AI Agents Are Killing the Engineering Pyramid — Here's What Replaces It

频道: YC Root Access
视频: https://www.youtube.com/watch?v=m00FTHk7570
原文语言: en
统计: 共 20 轮 · Diana 12 · Reynold 8

[0:04] Diana

Today I'm excited to have Rainol Shen who's the co-founder and chief architect of data bricks which is one of the largest AI data infrastructure for enterprises companies out there. Their last round was at over 130 billion plus in valuation. Pretty impressive.

今天我非常高兴请到 Reynold Xin，他是 Databricks 的联合创始人兼首席架构师。Databricks 是当今面向企业的最大 AI 数据基础设施公司之一，上一轮融资估值超过 1300 亿美元，相当亮眼。

[0:24] Diana

Thank you Diana. And the big shift right now is AI coding agents are working. How is that changing for how data bricks is building products internally?

谢谢你，Reynold。眼下最大的变化是 AI coding agent 真的开始好用了。这正在怎样改变 Databricks 内部做产品的方式？

[0:37] Reynold

Yeah, I think one of the thing is super interesting right now is uh in a way the AI agents are reshaping the organizational structure because it used to be the case that you have to have humans. You you build out this pyramid of engineering team for any mature product. um you have so maybe the manager the leader and then the senior engineers and sense of an army of more junior people um they will be contributing code and fixing a lot of bugs and I think that is changing because the uh AI agents are when designed well when we have the right harness it's capable of actually doing a lot of the coding work um and in some cases even the design work so I think it would actually reshape the uh or organizational structure to be a little bit more of a eye shape where uh I think teams will become more and more actually in a way topheavy um and have people that really understand what needs to be built but also how to build them um where it's leaving a lot of the grunt work to be done and completely automated by AI agents and that is a huge implication I think to the uh both the tech and the or structure

是的，我觉得现在特别有意思的一点是，某种程度上 AI agent 正在重塑组织结构。因为过去任何一个成熟产品，你都得靠人，得搭出这么一座工程团队的金字塔——可能上面是经理、是 leader，往下是资深工程师，再底下是一大批更初级的人，他们负责写代码、修一大堆 bug。我觉得这一套正在改变，因为 AI agent 只要设计得当、配上合适的 harness，其实有能力承担很多写代码的工作，某些情况下甚至能做设计的活。所以我觉得它真的会把组织结构重塑成更像一个'眼睛'的形状——团队会越来越头重脚轻，留下的是那些真正懂该做什么、也懂该怎么做的人，而大量的杂活、苦力活则完全交给 AI agent 自动完成。我觉得这对技术和组织结构两方面都是巨大的影响。

[1:42] Diana

interesting so have you seen then the product velocity of how you guys are shipping to be a lot faster

有意思。那你有没有看到你们出活、发版的速度因此快了很多？

[1:50] Reynold

I think one of the Interesting thing here is that um by the way when uh I think it's a good analogy here which is when steam engines when the world first discovered electric motors or so of invented electric motors um mo many of the factories were built with steam engines in mind and in the case of steam engine like people could actually google this they had it's it's actually very difficult to build a lot of different steam engines they're pretty big pretty bulky so the factories are designed with one gigantic steam engine in mind and there's a lot of conveyor belts and of surrounding the steam engine. So they're very tightly packed into a 3D structure. But when electric motors came out, one of the biggest changes, hey, you can build fairly small electric motors. You no longer need a single gigantic electric engine or steam engine. But uh because the existing factories are configured in a way to fit a single gigantic um steam engine, uh most factory just started by replacing that steam engine with a electric motor, a more powerful electric motor. that actually only led to fairly incremental gains um of factory sort of throughput um and over the course of like two or three decades people start engineers started thinking about hey how do we redesign the um factory for electric motors and that's when really unleashed the productivity gain and throughput from uh factories I think a similar thing is actually happening with the uh software uh factories um as well and one of the uh it's actually a lot easier to create a new software factory um fully embracing set of AI tools than just from scratch than just taking an giant existing system and slap a bunch of AI in it, you'll still get some incremental gains. There's a lot of tasks. It can be automated. Um but if you just think about without changing the processes and without changing maybe all the tooling and how your CI/CD works, it's actually it's very very difficult to get a massive um productivity gain.

我觉得这里有个很有意思的点。顺便打个比方，我觉得这个类比挺贴切：当年世界刚发明电动机的时候，很多工厂当初是按蒸汽机来设计的。蒸汽机这东西大家其实可以上网查，要造很多台不同的蒸汽机非常难，它们又大又笨重，所以工厂都是围着一台巨型蒸汽机来设计的，外加一大堆传送带绕着它转，整个被紧紧塞进一个三维结构里。可电动机一出来，最大的变化之一就是：你能造相当小的电动机，不再需要一台巨型蒸汽机或电动机了。但因为现有工厂的布局是为了塞进那一台巨型蒸汽机而设计的，大多数工厂一开始只是把那台蒸汽机换成电动机、换成一台更强劲的电动机，这其实只带来了相当有限的产能提升。直到大概两三十年后，工程师才开始琢磨：怎么为电动机重新设计整座工厂？那时候才真正释放出工厂的生产率和产能的飞跃。我觉得软件这个'工厂'现在也在发生类似的事。而且其实，全面拥抱一整套 AI 工具、从零搭一座新的软件工厂，比起拿一个庞大的现有系统硬塞一堆 AI 进去要容易得多。后者你也能拿到一些增量收益，确实有很多任务能被自动化，但如果你不改流程、不改所有工具链、不改你的 CI/CD 跑法，想拿到巨大的生产率提升其实非常非常难。

[3:45] Diana

That's a very good point. So, especially for companies that have been around longer and with this shift with AI, they have to be very thoughtful on how to do that so they don't end up with this analogy you're saying with a giant

这点说得很好。所以尤其是那些成立时间更久的公司，面对 AI 带来的这次转变，他们必须非常用心地去想该怎么做，免得最后落入你说的那个困境——剩下一台巨型……

[3:56] Reynold

hole in the middle that used to be the steam engine

……中间那个原本放蒸汽机的大窟窿。

[3:59] Diana

to then retrofit with AI. What you're saying is almost like to embrace AI for a company that's already further along,

……然后再往里硬塞 AI 去改造。你的意思几乎是说，对一个已经走得比较远的公司来说，要拥抱 AI，

[4:06] Diana

you have to almost create new space.

你几乎得另辟出一块新空间。

[4:09] Reynold

Exactly. Um it's one thing is by the way don't get me wrong it is important to replace that giant steam engine with the electric motor also but that won't give you all the gains the more important part is to start thinking about how do you reconfigure things but reconfiguration is slow because you don't want to be too disruptive either so it's actually a lot easier to create for example new teams new efforts new product lines new organizations to be more AI native compared with maybe the existing bigger machine. Hm. So tell us a bit about some of the products that data bricks is shipping that is enabling more native AI coding agents to interact with.

正是如此。顺便别误会，把那台巨型蒸汽机换成电动机也很重要，但光这么做不会让你拿到全部收益。更重要的是开始思考怎么重新配置整个体系——可重新配置很慢，因为你也不想搞得太有破坏性。所以相比改造那台更大的旧机器，去新建一些东西其实要容易得多，比如组建新团队、启动新项目、开新产品线、设新组织，让它们更 AI 原生。嗯，那给我们讲讲 Databricks 正在推出的、能让更原生的 AI coding agent 与之交互的产品吧。

[4:48] Reynold

Yeah.

好的。

[4:48] Diana

Which might not be some of the products that people are know you as much for.

可能有些产品并不是大家最熟悉你们的那些。

[4:51] Reynold

Yeah, exactly. Um actually one of our fastest growing product is something that doesn't even have a data bricks brand on it. Um it's it come from the Neon acquisition and the uh the whole point of the Neon product is that it's a PLG driven motion. It's super easy to sign up. So very different from the traditional data bricks enterprise motion.

对，没错。其实我们增长最快的产品之一，连 Databricks 的牌子都没挂。它来自我们收购的 Neon，而 Neon 这个产品的核心就在于它是 PLG 驱动的打法，注册超级简单，跟传统 Databricks 那套面向企业的打法很不一样。

[5:08] Diana

What is neon actually?

Neon 到底是什么？

[5:09] Reynold

Yeah, I was gonna and neon give you a serless postgress um that autoscales um super super rapidly and also allows you take a snapshot of the database and restore and branch off the database just like you can do with code. One other thing with AI agents is that um AI agents move incredibly fast and you can use AI agents to run a lot of experiments in parallel. Many of these experiments might not work out. some of them might. And for the ones that don't work out, you want it to be super duper cheap. For the ones that might work out, you actually want to be able to run on infrastructure that can scale you to the point of going to production um at scale. And uh I think historically sort of infrastructure, especially databases were designed to be heavy weight. Uh they were think of to support hey the most mission critical applications. Um but Neon's approach is hey, let's design something that is super super cheap. Um because you can start very very small just getting a Postgress database but um and if you want to branch off run a lot of experiments you can do that instantly but if any one of experiments actually start taking off and becomes maybe what you want to go into production you just use the same environment actually go uh autoscale to whatever you need. Um so that's like actually Neon's been growing um like crazy. Um it's when we acquired Neon was about actually less than a year ago the revenue has gone up more than 10x um just in less than a year and uh we're seeing massive adoption I think mostly because of the agentic workloads they're very very different from the past workloads

好，我正要说。Neon 给你的是一个 serverless 的 Postgres，能超级快地自动扩缩容，还能让你给数据库打快照、恢复、像对代码那样从数据库上拉分支。关于 AI agent 还有一点：AI agent 跑得飞快，你可以用它并行跑大量实验。这些实验很多可能不成，有些可能成。对于不成的，你希望它便宜到极致；对于可能成的，你又希望它能跑在那种可以把你一路扩到生产、扛得住规模的基础设施上。我觉得历史上那种基础设施、尤其是数据库，都是按重量级来设计的——它们是冲着支撑那些最关键的核心应用去的。但 Neon 的思路是：我们来设计一个超级便宜的东西。因为你可以从非常非常小起步，就只要一个 Postgres 数据库；想拉分支、跑一大堆实验，瞬间就能做到；可一旦某个实验真的起飞、成了你想推上生产的那个，你直接用同一个环境就行，它会自动扩到你需要的任何规模。所以 Neon 一直在疯狂增长。我们收购 Neon 其实还不到一年，营收已经涨了 10 倍多，就在不到一年里。我们看到极其迅猛的采用，我觉得主要就是因为这些 agentic 的工作负载，跟过去的工作负载非常非常不一样。

[6:42] Diana

is it because a lot of um AI coding agents or if you go on Taj or cloud and you ask for help me build this with a postgrad database neon becomes the recommended product. Yeah, that's uh one of the key reasons. Another one is um it's also powering um many sort of agentic coding platforms like rapid versel and many others that are coming in the pipeline. Um for many of this platform they suffer from exactly the same issue I talked about earlier which is they want each individual experiment or each app to be super cheap. Um but then if they do take off they want to be able to take it to a pretty far scale and that um it's just a difficult problem from a sort of conventional infrastructure point of view. This is fascinating because you guys data bricks have been really the hardcore infrastructure company and infrastructure historically is really heavy weight is meant to be done for really production grade and hardcore engineering systems the studio systems and in this new world and this shift when AI coding agent started to work there's this new thing with lightweight infrar that's becoming a thing and sounds like neon is one of them. Yeah, I think more generally and broadly than Neon, I do think infrastructure needs to evolve in the agentic era, which is uh it needs to be able to start super lightweight. It can't be this sort of a delicate thing that requires an army of people to babysit and costs like millions of dollars um for every little thing. Like it it needs to be able to support even at approximately zero cost to begin with. And when it does whatever that's being built on top of it um gen like actually demonstrates value um then it can actually start scaling up the cost.

是不是因为很多 AI coding agent——比如你去 Claude 上让它'帮我用一个 Postgres 数据库搭这个东西'——Neon 就成了被推荐的那个产品？对，这是关键原因之一。另一个是，它还在给很多 agentic coding 平台供能，像 Replit、Vercel，还有不少正在路上的。对这些平台来说，他们恰恰碰到我前面讲的同一个难题：他们希望每一个实验、每一个 app 都超级便宜，可一旦真的起飞，又希望能把它推到相当大的规模。从传统基础设施的角度看，这就是个很难的问题。这一点特别耐人寻味，因为你们 Databricks 一直是那种硬核的基础设施公司，而基础设施历来都是重量级的，是为真正生产级、硬核的工程系统、为那种'工作室级'的系统准备的。可在这个新世界里、在 AI coding agent 开始好用之后的这次转变中，出现了'轻量级基础设施'这么个新东西，听起来 Neon 就是其中之一。对，我觉得比 Neon 更普遍、更宽泛地说，基础设施在 agentic 时代确实需要进化——它得能从超级轻量起步。它不能是那种娇贵的、需要一大群人去伺候、每点小事都得花掉几百万美元的东西。它得能做到几乎零成本起步，等到在它之上搭起来的东西真正展现出价值，再开始把成本往上扩。

[8:22] Diana

What advice would you have for founders that are getting started and building for this new world and having this concept of lightweight infra rather than the old school heavyweight? Yeah, I think it's actually a great time right now for in sort of disruption infrastructure honestly u because pretty much every piece of infrastructure were designed to be shipper way even the word infrastructure sounded shipper heavyweight right it reminds you of PG&E and maybe oil pipelines um the and I think there's a lot of architectural evolution with the cloud and with agents that you could now actually start thinking about it many of infrastructure became super heavy weight not because there's a fundamental so technical limitation is mostly because it were designed initially just for high value sort of services but now I think um with agentic coding there's going maybe the individual value of this each service or each experiment is very low but in aggregate they can be very large so there's sort of a um opportunity to target the very long tail um to be building something pretty valuable and that's just something most incumbents have never even thought of so we have a very difficult time transitioning to

对于那些正在起步、为这个新世界做开发、抱着'轻量级基础设施'而非老派'重量级'这种理念的创业者，你有什么建议？对，说实话我觉得现在正是颠覆基础设施的绝佳时机。因为几乎每一块基础设施当初都是按超级重量级设计的——连 infrastructure（基础设施）这个词听起来都超级重，它会让你联想到 PG&E 这种电力公司、联想到输油管道。我觉得随着云、随着 agent 带来的大量架构演进，你现在真的可以重新审视这件事了：很多基础设施变得超级重，并不是因为有什么根本性的技术限制，而主要是因为它们最初就只是为那些高价值的服务设计的。但现在有了 agentic coding，可能单个服务、单个实验的价值都很低，可它们加总起来体量会非常大。所以这里有个机会，去瞄准那条非常长的长尾，做出相当有价值的东西——而这恰恰是大多数老牌厂商压根没想过的，所以他们会很难转过身来。

[9:28] Diana

awesome I mean This this sounds exciting future for everyone building. I think that's it.

太棒了。我是说，这对所有在做开发的人来说听起来都是个激动人心的未来。我觉得就到这儿吧。

[9:34] Diana

Thank you so much for coming and chatting with us, Rain.

非常感谢你来跟我们聊，Reynold。

[9:36] Reynold

All right. Thank you, Diana.

好的，谢谢你，Diana。