跨国深谈

上下文 (context)

30分钟阅读

跨国深谈← 返回←上下文 (context)
上下

上下文 (context)

概念
本站收录 33 集 · 109 条金句 · 关联 10 个

集里怎么说它

① 提到它的金句

109 条

Spotify 的 Tobi 杜撰了这个术语叫”上下文工程”，也就是把上下文传递给模型，在正确的时间传递正确的上下文，这至少占性能的一半。
Tobi from Spotify coined this term called “context engineering,” which is getting the context to the model, the right context at the right time, is at least half the performance.
—— Dan Shipper · [23:32]

指向原始笔记的链接

护城河是关于上下文和记忆的。
the moat is about context and memory.
—— Brian Balfour · [30:40]

指向原始笔记的链接

这些模型本身，如果你把它们并排比较，它们产生的结果是一样的，所以真正的决胜点是谁拥有更多的你的上下文，因为是上下文加上模型产生了最好的输出
these models by themselves, if you compare them side by side, they generate the same result, and so the actual difference-maker is which one has more of your context, because it’s the context plus the model that produces the best output
—— Brian Balfour · [30:44]

指向原始笔记的链接

我相信每个方向都有无限。所以这让我对几乎任何人说的话都相当持相反意见。所以如果有人像在 Twitter 上说什么，我有时和自己玩这个游戏，就是在什么语境下那实际上不成立？
I believe that there’s infinity in every direction. So that makes me pretty contrary on pretty much everything that anyone says. So if someone says something like on Twitter, I sometimes play this game with myself, which is in what context would that actually not be true?
—— Julie Zhuo · [80:20]

指向原始笔记的链接

那种上下文无法进入人脑，但它可以进入 AI 的上下文窗口。
That context can’t go into a human brain, but it can go into the AI context window.
—— Evan Reiser · [08:04]

指向原始笔记的链接

当我们试图请求一个 LLM 做这个错误分析时，我们通常发现它只是说追踪看起来很好，因为它没有理解某样东西是否可能是坏的产品味道所需的上下文。
What we usually find when we try to ask an LLM to do this error analysis is it just says the trace looks good because it doesn’t have the context needed to understand whether something might be bad product smell or not.
—— Shreya Shankar · [24:09]

指向原始笔记的链接

再说一次，只要把自己剔除并转移上下文，你就解决了今天 99% 的 AI 问题。
Again, just eliminate yourself and move the context, you solve 99% of the problems with AI today.
—— Lazar Jovanovic · [75:21]

指向原始笔记的链接

这是我认为大多数 PM 投入不足的层,正是这一层让 Claude 从一个通用的聊天机器人变成真正了解你的上下文
this is the layer that i think most pms under invest in it’s this layer that makes claude go from being a generic chatbot to actually knowing your context
—— Jyothi Nookula · [06:05]

指向原始笔记的链接

就像如果我只是让它生成，那将是垃圾。如果我把我自己的上下文给它，像我的策展、我的品味、我的 harness，它就会说，嗯，不要供应商之类的东西，或者它必须有四个声音。
Like if I would just have it generate, it will be rubbish. If I had like my own context to it, like my curation, my taste, my harness, it would be say, well, not vendors or kind of like things or like it has to have four voices.
—— Patrick Debois · [12:35]

指向原始笔记的链接

这是非常经典的苦涩教训药丸，但模型可以推理它们自己的内存和上下文结构，比你可以为它们规定一种构建自己内存的方式要好得多。
This is very classically bitter lesson pill, but models can reason about their own memory and context structure much better than you can prescribe for them a way to structure their own memories.
—— Lance Martin · [23:50]

指向原始笔记的链接

另一个幸运的意外是你获得了未受污染的上下文,所以每个智能体都专精于其工作并自然改进。
And other happy accidents are that you get unpolluted context, so each agent becomes specialized in its job and naturally improves.
—— 嘉宾 · [19:00]

指向原始笔记的链接

你的用户很可能比你更擅长编写上下文。
Your user is better at writing context than you are, most likely.
—— 嘉宾 · [19:36]

指向原始笔记的链接

这里创造的价值太多了，以至于在那个场景下值得 token 最大化。
There’s so much value created that it’s worth token maxing in that context.
—— 嘉宾 · [59:58]

指向原始笔记的链接

文件系统是……可能是目前上下文管理的最先进技术,智能体将利用文件系统来存储它们可能需要参考的信息,比如便签本。
The file system is… Probably the state of the art when it comes to context management today, where agents will utilize the file system to store information that they might need to reference, such as a scratch pad.
—— 嘉宾 · [03:29]

指向原始笔记的链接

对我们来说发生的事情实际上是，一旦我们删除了整个上下文，eval 开始表现得更好，这是反直觉的，但就是这样，是的，那个文件很大，列出的东西不一定有用，或者不应该进入指导文件。
And what happened for us is actually like the eval started to perform like much better once we deleted the whole context, which was counter uh intuitive, but it’s just like, yeah, that file was pretty big and was listing stuff that was not like necessarily useful or you know, that should not go into like a steering document.
—— Simon Boudrien · [38:11]

指向原始笔记的链接

但杠杆不在于权重。而在于你给它什么样的上下文，它有多相关，以及它是否发生在正确的步骤。
But the leverage is not in the weights. It’s in what context you give it, how relevant it is, and does it happen at the right step.
—— Garry Tan · [10:15]

指向原始笔记的链接

模型变得越好，差异化因素就越转移到上下文。
The better the models get, the more the differentiator moves to context.
—— Garry Tan · [33:51]

指向原始笔记的链接

上下文溢出到主上下文，因为上下文意味着 tokens，tokens 意味着钱，而且你有越多的上下文，LLM 在给你答案时就会越困惑。
Context spill over into the main context because context means tokens, tokens mean money, and the more context you have, the more confused the LLM is gonna be in giving you an answer.
—— Frank Coyle · [13:11]

指向原始笔记的链接

生产上下文重要得多，因为去检查仪表盘是一回事。说那个指标看起来不对劲是另一回事。
The production context is just way more important because it’s one thing to go check a dashboard. It’s another thing to say that metric smells off.
—— Justin Smith · [10:38]

指向原始笔记的链接

harness 的主要工作是在正确的时机把上下文带给模型。
The main job of the harness is to bring context to the model at the right point in time.
—— Harrison Chase · [02:07]

指向原始笔记的链接

这导致了一些非常令人惊讶的行为，因为实际上你离能够访问 10,000 个上下文桶只有不到两次 LLM 调用的距离，其中每一个都包含 200,000 个 token。
And that leads to some really surprising behaviors because you’re literally no more than two LLM calls away from being able to access 10,000 context buckets each one of which contains 200,000 tokens.
—— Flo Crivello · [32:49]

指向原始笔记的链接

智能不是差距所在。缺的是上下文。
Intelligence is not the gap. It’s context that’s missing.
—— Dennis Pilarinos · [09:58]

指向原始笔记的链接

我认为人们已经了解到，无论 AI 实验室在这些东西周围设置了什么护栏，你都不能依赖模型来阻止自己或理解上下文。
And I think what people have learned is that regardless of guardrails that the AI labs are putting around these things, you can’t rely on models to stop themselves or to understand context.
—— Nick Warner · [07:16]

指向原始笔记的链接

默认情况下,上下文是在游戏引擎上,而不是在游戏设计这部分,而我认为我们应该把这个理念翻转过来。
By default, the context is on the game engine rather than on the game design part, and I think we should flip that idea.
—— Arturo Nunez · [07:42]

指向原始笔记的链接

而是要想：这个前沿模型需要什么样的全部上下文，才能像我一样做决策。
But rather what is all the context this frontier model needs to be able to make decisions like I would make.
—— Hursh Agrawal · [07:29]

指向原始笔记的链接

如果你搭建了基础设施，如果你收集了上下文，如果你开始积累代码审查如何为你自动化的经验，那才是你能获得 10 倍速度的地方，也就是你向你的 CEO 或你自己或你的开发者承诺过的速度，否则它就是一个瓶颈。
If you put the infrastructure, if you gather the context, if you start accumulating how code review can be automated for you, then that’s where you’re gonna get the 10x velocity that you’re being promising your CEO or yourself or your developers, because otherwise it’s a bottleneck.
—— Itamar Friedman · [15:59]

指向原始笔记的链接

智能体提供结构和上下文这个关键洞察对我们来说非常重要，因为它意味着任何人都可以参与把他们的意图转化为实现。
That key insight of agents providing structure and context was a really big thing for us because it meant that anyone could participate in translating their intent into implementation.
—— Safia Abdalla · [13:38]

指向原始笔记的链接

在会议上我们谈论事情，但会议记录在之后才到；它们捕获了摘要。你仍然没有关于为什么做出决策的上下文。
In meetings we talk about stuff, but the meeting notes come after; they capture a summary. You still don’t have the context of why the decision was made.
—— Elaina O’Mahoney · [08:06]

指向原始笔记的链接

我们拒绝外包上下文层,因为那是我们的优势所在。
We refuse to outsource the context layer because that’s where our edge is.
—— Flora Liu · [18:02]

指向原始笔记的链接

而在生命周期营销方面,收到基于上下文的推荐的用户,采取下一步行动的可能性高出 63%。
And on the lifecycle marketing side, users who received context-aware recommendations were 63% more likely to take the next step.
—— Flora Liu · [19:06]

指向原始笔记的链接

我认为我们用数据想要实现的核心目标，是创建一个市场的完美虚拟副本——你正在瞄准的理想客户、客户公司和上下文。
And the core goal that I think we’re trying to accomplish with our data is to create a perfect virtual copy of the market, the ideal customers, the accounts, the context that you’re going after.
—— Everett Berry · [02:38]

指向原始笔记的链接

这就是为什么你的智能体需要一个上下文引擎才能完成真正的工作。
And that’s why your agents need a context engine to get real work done.
—— Peter Werry · [06:59]

指向原始笔记的链接

循环会累积。而你必须在整个过程中都让上下文的使用保持高效。
The loops compound. And you have to be efficient the entire way through with your context.
—— Peter Werry · [12:25]

指向原始笔记的链接

token 减少 50%,分类更快，答案更好。而这正是上下文引擎能做到的。
50% fewer tokens, faster triage, better answers. And that’s exactly what a context engine can do.
—— Peter Werry · [17:45]

指向原始笔记的链接

我们最初的尝试是把 Base64 数据直接传进代码，那是个非常糟糕的主意。它直接把上下文窗口撑爆了，方方面面都很糟。别那么做。
Our first attempt was just passing Base64 data into the code, and that was just a terrible idea. It just blew up the context window and was bad all around. Don’t do that.
—— Jesse Lumarie · [05:16]

指向原始笔记的链接

上下文窗口越来越大，但上下文永远不够。
Context windows are growing bigger, but there’s never enough context.
—— Roberto Milev · [05:20]

指向原始笔记的链接

主动提示 Claude 说，比如，我不明白你在这里说什么。我们能一起定义它吗？然后我会把它保存到我的上下文中，我就会知道你在说什么。从此往后真的非常犀利，也是我们在播客中还没见过的东西。
Actually proactively prompting Claude to say, like, I don’t understand what you’re talking about here. Can we define it together? And then I’ll save it to my context and I’ll know what you’re talking about. Moving forward is really, really sharp and something we haven’t seen on the podcast yet.
—— Claire Vo · [18:14]

指向原始笔记的链接

在不理解纹理和上下文的情况下创建技术红线的想法，不是我们认为可以强加给客户的界限。
creating a technological red line without understanding the texture and context is not a boundary line that we think that we can assert on top of our customer.
—— Nick Noone · [40:02]

指向原始笔记的链接

所以从上下文的角度来说,这里的愿景是:你应该永远不必操心上下文。
And so I would say from a context perspective, the vision here is that you should never have to worry about context.
—— Roshan · [48:14]

指向原始笔记的链接

随着时间推移，这实际上会让你花钱，因为每一轮都在喂入所有这些来自失败尝试和出错轮次的额外上下文。
That is actually going to cost you money over time because every turn is feeding all this extra context from failed attempts, turns gone wrong.
—— Santi Garza · [28:12]

指向原始笔记的链接

这是 Cursor 内部的一个巨大差异化优势——多仓库云端智能体，因为现实是，除非你有 monorepo 或者在小仓库里工作，你很可能确实希望一个仓库的上下文进入另一个仓库。
So this is a huge differentiator within Cursor, multi-repo cloud agents, because the reality is, unless you have a monorepo or unless you’re working in smaller repos, you probably do want context from one repo going into another.
—— Amrita · [30:16]

指向原始笔记的链接

而且我几乎认为，对这类系统来说，质量和数量是分不开的，因为上下文就是一切。
And I almost think you can’t separate quality and quantity when it comes to these systems because context is everything.
—— Sharadh Krishnamurthy · [41:38]

指向原始笔记的链接

管理上下文的能力——不只是在两小时的会话里，而是跨越数周或数月的工作——我认为这将是关键的，而且需要比今天长得多的上下文推理能力。
The ability to manage context, not just through a two-hour session, but across weeks or months of work is something that I think will be critical and will also require much longer context reasoning than we have today.
—— Alexander Whedon · [32:19]

指向原始笔记的链接

为此，我们真的希望人们觉得输入 token 是免费的——只管为你的问题考虑所需要的上下文就好。
To make it, we really want people to feel like the input tokens are free, like just consider the context that you need to for your problem.
—— Alexander Whedon · [45:32]

指向原始笔记的链接

你仍然在做那件事，感觉也是那样，它同样压在你身上——而不是真正地给同事来一记不看人的传球，然后说：「你可以搞定，这是背景信息，放手去干吧。」
You’re still doing the thing and it feels that way and it’s weighing on you in the same way versus truly throwing a no look pass to a colleague and being like, “You got this, here’s the context, go off and run.”
—— Roman Ugarte · [62:59]

指向原始笔记的链接

如果超过 30 或 40 个工具,我们就无法在不降低上下文质量或系统性能的情况下扩展。
We cannot scale it beyond 30 or 40 tools without degrading the context or degrading the performance of the system.
—— Ajay Prakash · [17:23]

指向原始笔记的链接

差距不再是智能了,是上下文。
The gap is not intelligence any longer. It’s context.
—— Brandon Waselnuk · [13:11]

指向原始笔记的链接

确保这些模型能访问全部上下文,因为它们会找到你的未知的未知。
Making sure that these models have access to all of the context because they will find your unknown unknowns.
—— Brandon Waselnuk · [08:22]

指向原始笔记的链接

那个上下文必须回到大脑里，这样你不仅在这次修好了它，而且在今后所有情况下都修好了它。
That context has to go back into the brain so that you not only fix it in this case, but in all future cases going forward.
—— Brian Armstrong · [13:57]

指向原始笔记的链接

这其实与模型的强大程度无关。Databricks 的关键，甚至很多公司想在 AI 上取得成功的关键，全都在于你数据的上下文。
It’s not really about the strength of the model. The key for Databricks and even the key for a lot of these companies to be successful at AI is all about the context of your data.
—— Ron Gabrisko · [19:57]

指向原始笔记的链接

从那以后,那套东西几乎全被扔出了窗外,现在流行的是提示词工程、上下文工程、循环工程。
Almost all of that has gone out the window since prompt engineering and context engineering, loop engineering.
—— Robert Lucero · [25:19]

指向原始笔记的链接

所以我们基本搁置了「存在一个巨大的、宽广的、平面的上下文，我们所有人都在其中游泳」的想法。它更像是一个上下文层级结构。
So we sort of set aside the idea that there is one big, broad, flat set of context that we’re all going to swim in. And it’s much more like a context hierarchy.
—— Charles · [41:01]

指向原始笔记的链接

这是一个演示文稿产品，但它永远做不出惊艳的演示文稿。它缺上下文，缺数据结构。
This is a presentations product and it’s never going to be capable of making amazing presentations. It was just missing the context. It was missing the data structure.
—— Keith Peiris · [04:14]

指向原始笔记的链接

现在,第三波会越来越走向持久化和异步的实体,也就是智能体成为我们基础设施中长期运行的进程,可以访问工具、组织中其他的智能体以及上下文。
Now, the third wave will be going more and more towards persistent and asynchronous entities, so agents being long-running processes in our infrastructure with access to tools and other agents around the organization and context.
—— Giselle van Dongen · [01:00]

指向原始笔记的链接

我们必须把工具结果推回主智能体的上下文里，让它以为自己做了工具调用，但其实没有。
So, we had to push the tools back into the context of the main agent so that it thinks it made the tool call, but it really didn’t.
—— Bohan Li · [04:08]

指向原始笔记的链接

所以更多上下文、更多音频会胜过旧的准确结果。
So more context, more audio is going to beat the old accurate one.
—— Bohan Li · [03:15]

指向原始笔记的链接

具体来说,你应该重建上下文,而不是对抗漂移。
Specifically, you should rebuild context and not fight drift.
—— Paula Dozsa · [08:15]

指向原始笔记的链接

但在易变的语音对话里,这是个陷阱,因为一旦用户转移话题,你复用的上下文就立刻错了。
But in a volatile voice conversation, it’s a trap because the second the user pivots, your reuse context is actively wrong.
—— Paula Dozsa · [08:25]

指向原始笔记的链接

构建这套系统的最大教训是,智能体的上下文主要来自代码库本身,而不是那么来自 Claude MD 文件。
And the biggest lesson in building that system is that an agent’s context comes mostly from the code base itself, not so much from the Cloud MD file.
—— Paula Dozsa · [10:33]

指向原始笔记的链接

最大的问题是，模型根本没有足够的上下文去真正理解演讲者、听众、以及两者之间的关系，而再多的通用推理也无法让你跨过这道坎。
But I think the biggest issue was that the model just didn’t have enough context to really understand the presenter, the audience, the relationship between the presenter and the audience and no amount of general reasoning gets you past that.
—— Keith Parris · [04:52]

指向原始笔记的链接

智能体可以帮你深挖这些问题，帮你找到可能驱动客户那些感知的指标，但通常不太擅长真正理解这项工作的端到端上下文。
And so, you know, agents can help you dig into those and can help you find um uh find the uh the the metrics that might be driving those perceptions from customers or driving those anecdotes, um, but generally aren’t as good at really understanding the end-to-end context of the work.
—— Marc Brooker · [19:28]

指向原始笔记的链接

所以区别也许在于：如果你把 Claude 当扩写器用——「这是一个提示词，把它变成一份六页的文档」——我认为那是在浪费读者的时间；而如果你把 Claude 当上下文提供者用——比如「加一段本周业务指标」——我觉得那没问题。
And so maybe the distinction there is if you’re using Claude as an expander, right, saying, here’s a prompt, turn it into a six-page document, I think that’s a waste of the reader’s time. However, if you’re using Claude as a context provider, right, like add a paragraph with the business metrics for this week, I think that’s fine.
—— Marc Brooker · [45:09]

指向原始笔记的链接

但有意思的是,它的准确性也差得多,因为再说一次,他们没有你公司里正确的上下文、正确的目标等等。
But interestingly enough, it’s also a hell of a lot less accurate because, again, they don’t have the right context, the right objectives and so on in your company.
—— Edo Liberty · [18:10]

指向原始笔记的链接

但我认为总体而言，大多数人都远远高估了需要这样做的场景数量，而且绝对低估了当人类和智能体都能访问公司正在做的一切事情的完整上下文时所产生的力量。
But I think by and large, most people far overestimate the number of things where that is required and definitely underestimate the power of what happens when both humans and agents have access to the full context of what a company is working on.
—— Wade Foster · [52:05]

指向原始笔记的链接

如果你能影响或在那些 token 里留下印记，那价值比出现在单个上下文窗口里高出数量级。
if you can get influence or imprint in those tokens, it’s worth, like, orders of magnitude more than being in a in a in single context window.
—— Ben Moore · [52:00]

指向原始笔记的链接

只要在服务上有 50 个用户，仅仅是这些单独的用户会话，就会超过你试图存储的模型权重
with just 50 users on your service, the user context, just those individual sessions end up being greater than the model weights that you’re trying to store.
—— Thomas Sohmers · [40:27]

指向原始笔记的链接

只要我们还在使用基于 transformer 的 LLM，这就是它们工作原理的数学本质，使用更少的上下文你就会得到更好的质量。
As long as we’re using transformer-based LLMs, that’s just like the math of how they work, and you will get better quality if you use less context.
—— Dex Horthy · [52:08]

指向原始笔记的链接

我们把文档视为非结构化上下文的通用容器。
We see documents as universal containers for unstructured context.
—— Jerry Liu · [06:31]

指向原始笔记的链接

我们做的很多工作和获得的经验，都处在一个更大的主题背景下：AI 应用的范畴最近发生了很大变化，从单纯的信息综合产品变成了真正在干活的智能体。
And I think a lot of the work that we’ve done and a lot of the learnings we’ve had are in this sort of broader thematic context of the scope of AI applications has changed a lot recently from just like information synthesis products to actual agents doing work.
—— Adit Abraham · [01:59]

指向原始笔记的链接

头号问题实际上就是给智能体提供它们所需的上下文:收集数据、打包数据,然后以一种 LLM 可以推理的方式呈现出来。
The number one problem is actually just giving the agents the context they need, gathering the data, packaging that data, and then presenting it in a way that the LLN can reason over.
—— Cody Menefee · [17:15]

指向原始笔记的链接

它找到目标后,就能访问让无人机跟踪跟随的工具——这不需要对规则做任何专门编码,而是用更智能体的方式:给智能体它需要的所有工具,让它根据无人机状态和可用的信息与上下文自己做决策。
It finds something, then it has access to the tools that allow the drone to track and follow, and that’s without any specific coding of that rules, but instead having a more sort of agentic, giving the agents all the tools that it needs to be able to understand the drone state and make decisions given the information and the context that’s available there.
—— Suchet Bargoti · [17:21]

指向原始笔记的链接

最优做法可能是把一个新的状态-动作-奖励放回上下文里，因为这非常快，但之后就需要某种类似睡一会儿的东西，几乎所有有智能的东西都会睡觉，告诉我一个不以某种方式睡觉的智能系统，对吧，然后……
The optimal thing to do may be to put a new SAR state action reward back into context because it’s very quick, but then there needs to be some like of go to sleep for a while, like almost everything that is intelligent sleeps, like tell me an intelligent system that doesn’t sleep, right, in some way, and then…
—— Francois · [27:46]

指向原始笔记的链接

我们实际上发现，这样做最终会有更新鲜的上下文，算是「头脑清醒」——就像人类有其他同事来验证他们的代码一样。
We actually found that if you do this, you end up with more fresh context and kind of fresh heads, same with when with humans have like other colleagues verifying their code.
—— Tereza Tížková · [13:04]

指向原始笔记的链接

我们实际上就是一个上下文引擎，或者说法务部门的 CRM。
we’re just a context engine or a CRM for the legal department.
—— Nick Fleisher · [11:10]

指向原始笔记的链接

但说实话，除了数学之外，几乎所有有趣的事情都是这种类型，即完整上下文永远对你不可用。
But to be honest, outside of math, almost everything that’s interesting is of this type, where the full context will never be available to you.
—— Liam Fedus · [50:36]

指向原始笔记的链接

让生活有趣的是，你没有全部上下文，但你必须在不确定性下做决策。
What makes life interesting is when you don’t have all the context, but you have to make decisions under uncertainty.
—— Liam Fedus · [50:48]

指向原始笔记的链接

而如果你保留这个上下文，也就是一大堆不同失败模式的运行日志，它们很可能会过度约束 Claude。
And if you keep this context where it’s running log of a bunch of different failure modes, they will probably over-constrain Claude.
—— Thariq Shihipar · [00:59]

指向原始笔记的链接

MCP 就已经用 100 个工具占满了你的上下文。可能你三分之一的上下文就这么没了，却什么工作都没做。
MCP would already eat 100 tools in your context. Maybe one third of your context would be gone without actually doing any work.
—— Jan Čurn · [03:00]

指向原始笔记的链接

如果你很少用到那 100 个工具，为什么要一直把它们全部放进上下文呢？
Why would you add all the 100 tools in your context all the time if you rarely need them?
—— Jan Čurn · [05:17]

指向原始笔记的链接

模型是 CPU,上下文窗口是内存(RAM),而记忆是磁盘。
The model is the CPU, the context window is the RAM, and the memory is the disk.
—— Dylan Couzon · [06:46]

指向原始笔记的链接

我们用 skills 来打补丁，但 skills 只是一张创可贴，因为你提供的上下文越多，智能体反而变得越笨。
And we patch this with skills, but skills is just a band-aid because agents become dumber with the more context you provide.
—— Sarah Simionescu · [04:01]

指向原始笔记的链接

那么有可能你的智能体用来回答问题的实际上下文中有高达 80% 被过时信息填满。
It can happen that up to 80% of your actual context your agent uses to answer a question is filled up with stale information.
—— Leo Platzer · [17:02]

指向原始笔记的链接

防御者拥有更好的上下文。
The defenders have better context.
—— Alon Schindel · [17:58]

指向原始笔记的链接

所以在每个单独的时间点，我们都不会出现既能读取疯狂未过滤的上下文、又能用我们的工具做事情的危险组合。
So at each individual point, we don’t have a dangerous combination of both being able to read in crazy unfiltered context and do things with our tools.
—— Jim Clark · [08:26]

指向原始笔记的链接

现在，当你在其上叠加时间维度的讨论时，你的 AI 系统应该在第 100 次调查中仍保有之前那次调查的上下文，否则你就是在再次从零开始转轮子。
Now, when you add temporal discussions on top of it, your AI system should have the same context of a previous investigation in the 100th investigation, or else you’re starting the wheel from the same time again.
—— Varun Krovvidi · [11:30]

指向原始笔记的链接

但作为目标真正重要的是，AI 解决那个具体问题所需的精确上下文量是多少。
But what matters as a goal is what is the precise amount of context AI needs to solve that specific problem.
—— Varun Krovvidi · [12:59]

指向原始笔记的链接

第二点，你真的会想拥有自己的数据层，因为现在比如一年前的轨迹就是有价值的上下文，你不会想让它们被锁在某个更商业化的系统里或类似的东西里。
And two, you want to really own the data layer because now the traces of, for example, a year ago are interesting context, and you don’t want them to be locked up in a more commercial system or something where
—— Marc Klingen · [16:20]

指向原始笔记的链接

AI 本来应该把我们从苦役中解放出来，对吧？但对大多数软件开发者来说并不是，因为我们在做更多的工作，我们有更多的上下文切换，我们正在失去那种大局观。
AI was supposed to free us from drudgery. Right? It isn’t for most software developers because we’re doing more work, we’ve got more context switching, we’re losing that big picture.
—— Sam Newman · [39:16]

指向原始笔记的链接

所以我认为你们最需要记住的一点是——而且你们已经知道了，尤其是在座的开发人员——上下文就是一切。
So I think the most important thing you need to remember is, and you already know that, especially the developers in the room, context is everything.
—— Eli Cohen · [14:47]

指向原始笔记的链接

你的 LLM 的表现取决于你喂给它们的上下文的好坏。
Your LLM are as good as the context that you feed them with.
—— Eli Cohen · [14:56]

指向原始笔记的链接

这就是为什么我认为如此多的 AI 试点或 POC 在失败,因为你没有给智能体相关的上下文以及在你的公司内如何行动的实战手册。
That’s why I believe so many AIPs or POCs are failing, because you’re not giving agents the relevant context and the playbooks how to act within your company.
—— Rafal Wilinski · [05:06]

指向原始笔记的链接

其中最重要的一点是，代码已经在磁盘上、在本地了，尤其是如果你的智能体有上下文的话，那为什么要用一个预检索系统把它复杂化呢？
And the largest part there is code is already on disk and local, especially if your agent has context, so why complicate it with a pre-retrieval system?
—— George He · [03:49]

指向原始笔记的链接

所以一百万 token 或两百万 token 的上下文，在性能方面不会帮到你。
So one million tokens or two million tokens, context will not help you in terms of the performance.
—— Vlad Luzin · [03:56]

指向原始笔记的链接

如果智能体所参考的未来上下文从未改变，那么一个错误的技能或指令就意味着，智能体今天犯的错误在未来的运行中仍会持续存在。
If the future context that that agent references never changes, then a misguided skill or instruction will mean that the agent, the mistake that the agent makes today will still persist in future runs.
—— Jake Broekhuizen · [03:48]

指向原始笔记的链接

只有当那个经验教训被转化为智能体在未来运行中可以参考的持久性上下文时，它才成为记忆。
It only becomes memory when that lesson is converted into durable context that the agent can reference on future runs.
—— Jake Broekhuizen · [04:36]

指向原始笔记的链接

它是经验如何变成上下文的方式。
It’s how experience becomes context.
—— Jake Broekhuizen · [15:51]

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

所以智能体的能力只取决于给它们的上下文。
So agents are only as good as the context that they’re given.
—— Kamalakannan Nandagopal · [04:30]

指向原始笔记的链接

但我们都知道，skills 和 MCP 的效果只取决于提供给它们的上下文。
But we all know skills and MCPs are only as good as the context that they’re provided to.
—— Kamalakannan Nandagopal · [02:44]

指向原始笔记的链接

但我们心中始终坚守的一个关键原则是：每一条落到上下文图谱上的数据点，都必须扎根于一个硬事实。
But one key principle that we had in mind always was every single data point that landed on the context graph always had to be grounded down in a hard truth.
—— Kamalakannan Nandagopal · [05:34]

指向原始笔记的链接

在智能体之间共享上下文窗口，就像是给其他智能体添加垃圾，因为并不是所有智能体都需要彼此之间的信息。
You share the context window between agents is like add garbage to the other agents because no all the agents need the information between them.
—— Elizabeth Fuentes Leone · [09:59]

指向原始笔记的链接

这个框架使智能体体验不再是一个关于模型的问题，而更多是一个关于上下文、能力和整体自我提升的问题。
This framing is what turns agent experience into less of a question of the model and more of a question of context, capability, and overall self-improvement.
—— Abhi Arya · [03:47]

指向原始笔记的链接

而上下文是什么，它就是你可以提供给智能体的那个神经系统。
And what context is, it is that nervous system that you can provide to your agents.
—— Matt Gibiec · [06:39]

指向原始笔记的链接

这些决策和建议在当时看起来可能是对的，但从长远来看它们不会是对的，因为没有足够的上下文让那个智能体理解你的环境究竟是什么样子。
And advices that might seem right at that point, but they won’t be right in the long term because there’s not enough context for that agent to understand how your environment really looks like.
—— Matt Gibiec · [06:19]

指向原始笔记的链接

但我们发现，真正有效的是把它保持得很窄，用渐进式发现来了解连接器可用的所有能力，并真正限制你塞进去的上下文数量。
But instead what we found is really keeping it narrow, using progressive discovery of all of the capabilities that the connector has available and really limiting the amount of context that you shove into
—— Pedro Lopez · [06:30]

指向原始笔记的链接

我思考推理的方式是：模型基本上是在从它们自己已经生成的内容中进行上下文学习。
And the way I think about reasoning is that models are basically in-context learning from things that they have generated already.
—— 嘉宾 · [12:25]

指向原始笔记的链接

我认为上下文学习所涵盖的，是模型为调整自身行为所做的一切，而这不是发生在权重学习内部的。
What I think of in-context learning as capturing is everything the model is doing to adapt its behavior that’s not happening within weights learning.
—— 嘉宾 · [20:30]

指向原始笔记的链接

我觉得上下文学习也是这样：如果你看聚合结果、跨一个大数据集看，事情可能看起来很平滑，也许是具有欺骗性的平滑，但实际上像单个故事这样的个别案例有着非常非线性的动态。
And I think in-context learning is like that too, where if you look at aggregate, if you look across a large data set, things might look smooth, maybe deceptively so, but actually individual cases like an individual story have these really non-linear dynamics.
—— 嘉宾 · [25:31]

指向原始笔记的链接

当每个人都想要个人智能体，或者每个企业都想要自己的、全天候 24-7 持续在线并运行大量自动化的个人智能体时，上下文窗口比内存中的模型权重更重要。
When everybody wants personal agents or every business wants its own personal agents that’s always on and working 24-7 and running a lot of automations, the context window matters even more than the model weights in the memory.
—— Arvind Srinivas · [30:26]

指向原始笔记的链接

② 出现在这些集

33 集

③ 关联

点进去有真内容 —— 本页主要出口

智能体 · Claude Code · harness · Cursor · Codex · 可观测性 · Claude · 评估 · RAG · 沙箱

反向链接