It Breaks a Village: Bevy’s 6th Birthday

摧毁一个社区：Bevy 的六周年生日

09 Sep, 2026 2026年9月9日

This post, like all my posts, was written by me. It came out of a lot of conversation and reflection with many people. It is long. I promise you there’s value in reading it in full, and with an open mind, and not merely summarising it or reading the headers alone. Opinions stated here are my own, not those of any person, organisation, or company I am associated with. Thank you to everyone in the Bevy community I’ve spoken to over this topic. (Last year: GiGF and Bevy’s 5th Birthday) 这篇文章和我所有的文章一样，均由我本人撰写。它源于我与许多人的大量交流与反思。文章很长。我向你保证，完整阅读并保持开放的心态，而不是仅仅阅读摘要或标题，是有价值的。文中表达的观点仅代表我个人，不代表我所关联的任何个人、组织或公司。感谢所有与我讨论过此话题的 Bevy 社区成员。（去年回顾：GiGF 与 Bevy 五周年生日）

Earlier this summer we had a string of positive things happen in the Bevy game engine. With the release of 0.19 the BSN macro finally merged. We have a declarative data model, UI is now tenable. Groundwork for actual arts tooling exists. This is a monumental achievement. 今年夏天早些时候，Bevy 游戏引擎发生了一系列积极的事情。随着 0.19 版本的发布，BSN 宏终于合并了。我们拥有了声明式数据模型，UI 现在变得可行了。实际美术工具的基础工作已经存在。这是一项里程碑式的成就。

Paired with that, Rustweek 2026 was the most positive experience I’ve had with techie people in real life, I got to meet many wonderful people and spend days and evenings having wonderful conversations. I did not have a single negative experience and everyone I met was lovely to talk to and spend time with, I would love to spend time with these people again. 与此同时，2026 年的 Rustweek 是我在现实生活中与技术人员交流的最积极的体验。我结识了许多很棒的人，度过了许多美好的日夜，进行了精彩的对话。我没有遇到任何负面经历，我遇到的每个人都非常可爱，我很乐意再次与他们共度时光。

Since then, things have been troublesome. I’ve been working on an overhaul of the Plugins API for a couple of months now. This aims to reframe the plugin system to make dependencies possible and future tooling more powerful. This design may or may not make it into the engine, but I am trying to argue that it is a good option. When I have an actual post about it, I’ll post it on my technical blog. My motivations for it are posted elsewhere, but I’ll be posting about it again soon. 从那以后，事情变得棘手起来。过去几个月里，我一直在致力于重构插件 API。这旨在重构插件系统，以实现依赖关系并增强未来工具的功能。这个设计是否会被引擎采纳尚不确定，但我试图论证这是一个不错的选择。当我写好相关文章时，我会将其发布在我的技术博客上。我对此的动机已在别处发布，但我很快会再次撰文说明。

This has been almost entirely on hold for the last 5 weeks as trying to write Bevy-related code right now makes me feel fully nauseous. Stomach churning, I-want-to-throw-up. My mind cannot keep from racing about what it means to be a contributor to open source in this moment in time. I may be alone in this specific feeling, but I am not alone in feeling that contributing to Bevy right now feels extremely bad. 过去 5 周，这项工作几乎完全停滞，因为现在尝试编写与 Bevy 相关的代码让我感到极度恶心。胃里翻江倒海，想吐。我的思绪无法停止思考：在当下这个时刻，作为一名开源贡献者意味着什么。也许我个人的这种感受是孤立的，但认为“现在为 Bevy 做贡献感觉非常糟糕”的人，绝不止我一个。

The “AI” Policy

“AI”政策

Note: There has been increased discussion over the last few days that may make it seem I am commenting on immediately recent things, the majority of my talking points here were first enumerated and drafted weeks ago. Bevy recently adopted a new “AI” policy that is (by stated intention) marginally more permissive than previously. On paper, it’s fine. The context of its development is where problems came up. 注：过去几天讨论的增加可能让人觉得我在评论最近发生的事情，但这里的大部分论点早在几周前就已经列出并起草了。Bevy 最近通过了一项新的“AI”政策，（按其声明的意图）比以前稍微宽松了一些。从纸面上看，这没问题。但其制定的背景才是问题所在。

The stated reasoning for this new policy is that we need to be “realistic” about people’s LLM use, and that the previous policy was “unenforceable.” Instances were cited where someone’s profile and Pull Request (PR) didn’t pass a sniff test but nothing could be proven for sure when the only evidence was a particular kind of code smell or a private / suspicious github profile. The burden of proof from this angle unfairly rested on an “accuser”, not “the accused”. This mirrors many justice systems, but a PR being rejected is not equivalent to imprisonment or a fine or death. It’s a mild social consequence, equivalent to saying “no thanks” to someone handing out flyers for a show at the Edinburgh fringe or leaflets of strange local religious sect. 这项新政策的理由是，我们需要对人们使用大语言模型（LLM）的情况保持“现实”，且之前的政策“无法执行”。有人举例称，某人的个人资料和拉取请求（PR）无法通过直觉测试，但当唯一的证据只是一种特定的代码异味或私密/可疑的 GitHub 个人资料时，无法证明任何事情。从这个角度来看，举证责任不公平地落在了“指控者”而非“被指控者”身上。这反映了许多司法系统的情况，但拒绝一个 PR 并不等同于监禁、罚款或死亡。这只是一种轻微的社会后果，相当于对在爱丁堡边缘艺术节发传单的人或散发奇怪地方宗教传单的人说“不用了，谢谢”。

This policy was mostly decided upon ahead of time amongst SMEs and other parties before public discussion opened up for a “debate” and editing period. During this period a lot of LLM enthusiasts started openly dreaming about even more permissive policies, what they would do once full agentic modes were allowed. Meanwhile, almost all arguments against “AI” use were more or less thrown out or ignored. Since this policy merged, the ratchet has been turning. This experience has been that of observing a knife get torn through the social muscles of the Bevy project. The most public fanfare around this new policy has been in the Bevy’s 6th Birthday post, if you don’t count the Discord. I feel complicated about writing about this out in the open, so publicly. But I do not feel like I have much of a choice if I want Bevy to feel like something I can continue to use and contribute to. 这项政策在公开讨论进入“辩论”和编辑阶段之前，大部分已在中小企业（SMEs）和其他相关方之间预先决定好了。在此期间，许多 LLM 爱好者开始公开畅想更宽松的政策，以及一旦允许完全代理模式（agentic modes）他们会做什么。与此同时，几乎所有反对使用“AI”的论点都被或多或少地抛弃或忽视了。自从这项政策合并以来，局势一直在恶化。这种体验就像是看着一把刀割裂了 Bevy 项目的社会肌理。除了 Discord 之外，围绕这项新政策最公开的宣传是在 Bevy 的六周年生日帖中。我对于如此公开地写这些内容感到心情复杂。但如果我希望 Bevy 依然是我可以继续使用和贡献的项目，我似乎别无选择。

Trust Is Broken and the Ground Is Uneven

信任已破裂，立场不平等

Community cohesion has fallen apart under pressure of persistent trolling and conflicting community values. The “AI” conversation has been positioned as two equal sides worthy of equal judgement. This is at odds with the context being a Game Engine as a creative tool people have sunk years of their lives learning, talking about, and promoting (myself included) in hope of rust having solid ground for artists / game developers. This context, vs the shallow buy-in of recent and more vocal LLM enthusiasts, was not acknowledged until late in the conversation. 在持续的骚扰和冲突的社区价值观压力下，社区凝聚力已经瓦解。“AI”对话被定位为两个值得同等评判的平等方。这与“游戏引擎作为一种创意工具”的背景相矛盾——人们投入了数年时间去学习、讨论和推广它（包括我自己），希望 Rust 能为艺术家/游戏开发者提供坚实的基础。这种背景与近期那些声音更大、但投入较浅的 LLM 爱好者的立场之间的差异，直到对话后期才被承认。

Acknowledgement did not mean material changes to the “AI” policy. The “unenforceable nature” of the previous LLM policy has read largely as a pretense rather than a valid reason for revision. The current LLM policy is, if anything, viewed as temporary or easy to bend, the assumption is that there will be a next, more permissive LLM policy, so why trust it? Why trust it when people are already disrespecting its limits? 承认并不意味着对“AI”政策进行实质性修改。之前 LLM 政策的“不可执行性”在很大程度上被解读为一种借口，而非修订的合理理由。当前的 LLM 政策如果说有什么意义，那就是被视为暂时的或容易变通的；人们假设未来会有更宽松的 LLM 政策，那么为什么要信任它呢？当人们已经不尊重其限制时，为什么要信任它呢？

Additionally, Bevy’s previous policy offered a safe space to people who otherwise have to deal with “AI” boosters, LLM enthusiasts, “AI” usage mandates etc. in their day jobs or personal relationships. This top-down cultural change has broken people’s trust on a deeper level than has been reckoned with. People have had their own positions viewed as merely a prefigure to “witch-hunts” and harassment campaigns. There isn’t trust that our positions are held in good faith, instead that we will merely lash out in existentially threatening ways that do not “address the reality” or “have real arguments.” This has extended to the point where merely asking if a PR is LLM Output gets push back, despite disclosure being a part of the current policy. 此外，Bevy 之前的政策为那些在日常工作或个人关系中不得不应对“AI”推崇者、LLM 爱好者、“AI”使用强制令等压力的人提供了一个安全空间。这种自上而下的文化变革在比预期更深的层面上破坏了人们的信任。人们的立场被视为“猎巫”和骚扰活动的预兆。人们不再相信我们的立场是出于善意，反而认为我们只会以威胁生存的方式进行反击，而不去“面对现实”或“提出真正的论点”。这种情况甚至发展到仅仅询问一个 PR 是否为 LLM 输出都会遭到抵制，尽管披露信息本身就是当前政策的一部分。

Backchannels Have Taken Root

幕后渠道已经扎根

This cuts both ways. Private communications on Bevy governance… 这在两方面都有影响。关于 Bevy 治理的私下沟通……