Fair enough.
Why should the tool you're using or rather its community be interested in accommodating that when you already state that your goal is to build a community for "the thing you built instead of the tool"?
Why should the tool [...] be interested in accommodating that
It does or doesn't have to. Not arguing for either, just responding to your comment in isolation about how they dont need to skip all of that.
They don't need to skip on that, but if the community doesn't want to be pestered by AI cruft, why bother them and be all whiny about how there's "gatekeeping" going on?
The code is open. The LLM hands you a result. You don't need that community to get there. You don't care about that community (only about one for your own product).
Leave the people alone.
LLMs make that possible for them without Rust skills.
I think this sentiment would feel much less disingenuous if Godot and GameMaker and Löve2D and Unity and a dozen other projects didn't already exist, serving people who do not know Rust.