Matt Jones
① 他说过的话
10 条
因为当我们每秒处理 700 万个 span 的时候,我们会在乎两个 web 应用之间漏掉一两个调用吗?
指向原始笔记的链接
Because when we’re processing 7 million spans a second, do we care if we missed one or two calls between two web apps?
—— Matt Jones · [26:49]
所以尽管我们每秒读取 700 万个 span,我们每秒写入 Neo4j 的大约只有 4000 条记录。
指向原始笔记的链接
So even though we’re reading 7 million spans, we’re writing something like 4,000 records a second to Neo4j.
—— Matt Jones · [21:36]
我们总是可以这样拆解:服务图是正在发生什么,代码图是它如何发生以及为什么发生,基础设施图是它在哪里发生。
指向原始笔记的链接
So we can always break it down as saying, you know, the service graph is what is happening. The code graph is how and why it’s happening. The infograph is where it’s happening.
—— Matt Jones · [46:34]
你基本上是在依据代码自己怎么想的来判断,而不是它实际在做什么。
指向原始笔记的链接
You’re kind of going on what the code thinks, not what it does.
—— Matt Jones · [46:31]
以至于当人们在集成时,默认方式就是通过 MCP,而我必须真的去问:你是不是其实只需要一个 API?
指向原始笔记的链接
Like, to the point that when people are integrating, the default is via MCP and I have to go, and I have to literally go, Chua, it’s like, do you just need an API for this, really?
—— Matt Jones · [13:55]
我当时是想在一个有 20 年历史的公司、有 20 年历史的代码库里找出那些见不得人的秘密。
指向原始笔记的链接
I was trying to find the skeletons in the closet in a 20-year-old company in a 20-year-old repo.
—— Matt Jones · [17:25]
说实话,我从没见过一个应用像 Kafka 那样放弃得那么快。
指向原始笔记的链接
Honestly, it’s like I’ve never seen an application give up as quick as Kafka could.
—— Matt Jones · [24:51]
所以你知道,如果是一次挥棒落空,那么好吧,我们损失了三四个月,但我们没有损失一年。
指向原始笔记的链接
So it was, you know, if it was a swing and a miss, then, okay, sure, we’ve lost three or four months, but we haven’t lost a year.
—— Matt Jones · [67:33]
但如果你一开始就模糊,我觉得结果永远会是模糊的。
指向原始笔记的链接
But if you start vague, I think it’s always going to be vague.
—— Matt Jones · [77:44]
而如果我们真的写出精心打造的、非常具体的提示词,向智能体提出具体的问题、给它们具体的方向,我们就能非常快地得到答案,而且它们似乎不会被带偏。
指向原始笔记的链接
Whereas if we actually write well-crafted prompts, very specific prompts, we’re asking the agents specific questions and giving them specific ways on where to go, We can get answers very, very quickly, and they don’t seem to get misled.
—— Matt Jones · [78:31]
② 出现在这些集
1 集
③ 他谈到的
点进去有真内容 —— 本页主要出口
Adyen · Architect · Neo4j · Flink · Kafka · 智能体 · 服务图 · Atrium · Git worktree · MCP
