集里怎么说它

• 《从拒用 AI 到全面拥抱：Dioxus 团队的智能体编程实战课》(12:27起):本集说代码架构仍是一门艺术：智能体代码落地的基底不好、贡献也不会好，且智能体不怕随便大规模重构，所以团队大部分开发时间现在花在思考架构与系统演进上。
• 《智能体崩溃后还能原地复活？聊聊 Pi Durable》(04:01起):本集的核心论点：涉及设计和架构的问题必须人来，因为『设计架构的过程』几乎不以文本形式存在于任何训练数据中，这正是模型卡壳的地方。

① 提到它的金句

31 条

而且我认为智能体的主流架构就是一个带工具的 for 循环。
And I think the prevailing architecture for agents is a for loop with tools.
—— Ankur Goyal · [06:28]

指向原始笔记的链接

理论上最优的架构对任何人都没有用。
A theoretically optimal architecture is not useful to anyone.
—— 嘉宾 · [17:40]

指向原始笔记的链接

所以你可以在网络上构建你想要的任何东西,但网络有一个不同的问题,那就是在过去的 25 年云架构中,我们一直在错误的方向上奔跑。
So you can build whatever you want on the web, but there’s a different problem on the web, which is that for the past 25 years of cloud architecture, we’ve been running in the wrong direction.
—— Kenton Varda · [05:07]

指向原始笔记的链接

实际上几乎作为副产品就给了你有效的隐私保护的 evals,而不需要在架构旁边额外加上什么东西。
Actually give you effective privacy preserving evals almost as a byproduct without needing to kind of bolt something onto the side of your architecture.
—— Christopher Lovejoy · [15:42]

指向原始笔记的链接

架构跟随模型更新，而你的 evals 必须跟随你的架构。
And so, you know, architecture follows model updates and your evals have to follow your architecture.
—— Ameya Bhatawdekar · [04:41]

指向原始笔记的链接

而他们始终让最重要的事情保持最重要，Cursor，也就是改变你编写软件工程的方式。他们在根本上相信这是一个产品问题，不一定是模型架构问题。
And they always kept the main thing the main thing, Cursor, which is changing how you write software engineering. And they believed fundamentally it was a product problem, not necessarily a model architecture problem.
—— Martin Casado · [31:38]

指向原始笔记的链接

所以关键点是，你团队中每一个正在做架构决策的人，那些人必须懂性能，而且他们必须做出能让那些在他们下游的其他人使用以后可以优化的架构的决策。
the crucial takeaway is everybody on your team who is making architectural decisions, those people must know performance and they must make decisions that will allow the other people downstream of them to use an architecture which can be optimized later.
—— Casey Moratori · [35:57]

指向原始笔记的链接

我们发现,与其在我们的业务之上构建一个 AI 层,我们设计了架构,让智能体能够作为与人类在同一个系统中的另一个操作者来运作。
We found that instead of building an AI layer on top of our business, we designed our architecture so that the agent can operate as another operator within the same system as humans.
—— Flora Liu · [07:09]

指向原始笔记的链接

如果我去看我们在项目开始时写的 PRD 和架构图，如果我和我们现在拥有的这个架构相比，80% 都对不上。
If I look at the PRD and the architectural diagram we wrote in the beginning of the project, if I compare it to this architecture we have now, 80% of it doesn’t match.
—— Sait Izmit · [13:17]

指向原始笔记的链接

我们为此确定的最强大的架构，是为每个账户都存在一个智能体，并维护该账户的持久状态。
So the architecture that we’ve landed on for this that I think is most powerful is an agent that exists for each account and maintains a persistent state of that account.
—— Everett Berry · [11:19]

指向原始笔记的链接

现实是你不可能花一年去构建某种完美无缺的复杂系统架构，所以你必须把各个垂直解决方案拼凑起来，然后把它们粘在一起。
The reality is that you can’t spend a year going and building some really complicated system architecture that is perfect, so you have to like piece together the vertical solutions and then stick them together.
—— Arman Vaziri · [19:25]

指向原始笔记的链接

你不再是去担心性能优化和 API 兼容性,而是在代码评审中更多地去思考架构。
Instead of you worrying about the optimization of the performance and the API compatibility, you’re going to be thinking more about architecture in your code reviews.
—— Will Bond · [14:19]

指向原始笔记的链接

我认为，如果你能把你的初创公司架构设计成能与这样的公司合作，你基本上能够满足其他所有人。
I think if you can architect your startup to work with companies like this, with this kind of architecture, you’re probably going to be able to satisfy basically everybody else.
—— Brian Lewis · [17:43]

指向原始笔记的链接

我会担心任何这样的记忆架构：它不依赖成本更低的嵌入器和重排序器，而是依赖 LLM 的多遍处理来帮你分类并缩小数据语料库。
I would worry about any memory architecture that instead of relying on lower costs embedders and re-rankers is relying on multiple passes of the LLM to help you categorize and shrink the corpus of data
—— 嘉宾 · [66:35]

指向原始笔记的链接

并不清楚的是,用新架构和新范式预训练一个新模型的第一个周期,第一个周期就奏效是疯狂的。
It’s not clear that the first cycle of pre-training a new model with a new architecture and a new paradigm, the first cycle of that working is insane.
—— Justin Johnson · [22:32]

指向原始笔记的链接

在大多数组织中,昨晚有一条 Slack 对话说你应该做 A 而不是 B,而如果智能体先找到了某份架构记录,它永远不会发现那条信息。
In most organizations, there’s a Slack conversation from last night that says you should be doing A instead of doing B, and the agent will never find it if it found some architecture record first.
—— Brandon Waselnuk · [05:28]

指向原始笔记的链接

这就是我们的护城河——我们是唯一拥有这个架构的公司。
That’s our moat, the fact that we’re the only ones with this architecture.
—— Renen Hallak · [27:00]

指向原始笔记的链接

我们终于有了一个我们知道可以无限扩展的架构。
We finally had an architecture that we knew could scale infinitely.
—— Michael Lee · [02:52]

指向原始笔记的链接

在合适的地方,我们不是用 LLM 替换检索和排序系统,而是把推理和理解蒸馏进某种生产级排序架构。
Where it seems right, we’re not replacing the retrieval and ranking system with an LLM, but rather distilling the reasoning and understanding into some production ranking architecture.
—— Raghav Saboo · [09:06]

指向原始笔记的链接

在线的 LLM 调用往往不是你所需要的产品架构。
The online LLM call is often not the product architecture you need.
—— Raghav Saboo · [21:26]

指向原始笔记的链接

它们不擅长语义。我会说它们在架构方面糟糕得难以置信。对我来说,架构是软件中最具人类创造性的部分。
They’re bad at semantics. I would say incredibly bad at architecture. So like to me, architecture is like the most human creative part of software.
—— Diogo Almeida · [29:08]

指向原始笔记的链接

所以就原始的时间节省和这方面的总体产出而言，在更多使用这项技术的初级工程师与更容易发现幻觉、理解所做改动周边架构的 staff 工程师之间，差距相当平。
So in terms of like raw time savings and the aggregate output of this, it’s pretty flat between junior engineers who are using the tech more and staff engineers who have an easier time spotting hallucinations, understanding the architecture around the changes they’re making and things like that.
—— Justin Reock · [10:04]

指向原始笔记的链接

所以，也许正因为我们可以跳过架构这一步，我们才更不应该跳过，对吧？
And so maybe just because we can skip the architecture step, we shouldn’t, right?
—— Moritz Beller · [29:09]

指向原始笔记的链接

而你塞进 agents.md 文件里的仓库架构，说实话，这几乎没起什么作用。
The repo architecture that you’re plugging into your agents.md files, like let’s be real, that’s barely moving the needle here.
—— Dan Adler · [07:37]

指向原始笔记的链接

我是说，我把微服务描述成一种「最后手段的架构」。
I mean, I’ve described microservices as being an architecture of last resort.
—— Sam Newman · [29:27]

指向原始笔记的链接

任何涉及设计或架构的东西,你还是需要一个人类,因为智能体只会在你的代码库里局部地乱搞一通,把事情弄得更糟。
For anything that concerns the design or architecture, you still kind of need a human because the agent will just locally shit into your code base and make things worse.
—— Mario Zechner · [04:01]

指向原始笔记的链接

我确实认为它们在架构部分差了一点,因为它们在结果上更差了,因为你给了它们多得多的控制权。
I do think they’re a little bit worse in the architecture part because now they are worse in the outcome because you give them way more control.
—— Armin Ronacher · [08:19]

指向原始笔记的链接

那是编程中唯一通常没有以文本形式编码在某个地方、因而 LLM 无法从中学习的东西,也就是构建某个东西架构的过程。
That is the part, that’s the only part in programming that is actually not usually encoded somewhere in textual form so an LLM can learn from it, that the process of building the architecture of something.
—— Mario Zechner · [16:29]

指向原始笔记的链接

当我们审视这个问题时，我们意识到，在分布式系统和微服务架构中，API 就是你的上下文层。
And when we looked at this, what we realized is in a distributed systems and microservices architecture, APIs are the context layer for you.
—— Kamalakannan Nandagopal · [04:01]

指向原始笔记的链接

所以我认为这个结构、这个架构其实归结为一个问题，那就是：谁为智能体犯错买单？
And so I think that structure, that architecture really comes down to one question and that is who pays for the agent being wrong?
—— Abhi Arya · [02:35]

指向原始笔记的链接

它们太个人化了，能做太多不同的事情，所以你真的必须更多地去思考底层基础、系统、架构、已经就位的脚手架。
They’re so personal and they can like do so many different things that you really kind of have to think more about the underlying foundation and the system, the architecture, the scaffolding that’s in place.
—— Alex Cornell · [33:30]

指向原始笔记的链接

② 出现在这些集

2 集

• 《从拒用 AI 到全面拥抱：Dioxus 团队的智能体编程实战课》 — 作为概念
• 《智能体崩溃后还能原地复活？聊聊 Pi Durable》 — 作为概念

③ 关联

点进去有真内容 —— 本页主要出口

智能体 · Claude Code · Jonathan Kelley · Armin Ronacher · Dioxys · Mario Zechner · Cognition · Pi · Rust · PyDurable