PDF 为何仍让 AI 失灵:Reducto CEO Adit 谈文档解析

Why PDFs Still Break AI: Reducto CEO Adit on Document Parsing

阿迪特·亚伯拉罕 Adit Abraham · Latent Space · 2026-11-06 · 约 51 分钟 · 原视频 ↗

打开互动全文版(中英对照 + 朗读 + 问答)→

本期速览 · Overview

Reducto CEO Adit 解析 PDF 为何仍是 AI 的难题,以及公司如何打造人类水平的文档解析能力。

Reducto CEO Adit explains why PDFs remain a hard problem for AI and how his company is building human-level document parsing.

要点 · TL;DR

核心观点 · Key points

反共识 · Contrarian takes

本期章节 · Chapters(共 21)

全文 · Full transcript(中英对照)

欢迎与病毒式牛排推文 Welcome and viral steak tweet

Host

大家好,欢迎来到 Late Space 烹饪秀,我们邀请创始人和研究人员来,就让他们尽情发挥。今天我们有一位非常特别的嘉宾,Reducto 的联合创始人兼 CEO Adit。欢迎你,Adit。

Hey guys, welcome to the Late Space Cooking Show, where we invite founders and researchers and just let them cook. Today we have a very special guest, co-founder and CEO of Reducto, Adit. Welcome, Adit.

Adit

嘿,谢谢邀请。

Hey, thanks for having me.

Host

当然。我们其实是通过 Adele 联系上的,因为一条推文火了,对吧?你煎了一块我觉得卖相非常好的牛排。

Yes, of course. And we actually got connected through Adele after a tweet went viral, right? Where you cooked, I thought, a really good-looking piece of steak.

Adit

我觉得那简直绝了。

I thought it was incredible.

Host

但大家都在 ratio 你。是这样吗?发生了什么?

But people were ratioing you. Is that correct? What happened?

Adit

是的。有一些指控,我们今天会回应。

Yes. There were some allegations made that we'll address today.

Host

好。我们今天就来挽回一下,然后开始。今天我们有牛排,还要做 chimichurri 酱,那我们就开始吧。你能把欧芹递给我吗,我们可以开始切了。

Okay. Yeah, we'll redeem today and get started. So today we have steak and we also have chimichurri that we'll make, and so yeah, we can get started. Do you want to hand me the parsley and we can start cutting it?

从Remember All到Reducto From Remember All to Reducto

Host

好。开始之前我也挺好奇的。你最初是带着一家叫 Remember All 的公司加入 YC 的,对吧?

Yeah. And before we start as well, I was kind of curious. You initially joined YC with this company called Remember All, correct?

Adit

对。

Correct.

Host

而且到今天,你的公司叫 Reducto。所以我想问,这跟《哈利·波特》有什么关联吗?

And even today, your company's called Reducto. And so I was wondering, are there any ties with Harry Potter?

Adit

我们经常被这么问。

We get that a lot.

Host

好吧。

Okay.

Adit

嗯,是的。故事是这样的:我们早就知道要从 Remember All 转型。进 YC 那天,我们就跟 Diana 说:“嘿,我们觉得这不是我们要做的方向。”当时我们也不知道具体会转到什么。大概的想法是,我们要为开发者做魔法般的工具,于是我们在群聊里发了个投票,列出每个名字和对应的可用域名。

Um, yeah. So the story here is we knew we were going to pivot from Remember All. The day we got into YC, we told Diana like, "Hey, we don't think this is what we're going to pursue." And we didn't know what we would pivot into specifically. There was this general idea of we'll make magical tools for developers, and we made a poll in a group chat saying here are the names and the domain names available for each.

Host

然后这个名字——

And this one, the name—

Adit

里面有些名字特别糟糕。不过是啊,从那时候到现在变化很大,当时做的是长时程记忆,或者只是给……

There were some horrible names in there. But yeah, I guess a lot has changed going from at that time it was long-horizon memory, or just memory for—

Host

给智能体的记忆。

Memory for agents.

Adit

给智能体的记忆。而今天你们在做文档解析或处理。

Memory for agents. And then today you're working on document parsing or processing.

Host

对。

Yep.

文档读取的难题 The hard problem of reading documents

Host

为了给观众一点背景,你能不能再拆解一下这个问题的核心难点,甚至讲讲从 OCR 到 IDP、再到今天用视觉语言模型和智能体的历史?

And I guess to contextualize for the audience, do you want to provide some more breakdown of the core crux of the problem, and even maybe some history going from OCR to IDP and now today with VLMs and agents?

Adit

好。做 PDF 处理的公司已经有几十年的历史了。我遇到一个人,他做 PDF 处理的时间比我的年龄还长。毫不夸张,他做过让打印机打印 PDF 的原始驱动。不过人们处理文档的场景显然已经很不一样了,这些只是很多人用来归档信息的通用格式,是企业里通用的一套数据。所以如果你要让模型承担大量人类的工作,读取时的保真度达到人类水平就非常重要。

Yeah. So there are a few decades worth of companies that have worked on PDF processing. I met this guy that had worked on PDF processing longer than I've been alive. Quite literally, he worked on the original drivers for printers to print PDFs. The context in which people process documents is obviously very different though, in that these are just horizontal formats that a lot of people will use for cataloging information, and it's a unified set of data that people will use across enterprises. And so if you're going to have models do a lot of human work, it becomes really important that you have human-level fidelity when you're reading it.

Host

嗯。

Mh.

Adit

麻烦的地方在于——我在那位做打印机驱动的人那里提到过——PDF 作为一种文件格式,最初的场景是让打印机忠实还原原件,对吧?所以它并不是围绕“智能体要读它”这个问题设计的。于是我们就深深扎进了这个兔子洞:怎样才能像人一样读这些文档。

And the tricky thing is—I hinted at this with the guy that worked on printer drivers—the original context for PDF as a file format was for printers to be faithful to the original, right? And so it wasn't really designed around this problem of agents are going to read it. And so we just got really, really deep in this rabbit hole of how can you read these documents the way that a human would.

Host

这是个非常难的视觉问题。

Which is a really hard vision problem.

Adit

我们常把它比作自动驾驶研究。车道辅助在车上已经存在很久了,但说你的 Waymo 能完全端到端自动驾驶,那是完全不同的事。这就是我们想在 Reducto 达到的标准。

We liken it a lot to self-driving car research. There's been lane assist in cars for a long time, and it's a very different thing to say that your Waymo will drive fully end to end autonomously. That's the sort of bar that we want to get to with Reducto.

文档智能体平台 An agentic platform for documents

Host

明白了。好,这说得通。我猜你们也是一个智能体式平台。那除了能跟很多智能体交互之外,这通常还意味着什么?你觉得它也会是一个横向平台吗,覆盖的不只是处理和解析,甚至还有编辑,并且有一个客户可以用的节点工作室?

Gotcha. Okay, that makes sense. And I guess you guys are also an agentic platform. And I guess what does that generally entail beyond just being interactable with a lot of agents? Like do you see it also being a horizontal platform where you cover beyond processing and parsing to even editing, and having a node studio that clients can use?

Adit

对,可以这样理解:模型与文档、或广义上的非结构化文件之间的任何交互,我们都希望 Reducto 是完成这种交互的最佳方式。这包括读文档,比如解析、对信息分类、编辑。人类工作中很大一部分就是人从文档里读、再往文档里写。这会覆盖从提取数据、与其他文档核对,到最终拿到他们需要的输出的一切。

Yeah, the way to think about it is any sort of interaction that a model has with a document or unstructured files in general, we want Reducto to be the best way for that interaction to happen. That means reading the document, like parsing it, classifying that information, editing it. So much of human work is people reading from documents and writing into documents. That will take care of everything from extracting that data, reconciling it against other documents, and then getting the end output that they needed.

模型路由与R1模型 Model routing and the R1 model

Host

明白了。你们会考虑路由吗?因为我在网站上看到 Reducto 底层有 12 个以上的模型。那针对不同任务和文件,在选更便宜的模型还是更贵的模型之间,决策框架是什么?

Gotcha. And do you guys consider routing? Because I read on the website that Reducto under the hood has 12-plus models. So what's the decision framework between choosing a cheaper model, let's say, versus a more expensive one for different tasks and files?

Adit

这里的麻烦在于,这很大程度上最终是个长尾问题,对吧?你会有一大堆看起来只是纯文本的页面,然后偶尔冒出一个非常复杂的表格,而你并不知道哪一页会有这种东西。所以我们一开始的想法是,好,我们用很多更小的传统 CV 模型做基线,然后用更大的模型处理真正难的任务。不过随着时间推移,我们在模型架构上摸索出了一套挺有意思的做法。我们周一要发布这个,是我们新的自研模型,叫 R1,它能端到端完成这件事。既包括捕捉版面布局这些部分,也包括最后一公里的修正,确保尽可能接近 100%。

So the tricky thing here is that a lot of it is ultimately a long-tail problem, right? And so you'll have a lot of pages that look like they're just text, and then every once in a while you'll have this really complex table, and you're not going to know exactly which page had that or anything like that. And so we started with this idea of, okay, we'll use a lot of smaller traditional CV models for the baseline, and then we'll use larger models for the really hard tasks. Over time, though, we've been able to have this pretty interesting approach to just the model architecture. We're going to announce this on Monday, which is our new in-house model. We're calling it R1, which is going to be able to do this end to end. So components of capturing the layout and all that, but also the last-mile corrections that you'd want to make sure that you're at as close to 100% as possible.

Host

明白了。好,这挺棒的。我们现在大概可以加欧芹了。我觉得把它切得差不多,然后再加其他配料。不过是啊,接着你提到的新模型,你们决定自研智能还是租用外部智能,背后的框架甚至心智模型是什么?因为去年,虽然只是测试,你们有个 ROM OCR,是建在 Qwen VLM 之上的。而随着这个新模型出来,看起来你们可能更专注自研,不再依赖供应商。

Gotcha. Okay. That's pretty great. We could probably add the parsley now. I feel like just have it relatively chopped and then add the other elements. But yeah, I guess going on to what you mentioned with this new model, what is your framework or even mental models behind determining whether to own your intelligence versus rent it? Because last year, even though it was for a test, you had ROM OCR, which is built on top of the Qwen VLM. And with this new one coming out, it seems like you're maybe focusing on that and not depending on providers.

Adit

对。我们的论点,或者说我们团队的任务,就是无论如何都要做最好的供应商。我们的客户不会去想 Reducto 用的是自研模型还是别的什么,他们只想要惊人的结果。所以我们不想用“是不是自研”来看这件事。随着时间自然发生的是,我们非常在意准确率、成本和延迟的那条帕累托前沿。

Yeah. So our thesis, or like mandate for us as a team, is we want to be the best possible provider no matter what. Our customers don't think in terms of is Reducto using in-house models or something else. They just want incredible results. And so we don't want to look at it as is it in-house or not. The thing that's happened naturally over time is we care a lot about that sort of Pareto frontier of accuracy, cost and latency.

模型专精与基础模型 Model Specialization vs Foundation Models

Adit

是的。我们不断发现,如果你关注某一特定类型的任务,针对该任务进行后训练的模型在三个维度上都表现更优。

Yeah. And we keep finding that if you are looking at a specific type of task, a post-trained model on that task outperforms on all three dimensions.

Host

是的。

Yeah.

Adit

因此,我们的技术栈随着时间的推移越来越转向内部研发。仍然有一些非常通用的任务适合使用前沿基础模型,但如今范围有限。

And so our stack has shifted more and more towards in-house over time. There are still really horizontal tasks where a frontier foundation model makes sense, but that's limited in scope today.

Host

太好了。好的。那么接下来,你想切桃子吗?这样我们可以烤一些,然后我可以……你要用这个新砧板吗?

Great. Okay. And so for the next thing, do you want to cut the peaches so we can maybe get some grilled and I can... Do you want to take this new cutting board?

Adit

好的。谢谢。然后我拿这个,这样我就可以……

Yes. Thank you. And then I'll just take this so that I can.

Host

还有,你说做饭时聊天会分心,这不是开玩笑的。

Also, you weren't joking about it being distracting to talk while you cook.

Adit

是的。

Yeah.

深度提取与基准测试 Deep Extract and Benchmarking

Host

是的。你们最近推出了 Deep Extract,并且有基准测试显示该工具链优于许多基础模型。是的。我想,你如何看待那些来找你说“为什么我不能等这些基础模型达到那个性能水平,不需要这个工具链,直接使用它们,而不是找你们?”的人?

Yeah. And you guys recently launched the deep extract um and had the benchmark showing that the harness outperforms a lot of the base models. Yep. Um, and I guess how do you view the people come to you and say, "Why can't I just wait for these base models to get to the performance level and not need this harness and just use that compared to going to you guys?"

Adit

我认为这次新发布最有趣的事情之一是……我不知道你有没有机会读那篇博客,我们发现了一个非常奇怪的精确率和召回率之间的权衡。所以,许多这些专门的文档处理平台在你不限制推理时间时,表现比前沿模型更差。

I think one of the most interesting things with the new launch was uh I don't know if you had a chance to read the blog, like we found this really weird trade-off between precision and recall. And so a lot of these specialized document processing platforms were worse than frontier models uh when you set uncap reasoning time.

Host

但前沿模型的失败模式在于召回率。所以它们会悄悄漏掉行,错过一些上下文之类的东西。

But the failure mode for the frontier models was on recall. So they were silently dropping rows um missing bits of context like those sorts of things.

Adit

是的。

Yeah.

Adit

如果你考虑你的提取层完全遗漏信息,那种损失模式在生产中你是找不到的,对吧?就像你根本不知道你一开始没能提取出什么。嗯,这些就是为什么有一个工具链来确保和审计结果变得非常非常重要。

And if you're thinking about your extraction layer just missing information altogether, that's the sort of loss pattern that you're just not going to find in production, right? like you don't know what you failed to extract in the first place. Um and those are the sorts of things where having a harness around making sure and auditing the results becomes really really important.

Host

是的。

Yeah.

Adit

嗯,我不认为这是二选一的问题,当然有些任务前沿模型很棒。但如果你在寻找针对该任务最高效的工具,你可能应该根据你正在做的任务进行专门化。

Um where I don't see it as either or like there are certainly tasks where a frontier model is great. But if you're looking for the most efficient tool for the task, you should probably specialize in terms of the task you're doing.

Host

是的。不,这有道理。我想,即使对于公告,你们也与 Micro One 合作创建了基准测试。是的。是的。我想,你如何看待评估基准测试与性能的关联,甚至是对基准测试刷分的锯齿状性能的担忧,在不同任务之间与真实世界任务相比。

Yeah. No, that makes sense. And I guess even for the announcement, you work with micro one to create benchmark. Yeah. Yeah. And I guess how do you think about um eval benchmarking tied with performance and even the concern of jagged performance for like benchmark benchmaxing y um across like the different tasks compared to like real world tasks.

Adit

所以显然我们在内部做了很多基准测试,让我感到沮丧的是,基准测试在很大程度上已经变成了营销材料。嗯,每个人在自己的基准测试上都是第一名,所以很快就变得非常非常嘈杂。嗯,所以我们与 Micro One 合作的原因是我们想要划清界限:Reducto 不选择文档。Reducto 不设定真实标准。产品不以任何方式操纵结果。嗯,只是为了有一个独立基准测试的基线。嗯,即使那样显然也不完美。我认为最终最重要的是让人们在自己的文档上尝试。

So we obviously do a lot of benchmarking internally and something that's frustrating for me is benchmarks largely have become marketing material. M um everybody's number one on their own benchmark and so there's just this it becomes really really noisy really quickly. Um and so the reason why we worked with micro one was we just wanted this like clear line drawn of reducto is not choosing the documents. Redduck is not setting the ground truth. Product is not you know manipulating the results in any which way. Um just so there's this baseline of it's an independent benchmark. Um even that obviously is not perfect. I think at the end of the day, the thing that matters most is for people to try things on their own documents.

Host

是的。

Yeah.

Adit

嗯,但至少它给人们一个方向感,开始思考这类事情,这就是我们关心的全部。

Um but at least it gives like a directional sense for people to start thinking about these sorts of things which is all we care about.

R1发布与定价 R1 Launch and Pricing

Host

明白了。我想,结合你们最近的发布 R1,这是否表明你们做事方式的变化?你们在开发过程中有什么有趣的发现吗?

Gotcha. And I guess tying in your recent launch R1 um does this indicate a change in how you guys will um do things? Is there anything interesting that you guys discovered while working on it?

Adit

嗯,我认为 R1 对我们来说是一个巨大的进步,因为有一段时间我们的锚点是 Reducto 将是你尝试过的最准确的东西,对吧?嗯,特别是当我们开始做更大规模的工作时,比如今天 Reducto 每月处理超过十亿页。嗯,这个数字变化很快。是的。进行那种从平价到前沿的对话,使其适用于各种不同的用例,变得非常重要。嗯,我如此喜欢 R1 的原因是它不仅仅是准确性方面的改进。它还将真正简化我们的定价以及客户在生产中采用产品的方式。所以有一段时间,如果你开启 AI 功能之类的,我们会有这些乘数。是的,这些都消失了。嗯,我们的 R1 就是你发送文档给我们,每页成本将低于我们基础模型的每页成本,而准确性比我们最准确的模型更好。

Uh I think R1 is a really big step forward for us because for a while we had the sort of anchor points of reductive is going to be the most accurate thing that you've tried, right? Um and especially as we start doing larger volume work like today reducto processes a little over a billion pages every single month. Um and that number keeps changing really quickly. Yeah. it becomes really important to have that sort of perto frontier conversation of making this available for all sorts of different use cases. Um, the reason why I like R1 so much is it's not just an improvement from the lens of accuracy. It's also going to really simplify our pricing and the way that our customers get to adopt the product in production. So for a while we would have these multipliers if you were turning on aic features and things like that. Yeah, that's all going away. Um, our one is just you send us the documents, it it'll be less than the per page cost of our base models for better accuracy than our most accurate models.

Host

哇。好的。是的。所以定价肯定好多了,远远好多了。就像比我们以前便宜好几倍。

Oh wow. Okay. Yeah. So like the pricing is definitely a lot better um by far and away. Like it's multiples cheaper than where we used to be.

Adit

明白了。哇。如果你像许多客户那样大规模使用 Reducto,那真是相当令人兴奋。

Gotcha. Dang. That is pretty exciting if you use Redducto at the scale that a lot of your customers do.

Host

没错。是的。

Exactly. Yeah.

高风险行业的准确性评估 Evaluating Accuracy in High-Stakes Industries

Host

是的。这让我想起,你提到你的许多客户所在的行业容错率非常低。嗯。我有点好奇你如何思考评估甚至为系统或性能设定指标,比如 99% 的准确率可能仍然不够,如果它在企业真正需要的少数用例上失败,比如在医疗保健或容错率非常低的行业。就像不对称的用例。你如何思考评估和考虑这一点?

Yeah. Which reminds me, you mentioned that a lot of your clients are in industries that have very low margin of error. Mhm. And I was kind of curious how you maybe think through um evaluating or even setting metrics for systems or even performance where let's say 99% accuracy may still not be enough if it fails on like the few use cases that um an enterprise really needs in let's say healthcare or a very low margin error industry. Like asymmetrical um use cases. How do you like think through evaluating that and considering that?

Adit

所以首要问题是,理想情况下我们的第一遍应该尽可能准确。嗯,显然这是我们很多精力所在。但第二件事是,永远不会有真正完美的界限。就像即使你非常非常擅长阅读页面上的内容,也许会有歧义之类的东西。嗯,所以可追溯性变得非常重要。比如能够知道某物是从哪里提取的,能够说出如果人工审核时应该看什么。嗯,还有准确的置信度评分,这样他们知道何时该查看某个东西。嗯,如果你要做医疗文档工作流之类的事情,那变得非常非常重要。是的。嗯,所以在内部做基准测试时,我们不仅衡量提取准确性。我们说,在那些不完美的事情中,我们能多可靠地升级给人类说这不完美?所有这些事情也很重要。

So the first order problem is ideally our first pass should be as accurate as possible. Um and obviously that's where a lot of our head space is. But the second thing is there's never going to be a line for truly perfect. Like even if you get really really good at reading what's on the page, maybe you have ambiguity or things like that. Um and so traceability becomes really important. like being able to know where something was extracted from, being able to say uh exactly what a human should look at if they're reviewing it. Um and also having things like accurate confidence scoring so they know when to look at a certain thing. Um that becomes really really important if you're going to do something like healthcare document workflows. Yeah. Um and so internally when we do benchmarks, we don't only measure ourselves on extraction accuracy. We say like of the things that weren't perfect, how reliably could we escalate to a human that this is not perfect? All of that kind of stuff really matters too.

Host

好的。

Okay.

失败追踪与持续评估 Failure Tracking and Continuous Evaluation

Host

那么,你们会跟踪所有最难的失败实例,然后把它放进一个数据集里运行,以确保你们的测试框架没有回归吗?

And so do you I guess keep track of all the hardest failure instances and then put that in like a data set that you could run to make sure that there are no regressions with like your harnesses?

Adit

我们在这方面做了很多工作。是的。

We do a lot along those lines. Yeah.

Host

好的。你能详细说说吗?比如你们有没有使用特定的基础设施工具或框架……

Okay. Can you double click into that like are there specific infrastructure tools that you use or frameworks that

Host

我之前应该问一下。你希望我怎么选择功能?

I should have asked before. How do you want me to choose features?

Adit

这完全取决于你。如果你想那样切片,没问题。你想切得更细。我本应该做薄垂直切片。好的,

So this is completely up to you. So if you want to do it slices like that that's fine. You want to do it thinner. I should have done thin vertical slice. Okay,

Adit

我们也有很多桃子,所以我们可以吃掉那些,然后

we also have a lot of peaches, so we can eat those and then

Adit

用更薄的来烤,放在你的三明治上。

use the thinner ones to grill and put on your sandwich.

Host

嗯,但是是的,再次回到你的框架或工具,或者你如何考虑捕捉这些失败案例,然后对它们进行基准测试。

Um, but yeah, again, back to like your framework or tools or how you think about um catching these failure cases and then benchmarking against them.

Adit

是的,我们不太公开谈论的一件事是我们的机器学习团队在持续评估上投入了大量时间。所以能够发现那些可能不是不正确,甚至只是不完美的地方,这让我们有能力

Yeah, something we don't talk about as much publicly is our ML team puts a lot of time into continuous evaluations. So being able to spot things that are maybe not incorrect but even just imperfect in production um which gives us the ability

Host

嗯。

Mhm.

Adit

开始理解对于真实世界用例来说什么是分布外的。

to start understanding what is out of distribution for real world use cases.

Host

是的。

Yeah.

Adit

嗯

Um

Host

这不是关于单个数据点本身。更多的是关于失败模式的模式。就像我们在医疗保健领域非常关心能够知道。是的,你可以说你在意手写,也许你对你的手写性能进行了基准测试。嗯,但你可能没有很多数据点是关于在倾斜页面上手写,你知道有人在网格线外涂鸦。嗯,那些只是知道你在哪里不足是第一步,让你能够捕捉到你想要在下次训练运行中包含的那些东西。嗯。

and it's not about the individual data point itself. It's more so about patterns of failure modes. Like we care a lot about being able to know in healthcare. Yes, you can say that you care about handwriting and maybe you benchmark your handwriting performance. Um, but you might not have a lot of data points that are handwriting on a tilted page where you know somebody scribbled outside the grid lines. Um, and those just knowing where you're falling short is the first step for you to be able to catch the sorts of things that you want to include in your next training run too. Mhm.

Host

你们有行业特定的,嗯,我猜是标准吗?因为我假设医疗保健公司的需求可能与律师事务所的需求不同。嗯,但律师事务所之间可能更相似,医疗保健公司之间也可能更相似。尽管行业不同。

And do you have industry specific um I guess criteria cuz I assume the needs at a healthcare company may want versus a law firm will probably differ. Um but like the law firms may have closer similarities and the healthcare companies may have closer similarities as well. Even though the industries are different.

Adit

是的。嗯,这是我们早期经常争论的问题,尤其是

Yeah. Um this was a debate we had a lot in the early days especially

Host

因为呃,构建一个垂直特定的文档处理平台要容易得多。

because uh it's a lot easier to build a vertical specific document processing platform.

Adit

是的。嗯,你做出的权衡是,我认为很多文档最终归结为非常相似的基本元素。

Yeah. Um, the trade-off that you make there is I think a lot of documents ultimately boil down to very similar primitives.

Host

就像人们认为这是理所当然的。作为一个人,当你看着一页纸时,你会做出很多基本的判断,这些只是瞬间的判断,对吧?就像

Like people take this for granted. You as a human when you look at a page make a lot of fundamental judgments that are just snap judgments, right? Like

Adit

每次两个段落之间有间隙,

every time there's a gap between two paragraphs,

Host

你知道,好的,这是一个新的语义信息块。你不会有意识地感知到这一点,但它是直觉的。

you know, okay, this is a new semantic block of information. You don't like consciously perceive that, but it's intuitive.

Adit

是的。其他一切也归结为类似的事情,就像你自然理解你知道表格结构如何关联两个维度的信息,所有这类东西。

Yeah. And everything else kind of boils down to a similar thing like you naturally understand you know how a table structure is associating two dimensions of information all that kind of stuff.

Host

是的。

Yeah.

Adit

嗯,我们为确保对杂乱的表格和通过传真发送的医疗保健数据具有韧性所做的工作,意味着我们自然也能更好地为金融领域的客户提供可靠的解决方案。

Um and the work that we do on making sure that we are resilience to you know messy tables and healthcare that were faxed means that we are naturally better at offering a reliable solution for our customers in finance as well.

Host

嗯

Um

Adit

所以这总体上是一项更艰难的任务。是的,

and so it's like a a harder undertaking overall. Yeah,

Host

我认为净效果是,我们可以为这些客户提供最具挑战性的案例服务。

I think in nets it just means that we can serve the most challenging cases across these customers.

Adit

明白了。所以你们从不同行业获得的所有学习都有重叠,因为你们专注于相同的基本元素,并且像

Gotcha. So all the learnings that you have from different industries kind of have that overlap since you're focused on the same primitives and like

Host

是的。嗯,基本上是将人类在看文档时的直觉构建到你的模型中。

Yeah. Um basically building the intuition that a human has when they look at a document into like your models.

Adit

这就是目标。是的。显然会有细微差别,比如如果你在法律领域,你可能非常关心删除线和下划线之类的东西,嗯,在医疗保健中就不那么重要,但拥有那种基线直觉对我们来说非常非常重要。

That's the goal. Yeah. And obviously there are going to be nuances like if you're in legal you probably care a lot about something like strikethroughs and underlines um matters a bit less in healthcare but just having that baseline intuition is really really important for us.

Host

是的,这说得通。这也是 Reducto 的未来愿景吗?它永远不会垂直化,而是彻头彻尾地水平化,嗯,这样任何行业、任何公司都能拥有一个非常专业或非常通用的解决方案,并且表现良好。

Yeah that makes sense. And is that kind of the future vision as well of reducto where which won't ever become verticalized will be horizontal through and through um so that any industry any company will have a very specialized or very general solution that performs well.

Adit

是的。我们不希望处于这样一种情况:对于像解析这样的客观任务

Yeah. We don't want to be in a situation where for objective tasks like parsing

Host

你可以非常清楚地说在给定的页面上,就像嘿,这就是页面上的确切内容,对吧

you can very clearly say on a given page like hey this is exactly what was on the page right

Adit

嗯,我们希望那是水平化的,我们希望那是零样本的

um we want that to be horizontal we want that to be zero shot

Host

对于主观意见的事情,比如财务团队对某个指标可能有不同的定义,是的

for things where it's a matter of opinion like maybe a finance team has a different definition of a certain metric yeah

Adit

那些事情应该是可提示和可引导的,并随着时间的推移而改进

those things should be promptable and steerable and improve over time

Host

没错,好的,这说得通。我猜总的来说,关于 Reducto,嗯,你如何看待智能体和 LLM 的整个繁荣?嗯,这是否是你们为了采用而做了很多文化变革,比如强迫你的工程师使用编码智能体,嗯,甚至内部有很多工具,还是主要是你们试图将所有努力集中在客户将使用的产品上,或者甚至是两者的混合?

true okay that makes sense and I guess Just overall with respect to reductto um how have you kind of perceived this whole boom of agents and LNS? Um is it something that you've kind of made a lot of cultural changes to to I guess adopt and like force your engineers to use coding agents um or even like internally have a lot of tooling or is it mainly like you're trying to focus all of these efforts towards a product that your customers will use or even a mix of both?

Adit

是的,我的联合创始人嗯很年轻。嗯,这是他大学毕业后的第一份工作,所以我认为他恰好是你会遇到的最 LM 原生的人之一。嗯

Yeah, my co-founder um is very young. Um this is his first job out of college and so I think he he happens to be one of the most LM native people you'll meet. Um

Host

就像你给他发消息,回复会像是 ChatGPT 发的。嗯,所以鉴于此,我认为他只是自然地,并不是我们在强制我们的工程师说你们必须使用 Claude Code 或类似的东西。

like you'll send him a message and it'll be like sent by a chat GPT back. Um and so given that I I think he just naturally it's not that we're mandating our engineers and saying like you must use Claude Code or anything like that.

Adit

嗯,人们看到他和团队其他人做的那些事情,就很自然地想要去做。

Um it is very natural for people to see the sorts of things that he and others on the team are doing and like want to do it.

Host

是的。

Yeah.

Adit

对我来说最有趣的事情之一是,我们的销售团队现在大量使用 Devin。嗯

One of the most interesting things for me is like our sales team now uses Devon a ton. Um

Host

哇。

Wow.

Adit

就像人们会为潜在客户制作定制网站之类的。

Like people will make custom websites for prospects and things like that.

Host

是的。

Yeah.

Adit

嗯,这并不是真正的自上而下的推动。我认为人们只是开始看到,嘿,我在 Google Slides 上花了大量时间,而事实证明 Claude Design 可以做出比你投入精力更好的提案。

Um and it wasn't really a top down push. I think people just started seeing like, hey, I'm spending a ton of time in Google Slides and it turns out that cloud design can just make a proposal better than you know what you put effort into.

Host

是的。

Yeah.

Adit

嗯,所以他们就会那样做,这更多的是展示价值可能是什么,而不是其他任何东西。

Um, and so they'll just do that like it's more about showing what the value could be more so than anything else.

Host

明白了。所以这是以身作则吗?比如你的联合创始人大量使用它,然后你的员工就像,“哦,哇。也许我也应该试试,变得更高效?”

Gotcha. So, was it like a lead by example like your co-founder was using it a lot and then your employees were like, "Oh, wow. Like I should maybe try it out as well and become more efficient?"

Adit

是的,是这样,加上嗯,我们从第一天起就允许 LLM 和现场面试,或者只是一般的面试

Yeah, it was that plus um we've allowed LLM and on-sites since day one or just interviews in general

Host

而我们的现场面试实际上就像一整天的现场。所以,你和团队待 8 个小时,呃。

and our on-sightes are actually like a full day onsite. So, you spend 8 hours uh with the team.

Adit

哦,哇。是的。

Oh, wow. Yeah.

Host

而这些问题最终真的很难实际解决,除非你正确使用 LLM。

And the problems end up being really difficult to actually solve for unless you use LMS properly.

招聘理念与痴迷 Hiring Philosophy and Obsession

Host

我认为从本质上讲,最终通过现场面试的人通常是那些知道如何有效使用编程工具的人。

I think just by nature, the people that end up making it through the on-site typically are people that know how to use coding tools effectively.

Adit

是的。

Yeah.

Adit

我们会看到一些极端案例,比如我们团队有个叫 Arth 的人。呃,如果 Arth 在看这个,很抱歉让你为难了。嗯,Arth 加入我们时还是伯克利的学生。他来现场面试时大概 20 岁。

And we'll see some extreme cases where there's this guy on our team named Arth. Uh if Arth is watching this, I'm sorry for putting you on the spot. Um Arth joined us while he was still a student at Berkeley. Um he was like 20 years old when he came on site.

Host

哇。

Wow.

Adit

所以 GPT-3 刚出来时他还在读大一。是的。嗯,如果你看他写代码,就像看一个瘾君子一样,同时处理的事情数量之多。嗯,

And so he was a freshman when GPT3 first came out. Yeah. Um, and if you watch him coding, it's kind of like watching a crackhead just like in terms of the number of things that he has going on at one point. Um,

Adit

尤其是他,最疯狂的是我记得我们给了他标准的软件工程现场面试。

and the craziest thing with him in particular was I remember we gave him our standard software engineering onsite.

Host

嗯。

Mhm.

Adit

他大概两小时就全做完了。

And he did all of it in like 2 hours.

Host

哇。

Oh wow.

Adit

嗯,所以 Ronic 在 Slack 上私信我,非常紧张地说:“糟糕,他会觉得很无聊的。我们该怎么办?”于是我们又给了他整个机器学习工程的现场面试。嗯,这就是我们想确保纳入团队的那种人。

Um, and so Ronic was DMing me on Slack, really stressed, saying like, "Shoot, he's going to be really bored. Like what do we" And so we gave him our entire ML engineering onsite, too. Um, and that's the kind of person that we want to make sure we include on the team.

Host

哇。是的,那真是完美的招聘。我觉得这个人对工具非常非常原生,而且行动非常快。

Wow. Yeah, that's a perfect hire. I feel like someone who's very very native with the tools and also moves very fast.

Adit

是的。

Yep.

Host

嗯,很好。好的。你想尝尝这个 chim chur 吗?是的。

Um, great. Okay. Do you want to try some of the chim chur? Yes.

Adit

然后我觉得一个桃子应该就够了。

And then I think a peaches that should be good enough.

Host

嗯,那边还有勺子,如果你想用的话。是的,我们准备好了喷枪。

Um, there's also a spoon over there if you want to use. And yeah, we got it set up for blowtorching.

Adit

酷。

Cool.

Host

但我想趁我们做这个的时候,深入聊聊招聘。你们以招聘非常慢和团队非常精简而闻名。最初你们最早的员工之一,我想你的第一个员工是 Ephe,他基本上是业内最优秀的之一,嗯,在这个特定领域,甚至你们其他的增长招聘也非常专业,有自己的业务,甚至 Adele 等人。所以我想现在从只有你和 Ronic 到现在成为一个规模化的团队,你在顶级员工中是否看到一些反复出现的特质,甚至今天在招聘时你也会寻找这些特质?

But I guess while we're doing that, I want to double click on this with hiring. So you guys are notorious for hiring very slow and moving with a very lean team. initially your one of your earliest hires. I think your first hire was Ephe who's basically one of the best in the industry um for this specific topic um and even your other like growth hires um very specialized have their own business even Adele others and so I guess now going from just you and Ronic to now being a scaled up team are there traits that you see recurring among like your top hires that you even today when you're recruiting try to look for

Adit

我们的第一文化价值观,嗯,我们的第一文化价值观是对你的手艺有执念。嗯

Our number one cultural value um our number one cultural value is having obsession for your craft M

Adit

嗯,我认为你可以把执念看作只是欲望,但真正我关注的是,如果你和任何真正痴迷于在某件事上做到卓越的人交谈。

um and I think that like you can think of obsession as just desire, but like the the real thing that I look out for is if you talk to anybody that is really really obsessed with this idea of being great at something.

Host

是的。

Yeah.

Adit

嗯,厨师、运动员,任何人,

Um cooks, athletes, whoever,

Adit

你总是发现他们对待日常的方式有一种严谨。嗯

you always find that they have this like rigor to the way that they approach the day-to-day. M

Adit

嗯,比如 Yee 是一位不可思议的机器学习研究员,不仅仅是因为他在我们做的领域读了博士。嗯,他可能是最痴迷于检查数据的人之一。就像人们非常非常仔细地查看数据集。嗯,他可能记住了单个 PDF 页面,就像

um like Yee is an incredible ML researcher, not just because he did a PhD in what we do. Um he is probably one of the most obsessive people about inspecting the data. Like people look through the data set really really closely. Um probably has like individual PDF pages memorized like

Host

哇。

Oh wow.

Adit

我不知道有谁在 PDF 上画框的时间比他更多。

I don't know anybody that has spent more time drawing boxes on PDFs.

Host

哇。

Wow.

Adit

嗯,这很重要。

Um and that matters.

Host

无法规模化。

Don't scale.

Adit

没错。是的,这真的非常重要,即使在我们某个时候数据标注员比全职员工还多的时候。

Exactly. Yeah. like it it's really really important even when we had at some point we had more data labelers than we had full-time employees.

Host

哇。

Oh wow.

Adit

嗯,但他仍然是那种亲自查看数据并真正痴迷于此的人。

Um but he would still be the type to look at the data himself and like really obsess over it.

Adit

我认为工程师需要对客户支持之类的事情痴迷。这不仅仅是关于炫酷的功能发布。而是关于确保在生产中长尾案例得到解决,诸如此类的事情。

I think engineers need to be obsessed about things like customer support. Like it's not just about the flashy feature launch. It's about making sure that in production the longtail cases are addressed like all that kind of stuff.

Host

是的。

Yeah.

Adit

嗯,我认为强迫某人想要做这件事真的很难。就像如果他们把工作看作他们想要做到不可思议的事情。

Um and I think that's a really hard thing to force somebody to want to do. Like if they look at their work as this thing that they want to be incredible at.

Host

是的。

Yeah.

Adit

嗯,他们自然有这种想要以那种方式改进的倾向。

Um they naturally have this tendency to want to improve in that way.

Adit

所以我们要试着筛选出这样的人。

Got so we try to select for that.

Host

好的。那么你如何评估或从潜在员工那里发现这一点?你只是看他们的经验?他们过去做了什么,或者面试中有一些问题或流程,让你能够确定他们对craft的痴迷程度,甚至在过去?

Okay. And I guess how do you assess that or get that out from a prospective employee? Do you just look at their experience? What they've done in the past or are there questions processes in the interview that you kind of are able to determine how obsessive they are with their craft or even in the past?

Adit

我认为痴迷的人往往在他们的工作中以许多不同的方式痴迷。嗯,

I think obsessive people tend to be obsessive in many different ways throughout their work. Um,

Adit

也许有时它出现在他们的全职工作中。也许有时它出现在爱好中。是的。

and maybe sometimes it shows up in their full-time job. Maybe sometimes it shows up in like hobbies. Yeah.

Adit

我非常喜欢问人们一些他们世界级水平的事情。

I really like asking people about something that they are just world class at.

Adit

你会看到他们谈论那件事时眼睛发光的样子。嗯,甚至可能是电子游戏,像那样痴迷的人,只要你引导他们走向正确的事情,正确的问题。

And you see the way that their eyes kind of light up um when they talk about that thing. Um, could be a video game even like people that are obsessive in that way as long as you orient them towards the right things like the right problems.

Host

是的。

Yeah.

Adit

自然会转移过来。

Naturally transfer over.

Host

明白了。顺便说一句,有很多桃子切片。

Gotcha. Plenty of peach slices by the way.

Adit

是的。我拿几个,我们也可以用喷枪烤它们。不,这有道理。我认为执念确实会延续,然后我们再聊。

Yeah. I'll grab a couple and we could also blow torch them. No, that makes sense. I think obsession does carry over and then we'll talk.

Host

是的。

Yeah.

Host

你想烤多焦?随你喜好。老实说,你可以把喷枪放进去。

How charred do you want it to be? as as much as you prefer. Honestly, you could put a torch in there.

Adit

老实说,我以为它会算数。

Honestly, I thought it would count.

Host

是的。

Yeah.

Host

你想让我就把它扔在锅里?

You want me to just toss it on a pan?

Adit

是的,我们应该就扔

Yeah, we should just toss

Host

我在这里有点挣扎。

I'm struggling a bit here.

Adit

我们用脉冲炉。嗯,你想试试吗?所以你只需打开它。

And we'll use the impulse stove. Um, you want to try? So, you just turn it on.

Host

我太兴奋了。

I'm so excited for this.

Adit

是的。所以你可以扭动它。

Yes. So, you can just twist it.

Host

哦。任何温度。嗯嗯

Oh. any temp. Um um

Host

是的,我看过创始人煎牛排的视频,我租公寓,但如果我不租,我会买一个。

yeah, I've seen the videos of the founder cooking a steak and I rent my apartment, but if I didn't, I would be purchasing one.

Adit

是的,我觉得这个炉子非常非常好。它加热超快。甚至烧水,也很容易。

Yeah, I think this stove is very very nice. It heats up super fast. Even boiling water, it's it's very easy.

Host

是的。好的。从非常早期阶段到现在 B 轮,再次恭喜。

Yeah. Okay. Going from very early stage now, series B, congratulations again.

Adit

谢谢。

Thank you.

Host

嗯,从这么小的团队到现在我猜是增长阶段,有没有什么你必须忘记或甚至改变的事情?

Um, have there been any things you've had to unlearn or even change going from such a small team now to I guess growth stage?

Adit

是的。嗯,第一,我认为公司的问题数量,假设一切顺利,实际上永远不会减少。

Yeah. Um, one I I think the number of problems at a company assuming things are going well never actually decrease in size.

Adit

嗯,你只是随着时间的推移处理不同的问题。所以就像第一天,我的很多工作只是当时需要的任何琐碎事情。嗯,所以就像单个交易或单个支持票。显然随着你有更多人,这会改变。

Um, you just get to work on different problems over time. And so like day one a lot of my job was just whatever in the weeds thing I needed at that moment. Um and so it's like an individual deal or like an individual support ticket. Obviously that changes as you have more people.

Host

是的。

Yeah.

Adit

嗯,放弃这个很奇怪。我认为作为创始人,很长一段时间,如果我不经常在 Slack 上,我会感到焦虑。

Um and it's weird to let go of that. I think as a founder like for a long time I would have this anxiety if I wasn't constantly available on Slack.

Host

嗯。

Mh.

Adit

嗯,因为感觉如果我不看着,好像就会出错。是的。

Um because it just felt like if I'm not looking like something could go wrong. Yeah.

Adit

嗯,随着时间的推移,这变得不那么真实了。就像你开始意识到公司在某种程度上有了自己的生命。

Um that's been less true over time. Like you start to realize that the company has a life of its own on some level.

Adit

嗯

Um

Adit

但你仍然总是花很多时间与你的顶级客户在一起。

but you always still spend a lot of your time with your top customers.

创始人角色的演变 Evolution of founder role

Adit

随着时间推移,你会开始处理更多与人相关的事务。比如我从未想过自己会成为那种写文化文档的创始人,但奇怪的是,你的很多工作开始变成这类事情。

You start having more people-related things over time. Like I never thought I would be the type of founder to write culture documents, but weirdly a lot of your job starts being things along those lines.

Host

是啊。

Yeah.

Adit

但这都很有趣。我经常发现自己不擅长当前阶段的工作,然后三个月后你变得擅长了,接着又开始做新的工作。

But it's all fun. Like I constantly find that I suck at the current phase of my job and then you get good at three months and then you start doing a new job.

Host

是的。随着你的成长和规模扩张,它不断演变。

Yeah. It constantly evolves as you grow and scale.

Adit

没错。

Yep.

团队扩张中的优先级设定 Priority setting as team scales

Host

这个过程大体上是一样的吗?因为我记得听你说过,在早期阶段你会给工程师一周一个优先级。

Is the process generally the same? Cuz I recall hearing you say in the early stage you would give one priority to an engineer for a week.

Adit

那可能直接与你想要实现的目标相关。是的。现在团队规模扩大了,你仍然试图强制执行一周一个优先级并完成它,还是更像一个大科技公司那样的组织?

That would kind of be maybe directly related to the goal you want to achieve. Yeah. And now that the team has scaled up, do you still try to enforce that one priority get it done in a week or is it a little more like an organization like a big tech company?

Host

是的。我们仍然每周开全员会议。所以工程团队应该对我们的顶级交易有很好的了解,反之亦然。所有这些都没有改变。我确实认为我们的一些重要项目很难压缩到一周内,所以也许某些事情的时间范围已经改变了。

Yeah. We still meet as a full team every single week. So like engineering should have good context on our top deals and vice versa. Like all of that has not changed. I do think some of our important projects are hard to dismiss into like one week and so maybe time horizons for some things have changed

Adit

而且仍然有一些事情是,你知道,像是当天的优先级,甚至不是一周的优先级。

And there are still things that are on the order of you know like it's a priority of the day not even priority of the week.

Host

但我们确实努力确保明确我们真正关心的一系列事情。

But we do try to make sure that there's clarity on like what are the set of things we really care about.

Adit

是的。我的想法已经从每周更多地转向月度优先级和季度优先级。

Yeah. And my thinking has shifted a bit more from weekly to more so like monthly priorities and quarterly priorities as well.

Host

明白了。好的。

Gotcha. Okay.

Adit

但这并不完全像大科技公司组织,我们没有像 OKR 那样每年要应对的一系列事情。

But it's not quite like big tech organization like we don't have a annual set of things like OKRs that we're going up against or anything.

Host

是的。我们仍然试图像初创公司一样精益运营,而且

Yeah. We're still trying to operate it lean like a startup and

Adit

很好

great

Host

因为你不断学习新信息,对吧?就像

cuz you're constantly learning new information, right? like

Adit

如今市场变化如此之快,如果你不根据当前情况调整,你就会落后

the market changes so quickly today that if you don't adjust as a function of whatever the current thing is like you're behind

Host

没错,尤其是在这个 AI 时代

no s there especially in this age of AI

Adit

而且编程领域发展非常快

and coding it's moves very fast

拥挤市场中的防御性 Defensibility in a crowded market

Host

那么稍微转向商业方面,我知道你们所处的领域有些拥挤,你们有其他竞争对手,每个人都在尽可能快地提升性能。是的。那么你如何考虑防御性,尤其是在你可能无法训练所有客户数据的情况下,也许性能无法随着规模扩展而扩展,以及你如何差异化并保持领先?

But kind of then to pivot more to the business aspect I know that you guys are in a space that is somewhat crowded you have other competitors and everyone is going as fast as possible to improve the performance Yeah. So how do you kind of think about defensibility um in the case where you may not be able to train out all the data of your customers um and so maybe the performance can't scale as the volume scales um and how you differentiate and kind of stay ahead of pack.

Adit

是的。我的意思是,像我之前提到的,擅长持续评估最终成为我们思考每月处理十亿页的关键部分,比如你如何

Yeah. So I mean things like what I mentioned earlier of being really good at continuous evaluations ends up being a key part of how we think about your processing a billion pages per month like how do you

Host

围绕那个建立飞轮。

have a flywheel around that.

Adit

嗯。我认为我们的很多工作不仅仅是针对你知道的某种单一布局之类的。而是这种横向性质,而且非常好的一点是,我们的客户群越广泛,我们对问题范围的直觉也越好。明白了。

Mhm. I think a lot of our work is not just on you know an individual type of layout or anything like that. It's this like horizontal nature and a really nice thing is the more the broader our customer base gets the better our intuition around the range of problems becomes too. Got it.

Host

嗯

Um

Adit

我最终认为竞争市场真的很好,因为它非常精英化。嗯,所以服务一些领先的 AI 原生初创公司,比如 Harvey,就意味着如果我们不继续保持最好,他们就会自己寻找最好的。

I ultimately think that competitive markets are really nice because like it's very meritocratic. Um and so serving you know some of the leading AI native startups uh like a Harvey just means if we don't continue being the best like they will look for the best themselves.

Host

是的。

Yeah.

Adit

嗯,我们的立场一直是,我们只需要迎接那个挑战。

Um and our stance has always been like we just need to rise to that challenge.

Host

是的,有道理。

Yeah, that makes sense.

Adit

所以你会说你可能首先是竞争性的。你想赢,而且

So you would say you're probably competitive first. You want to win and

Host

而且是的,有一个公平的竞争环境。

and yeah have a level playing field.

Adit

是的。

Yes.

Host

很好。是的,有道理。我想是的,尤其是这些天,随着格局的变化,很多市场会非常拥挤,问题就在于你是否能成为那个快速行动并竞争的人。所以是的,很高兴听到。我认为比以往任何时候都更少关于系统性模式,而更多关于执行。

Great. Yeah, that makes sense. I think yeah especially these days too um with how the landscape is yeah a lot of markets will be very crowded and it's just a matter of can you be the one to move that fast and u compete so yeah it's good to hear I think more so than ever before it's less about like systemic modes and more so on execution

Adit

是的,不断如此,当然当然。很好,好的。我想桃子看起来不错,把它们转移到盘子里,然后你开始做这个牛排。那么是的,有点回到我们之前提到的关于 Reducto 的未来。

yeah constantly for sure for sure great okay I think the peaches look good to transfer them to the plate and then you see get started on this steak so yeah kind of going back on what we touched before about the future of reductto.

未来产品方向 Future product direction

Host

该死,我能再要点油吗?

Damn, can I get some more oil?

Adit

好的。

Yeah.

Host

嘿。

Hey.

Adit

是的。关于 Reducto 的未来,你是否认为自己可能会扮演像 Rippling 那样的角色,拥有许多不同的横向产品,有点像这种复合型初创公司提供大量服务,甚至可能接管文件存储或客户可能拥有的超出提取、解析、编辑之外的其他需求?

Yeah. on the future of Reducto, do you kind of yourself maybe taking the role like a rippling where you have a lot of different products that are horizontal um and kind of having this like compound startup that provides a lot of services and maybe even taking over like file storage or um yeah other needs that your clients may have beyond just extraction, parsing, editing.

Adit

是的,我们最关心的是与文档本身的实际操作。好的。所以我们思考的方式不太关注文件存储之类的事情,对我们来说问题始终是:我们能否成为那项任务的世界最佳。当你考虑文档分类、解析和编辑等事情时,答案是肯定的。

Yeah, we care the most about the actual actions with the document itself. Okay. So the way that we think about it is less so focused on file storage and things like that like the question for us is always can we be the best in the world at that task. And as you think about things like classification on documents and parsing and editing like the answer is yes.

Host

嗯,至少到今天为止,像仅仅存储数据这样的事情对我们来说不太有趣。

Um, at least as of today things like you know just storing the data is less interesting for us.

Adit

是的。嗯,我们只有在那种方向能为客户带来更好体验时才会真正推进。

Yeah. Um, we would only really push on those sorts of directions if it in that leads to a better experience for our customers.

Host

明白了。

Got it.

客户成功故事 Customer success stories

Host

那么每当你的客户来的时候,我也听说他们通常带着一组之前所有测试都失败的文档,但不知怎么的 Reducto 就能处理。你有没有什么有趣的故事可以分享,比如关于实例或者任何你能分享的,那种发生了非常神奇的时刻?

And so whenever you get your customers, I also heard that um they usually come with a set of documents that have failed all the previous tests y before and somehow it works with reductto. Do you have any fun stories to share I guess about instances or like whatever you can um of like moments where something was just very magical that happened for?

Adit

是的。嗯,我最喜欢的那种客户演示循环是当他们发现一个以前无法实现的用例。

Yeah. Um, my favorite type of just like customer demo loop is when they find a use case that previously could not work.

Host

嗯。

Mhm.

Adit

嗯,然后他们最终能够解决那个问题。就像他们开始引入他们之前认为超出范围的事情。

Um, and they end up just like being able to solve for that thing. Like they start introducing things that they just seen as out of scope.

Host

嗯。

Mhm.

Adit

嗯,还有一些事情我们发现,解析准确率最终会转化为人们没有考虑到的下游效应。比如我们的一个法律科技客户最终发现,当文档由 Reducto 解析时,客户与文档的互动增加了一倍。

Um, there's also a few things where we find that like parsing accuracy ends up translating into downstream effects that people don't consider. Like one of our legal tech customers ended up finding that customers interact with documents twice as much when it's parsed by reducto.

Host

哇。

Oh wow.

Adit

嗯,比如他们会问更多问题。嗯,这显然不是我们销售的功能。嗯,但他们只是发现答案质量和引用精度提高了,以至于整体上客户更有可能参与。是的。

Um like they ask more questions. Um and it was something that like it's obviously not a feature that we sell. Um but they just found that like answer quality improved and citations precision improved such that overall customers were more likely to engage. Yeah.

Host

这很重要。这正是我们最终想要推动的事情,但也许它不会立即在邮件流程中显现出来。

And that's huge. Like that's the kind of thing that we ultimately want to drive, but maybe it doesn't show up immediately in an email process.

Adit

明白了。好的。是的。

Gotcha. Okay. Yeah.

创始关系与团队动态 Founding Relationship and Team Dynamics

Host

我想即使现在在团队里,Ronic,你们俩都上了麻省理工学院,对吧?然后你们是通过他做助教的一门课认识的。

And I guess even now on the team with you, Ronic, you guys both went to MIT, right? And then you met through a class that he was RA for or like being a TA.

Adit

嗯,是的,当时他还是大一新生。

Um, and yeah, when he was a freshman.

Host

是的。你过去说过,你对他的能力和才华印象深刻,然后你们一起工作。我想现在大概已经两年了。

Yeah. And you have said in the past that you're very impressed by just his ability and his talent and then work together. And I guess over the course of now it's been like two two years.

Adit

两年——快三年了。

Two—we're coming up on three.

Host

三年。是的。在这整整三年里,有没有哪些时刻让你意识到你们真的很合拍,这些特质对联合创始人来说非常重要?或者有没有一些时刻,你们似乎违背了传统的创业建议?

Three years. Yeah. Like during those entire three years, have there been moments where you realized like you guys really clicked and these traits are very important for co-founders to have or is the moments where you know seem non-intuitive that the conventional startup advice they give that you guys kind of go against?

Adit

是的。嗯,我觉得到目前为止,我们创始关系中最让我喜欢的一点是,我们仍然从未有过真正意义上的争吵。

Yeah. Um, I think one of my favorite things about our founding relationship to date is, uh, we still have never had what I would truly classify as like an argument.

Host

好的。

Okay.

Adit

嗯,我们有过很多关于事情的讨论,也谈过不同的意见。嗯,但我们有一个非常好的动态,就是根据事情本身的优劣来讨论,以及解决争端。

Um, like we've had plenty of discussions about things and talk through deferring opinions. Um, but we have a really really good dynamic of talking through things on their own merits uh, and like resolving disputes.

Host

嗯,好的。对我来说,老实说,除了需要有一个基线,如果你真的尊重对方,你认为他们带来了互补的技能组合等等,这是最重要的。

Well, okay. And to me, that's honestly the most important thing outside of there needs to be this baseline if you really respect the other person. You think that they bring a complimentary skill set and all that.

Adit

嗯,但创业真的很难,重要的是你觉得你可以和联合创始人一起经历起伏而不心怀怨恨。嗯,这很特别。我觉得这在初次见面时很难强求或预测。

Um, but startups are really hard and it's really important that you feel like you can go through the ups and downs with your co-founder without being resentful. Um, and that's special. Like I think that's hard to force or forecast when you're first meeting somebody.

Host

是的。你和 Ronic 有互补的技能组合吗?

Yeah. And do you and Ronic have complimentary skill sets?

Adit

我觉得有。比如我在麻省理工学院完成了整个本科的计算机科学学业,但嗯,相比之下,我几乎可以说是非技术性的。嗯,是的,通常我们会在产品上有重叠。嗯,他只是在全公司范围内领导很多工程工作。嗯,而且他倾向于在这方面有很强的意见。然后反过来,在市场推广方面是我。

I'd say so. Like I did my entire undergrad in CS at MIT, but um in comparison, I might as well be non-technical. Um yeah, like typically we will overlap on product. Um he just leads a lot of our engineering work across the board. Um and like is a type to have a lot of strong opinions there. And then vice versa for go to market on my side.

Host

明白了。好的。但两个人都足够技术化,他可以参与对话,但我不会——我永远不会和 Ronic 参加黑客松比赛。嗯,我觉得他一生中从未输过黑客松。

Gotcha. Okay. But both technical enough where he can u conversations but I'm not going to I would never enter a hackathon against Ronic. Um I don't think he's lost a hackathon in his life actually.

Adit

哇。

Dang.

Host

你们也一起参加过黑客松,对吧?

You guys also did a hackathon together right?

Adit

是的。

We did.

Host

Anthropic 的那个,当时早期的 Quad 模型刚出来。

The anthropic one when quad the early quad model came out.

Adit

他们嗯给了 Quad 的早期访问权限。嗯,这实际上是我们创业旅程的开始。嗯,是的,所以当时 Redto 的整体框架就是在那时形成的,当时模型已经非常令人印象深刻了。

They were uh giving early access to Quad. Um that's actually sort of how our founding journey started. Um yeah, so back then that's where the whole like framing for Redto came in where models were already really impressive.

Host

是的。

Yeah.

Adit

嗯,但它们的上下文窗口限制很糟糕,所以这就演变成了长期记忆的想法。是的。以及后来 Redto 成为的一切。

Um but they had horrible context window limits and so that spiraled into this like long-term memory idea. Yeah. And everything else that reductive became after.

Host

明白了。是的。时代确实变了,因为当时 rag 非常火,你们仍然在做很多分块处理,对吧?现在随着窗口变大,分块可能不那么引人注目了。嗯,但很多问题最终都是搜索问题。好的,不管你是叫它 rag 还是别的什么,很多产品仍然是某种形式的搜索,然后在搜索结果之上进行某种推理。明白了。我个人认为从定义上讲,这就是检索增强生成。是的。没错。你能详细说说每个问题最终都是搜索问题这个观点吗?嗯,因为我知道即使在编码方面,ref 和文件搜索在 rag 之后也成了下一个热点,我猜对于文档解析和处理,框架是一样的吗?你们有什么不同做法吗?有什么见解可以分享吗?当然。

Gotcha. Yeah. Times have definitely changed because for sure like then rag was a very big thing and you guys still do a lot of chunking right with your processors it less so now with less windows are okay yeah like I think chunking is less of maybe what piques somebody's interest um but a lot of problems ultimately are search problems okay and whether you call it rag or you call it something else like plenty of products are still some form of search and then some form of inference on top of the search results Gotcha. Which I personally see as like definitionally it's kind of retrieval augmented generation. Yeah. Right. That's true. Can you double click on that with the notion of every problem kind of being still down to a search problem. Um because I know like even with coding like ref and file search kind of became next thing after rags and I guess for document parsing for processing is the framework the same do you differently? Are there like insights that you can share? Of course.

Adit

是的。所以嗯,我的意思是,很多时候当人们想到 rag 时,他们会想到更传统的方式:你把东西分块,嗯,你嵌入它们,然后你使用余弦相似度搜索来确定什么传入你的模型上下文。是的。嗯,人们对 rag 有反感,因为比如“哦,我不是在构建聊天机器人。”嗯。嗯,我们的很多客户仍然会有,你知道,比如一个智能体的文件系统,智能体在文件系统中爬行以找出相关上下文在哪里,然后它基于那个上下文进行推理,并把多个文档的信息拼凑起来。是的。嗯,我的意思就是那个过程,无论是智能体在做,还是在你的搜索产品中清晰可见,都是你在搜索信息,拼凑上下文,综合它,并在最后采取某种行动。

Yeah. So um I mean a lot of times when people think of rag they think of like the more traditional you chunk things um you embed them and then you use cosign similarity search to figure out what to pass into your models context. Yeah. Um and people have an aversion to rag on the basis of like oh I'm not building a chatbot. Mhm. Um, a lot of our customers will still have, you know, like a file system for an agent and the agents is crawling through the file system to figure out where the relevance contextes context is and like it's reasoning on that context and piecing together information for multiple documents. Yeah. Um, what I mean is just like that process, whether it's the agent doing it or whether it's, you know, clearly visible in your search product, is you're searching for information, piecing together that context, synthesizing it, and making some sort of action at the end of it all.

Host

我明白了。好的。是的,这说得通。这像是一个很直观的过程。嗯,让我们看看如果太好了。好的。我想我的部分完成了。我觉得你可能可以开始看了,但是的,我想现在更多是关于人们如何找到你以及你在寻找什么。我假设你们在招聘不同职位。有

I see. Okay. Yeah, that makes sense. It's like a pretty intuitive process. Um, let's see if Great. Okay. I think my stake is done. I feel like you can probably start looking at but yeah and I guess now going towards more how people can find you and what you're looking for. I assume you're hiring across different roles. Are there

Adit

我想我们现在有大约 40 个职位空缺。

I think we have like 40 roles open at this point.

Host

好的。你们现在的瓶颈是什么?主要是销售,还是你们仍在寻找很多研究人员?

Okay. What what are you bottlenecked by right now? Is it mainly sales or is it you still trying to find a lot of researchers?

Adit

各个领域都有。嗯,所以我们在工程方面招聘很多,比如基础设施问题对我们来说真的很重要,尤其是当我们开始规模化时。嗯,另外我确实想之后洗个手。是的。嗯,洗手时的错误。是的。所以我们在招聘产品工程团队、机器学习职位,嗯,还有平台。在销售团队,我们年初只有一个 AE 和几个席位。哇。现在我们销售团队有 30 人,或者哇。嗯,所以整体招聘很多。是的。嗯,我预计短期内不会放缓。就像我们团队仍然感到产能受限。

It's across the board. Um so we're hiring a lot on engineering like infrastructure problems really matter to us especially as we start to scale. Um also I do want to wash my hands after. Yeah. Um mistakes in the washing hand. Yeah. So we're hiring for our product engineering team, ML roles, um platform as well. On the sales team, we started the year with one AE and seats. Wow. And we now have 30 people on the sales or Wow. Um so a lot of hiring overall. Yeah. Um I expect that's not going to slow down in the near term. Like we still feel capacity constrained as a team.

Host

明白了。是的。嗯,那是个好位置。我觉得成长的挑战比处于转型地狱的挑战要好,正如你过去提到的,嗯,无法坐下来解决问题。是的。嗯,那比只是成长的时刻要难得多。所以是的,我觉得应该不错。是的。所以你可以转移到这里,然后我们可以洗。我以前从未用叉子煮过牛排。凡事都有第一次。太棒了。太好了。我们回来了。所以,我们的牛排准备好了,我们可以开始切了。

Gotcha. Yeah. Well, that's a good place to be in. I feel like challenge of growing rather than the challenge of being in pivot hell as you mentioned in the past of um not being able to sit down and solve a problem. Yeah. Uh is a lot harder than being at the moment where you just kind of grow. So yeah, it should be good, I think. Yeah. So you can transfer here and then we could wash. I've never cooked a steak with a fork before. first for everything. Amazing. Great. And we're back. So, our steaks are ready and we can start cutting.

从解析到编辑 From Parsing to Editing

Host

你想切面包,我来切牛排?不过,还是说回 Reducto,你们已经完成了这个转型:从企业可以来提取和解析文档的地方,到现在甚至能处理编辑,对吧?所以你觉得有很多机会,或者正在转向承接整个用例,因为我猜从阅读到现在真正分析和编辑以及知识工作,可以覆盖很多用例。

Do you want to cut the bread and I'll cut the steaks? But yeah, I guess more so again on Reducto, you guys have made this transition from being the place where companies can come to extract and parse documents to even now handle editing, right? And so do you see a lot of opportunities or transitioning to taking on this whole use case because I assume going from reading to now like actually analyzing the editing and the knowledge work can cover a lot of use cases.

Adit

是的。我总体的看法是,我不认为白领工作已经迎来了类似 Quad Code 那样的时刻。而且我觉得,如果你想想很多白领工作,它不只是信息综合,比如人们从文档中提取信息,

Yeah. My general view here is I don't think white collar work has quite had the equivalence of like a quad code moment. Um, and I think if you think of a lot of white collar work, it's not just information synthesis, like people pulling from documents,

Host

但我们把这些三明治做多大?

but also how big are we making these sandwiches?

Adit

嗯,你想多大就多大。

Um, as big as you want.

Host

我们可以直接切开。

We could just cut it.

Adit

是的。

Yeah.

Host

好的。

Okay.

Adit

嗯,不只是人们从文档中提取信息,而且在这些流程的最后,你会写入电子表格或 Word 文档,不管是什么,对吧?所以我们真的想确保我们不只是考虑最初的解析和提取层。比如

Um, it it's not just people pulling from documents, but also like at the end of a lot of these processes, you write into a spreadsheet or you write into a word doc, whatever it is, right? Um, and so we really wanted to make sure that we are not just thinking about that initial layer of parsing and extraction. Like

Adit

我们希望让模型能够尽可能可靠地红线修订合同。嗯,我们希望它们能够

we want to make it possible for models to redline contracts as reliably as possible. Um, we want them to be able to

Adit

填写贷款申请,做任何人类必须做的事情,对吧?

fill out loan applications, do whatever a human would have had to do, right?

Host

是的。

Yeah.

Adit

嗯,那里出现了一系列问题,真的很难做到。这是一个很好的切入点。

Um, and there's a range of problems that comes up there that is just really hard to do. That's a great stake.

Host

是的。嗯,就像我们之前谈到的一件事,人类每年在文档本地化上花费数十亿美元。

Yeah. Um, like one of the things we were talking about earlier is humans spend billions of dollars per year on document localization.

Adit

呃,所以如果你是美国的企业,想和中东的公司做生意,嗯,你可能需要把内容翻译成阿拉伯语,对吧?

Uh, and so if you're a business in the US and you want to do business with a a company in the Middle East, um, you probably are going to want to translate your content into Arabic, right?

Host

如果你想想今天的情况,这非常手动。模型在翻译方面很擅长。

And if you think about that today, it's very manual. Models are great at the translation.

Adit

是的。但将编辑应用到文档真的很难,因为你需要拿一份用英文生成的文档并更改布局。比如当你把它改成从右到左的语言时,现在你需要考虑文本可能会流到下一页。

Yeah. But applying that edit to a document is really hard because you're going to need to take a document that was generated in English and change the layout. Like when you change it to a right to left language, now you need to consider like maybe the text flows over onto the next page.

Host

格式不同了。

Formatting is different.

Adit

没错。嗯,所以这里有一整套问题,理想情况下我们可以为客户抽象掉。

Exactly. Um and so there's a whole set of problems here that ideally we can abstract away for our customers.

Host

是的。

Yeah.

Adit

嗯,我们仍处于那个阶段的早期。我们不久前推出了表单编辑端点。嗯,到现在快 6 个月了。

Um we're still early in that phase. We launched a form editing endpoint not too long ago. um close to like 6 months at this point.

Adit

嗯,但我们有很多即将发布的东西,我认为这将会是一个子功能的变化。

Um but we have a lot that we're about to release soon that I think will be a sub function change here.

Host

明白了。不过这听起来像是一个相当不同的问题。听起来很像计算机使用,比如如果你在填写

Gotcha. This kind of sounds like a pretty different problem though. It sounds very like computer usees like if you're filling out

Adit

肯定有一些相似之处。嗯,这里肯定也有很多更接近代码生成的工作。嗯,但我认为从根本上说,非常非常擅长理解视觉结构之类的东西

there's some parallels for sure. Um there's definitely a lot of work that's closer to codegen here as well. Um, but I think fundamentally being really really good at understanding things like visual structure

Adit

使我们也有可能非常擅长编辑应用之类的事情。

makes it possible for us to also be really good at things like the edit application.

Host

明白了。好的。那么,你们是统一处理这个问题,还是为这些不同问题设有独立团队,还是都在一个 umbrella 下?

Gotcha. Okay. So, do you kind of treat the problem or do you have like separate teams for these different problems or is it all under one umbrella?

Adit

呃,我们并没有真正有意义地划分团队。也许这涉及到我们之前谈到的,你知道,我们如何为工程团队构建优先级等等。

Uh, we don't really separate teams that meaningfully. Maybe this touches on what we were talking about earlier with, you know, how do we structure prioritization for the engineering team and all that.

Adit

嗯,我们有一个 ML 团队,但在那个 ML 团队内部,人们研究各种不同的问题。

Um, we have an ML team, but within that ML team, people work on all sorts of different problems.

Adit

肯定有人专攻某些类型的事情。我们刚刚完成了一次非常棒的招聘。嗯,

They're definitely people that specialize in certain types of things. We just made a a really awesome hire. Um,

Adit

他叫 Leang。你可能见过他在 Twitter 上因为开源办公智能体而爆火。

his name is Leang. You might have seen he went super viral on Twitter for open source office agents.

Host

哦,我想我看到了。是的。

Oh, I think I saw this. Yeah.

Adit

是的。嗯,真的很难招到。他刚从新加坡搬过来,

Yes. um really hard hire to make. He just moved over from Singapore,

Adit

但他对这类编辑问题非常兴奋,所以他在很多方面带头。就像那种个人问题热情的一致性非常重要。

but like he's really excited about this sort of editing problem and so he's taking the lead on a lot of that. Like those sorts of just person problem enthusiasm alignments is really important.

Host

那么你如何找到对使命同样有雄心的人?是大量 outbound 还是 inbound?嗯,你们会以特定方式筛选人吗?是的。

And how do you find the right people who are um as ambitious, let's say, for the mission? Is a lot of outbound or inbound? Um do you kind of like filter out folks like a specific way? Yeah.

Host

嗯,所以他们非常使命驱动。我猜这是你在招聘中考虑的一个非常重要的因素。

Um so that they're very like mission driven. I assume that's like a very big thing you consider in hiring.

Adit

是的,我们有稳态招聘。比如我们有内部招聘人员,与机构合作,所有这些。但每隔一段时间,你会在 Twitter 上发现有人奇怪地痴迷于 PDF,嗯,频率比你预期的要高。

Yeah, we have steady state hiring. Like we have recruiters in house, we work with agencies, all of that. But every once in a while you find somebody on Twitter that is weirdly obsessed with PDFs um at a more frequent cadence than you would expect.

Host

是的。

Yeah.

Adit

所以每当我们看到类似的情况,这很能说明这个人,他们随意地花时间思考文档处理。是的。

And so like anytime we see something like that, like it says a lot about the person that they're spending their time thinking about document processing just willy-nilly. Yeah.

Adit

我们总是会尝试联系。

We'll always try to get in touch.

Host

是的,这有道理。是的。我猜当你说你们还没有迎来 Claude Code 时刻。

Yeah, that makes sense. Yeah. And I guess when you're talking about you haven't had a Claude Code moment.

Adit

是的。

Yeah.

Host

嗯,你认为什么会成为类似寒武纪大爆发的催化剂?仅仅是看到用例有多大价值,还是更像是特定行业的扩散问题,也许只是编辑用例的更多采用?

Um what do you think would be a catalyst for that like Cambrian explosion? Is it just a matter of seeing how valuable the use case could be or is it like more so diffusion problem of like specific industries and maybe just more adoption for like the editing use case?

Adit

是的。嗯,我认为模型智能就在那附近。嗯,很多前沿实验室显然在考虑模型要解决的其他问题之前,先攀登了代码生成。

Yeah. Um one I think model intelligence is right around there. Um and a lot of the frontier labs clearly hill climbed on codegen before they were thinking about some of the other problems for models to solve.

Host

是的。

Yeah.

Adit

嗯,我认为代码在很多情况下是一个可验证的问题,嗯,工程师最有可能采用,肯定有很大一部分

Um I think whereas code is in many cases like a verifiable problem um and engineers are the most likely to adopt there's definitely like a large component of

Adit

企业将数据放在各种不同的地方,而且有一些奇怪的启发式方法,比如你和大型金融机构谈论某个纸质流程是如何完成的

enterprises have their data in all sorts of different places and there are like weird heruristics of you talk to a large financial institution about how like some paper process is done

Adit

他们就会有机构知识,比如哦,如果我在屏幕左下角看到这个,我知道它来自肯塔基,因此或者我做 X。

and they'll just have institutional knowledge of like oh if I see this in the bottom left of the screen I know it's from Kentucky and therefore or I do X.

Host

是的。

Yeah.

Adit

嗯,这种情况很少被清楚地写下来。

Um it's rarely the case that that is clearly written down.

Adit

嗯,所以我认为有很多工作要做

Um and so I think there's a ton of work for

Adit

工具领域的人要做的事情,比如聚合上下文,帮助他们构建适当的知识库,给模型足够的上下文以便它们能够对数据进行推理。

people in the tooling space to do to like aggregate that context, help them build the appropriate knowledge bases, give models the sufficient context for them to be able to reason on that data.

Host

是的。

Yeah.

Adit

嗯,但我不认为我们离那个点很远。好的。

Um but I don't think that we are far from a point where that happens. Okay.

Host

大规模地。

At scale.

Host

明白了。所以听起来更像是数据和上下文的碎片化似乎是问题所在,而不是性能问题,

Gotcha. So it sounds more so like a fragmentation of the data um and context that seems to be the issue more so than performance that

Adit

模型在长程推理方面已经变得如此出色,以至于智能是最大瓶颈的可能性越来越小。

models have gotten so good at long horizon reasoning that it feels less and less likely that intelligence is the biggest bottleneck.

Host

是的。

Yeah.

数据挑战与垂直方案 Data Challenges and Vertical Solutions

Host

那么在这种情况下,这是否需要我们更加集中精力?因为我觉得,处理来自俄亥俄州这篇论文的所有这些不同数据源——他们有很多老式文件系统——并不是软件一定能解决的。所以也许通过文档处理来解决并不是件简单的事。你有不同的思考吗?

And so in that case, is that something that we need a more concentrated effort? Because I feel like wrangling all these different sources of data from this paper in Ohio where they have all these older file systems isn't something that software necessarily can solve. So maybe this is not something that's trivial by approaching it through document processing. Have you thought about it differently?

Adit

是的。我的意思是,很酷的一点是,我们能够与各个不同垂直领域的许多领先团队非常紧密地合作。所以我们一直在与 Basis 团队合作,他们正试图自动化 N10 税务申报。我们与 Harvey 和 Lora 等公司紧密合作,他们在法律领域做着令人难以置信的工作。他们专业化的好处之一是,他们开始弄清楚在特定情境下,端到端交付价值需要什么。这很棒。

Yeah. I mean the cool thing is we get to work really, really closely with a lot of the leading teams in all sorts of different verticals. And so we've been working with the Basis team, who's trying to automate N10 tax filing. We work super closely with companies like Harvey and Lora who are doing incredible work in the legal space. And part of the great thing about them specializing is that they start to figure out in a certain context what it takes to deliver that value end to end. Which is great.

Host

明白了。好的,有道理。

Gotcha. Okay, that makes sense.

Adit

这是小菜一碟。

It's steam a piece.

Host

哦,好的。很好。

Oh, okay. Good.

Adit

嗯。

Mhm.

Host

好的,我们可以组装一个,然后吃掉,然后就这样。好的。是的。我想问,除了这次发布之外,还有没有其他里程碑或你期待的事情?再次祝贺你,并祝今年剩下的时间一切顺利。

Okay, we could assemble one and then eat it and then call it. Okay. Yeah. And I guess are there any other milestones or things that you're looking forward to outside of this launch? Congratulations again for the rest of the year.

Adit

是的。总的来说,我们的模型正在变得越来越好。

Yeah. Across the board, our models are getting much, much better.

Host

你想要光环吗?

You want the Halo?

Adit

是的。你想和他们一起戴上这个吗?

Yeah. You want to put this on with them?

Host

当然。

Sure.

Adit

是的。所以我们的结构化提取模型也有了巨大的改进。这不仅仅是准确率的问题。我们在引用性能和模式遵循等方面做得更好了。而且成本效益更高。我们今天宣布了这一点。历史上,一页结构化提取的总成本大约超过 4.5 美分。现在我们可以以一半的成本完成,同时提高性能。

Yeah. So we have a massive improvement to our structured extraction models as well. This isn't just accuracy again. We're getting better at things like citations performance and schema adherence. And are doing that at a much more effective cost. We announced this today. Historically a page of structured extraction would have cost you all in like over 4.5 cents. And now we're able to do that at half the rate while also improving performance.

Host

哇。

Oh wow.

Adit

你没事吧?

Um you okay?

Host

我噎住了。

I choked up.

Adit

是的。哇。这真是大幅削减。

Yeah. Wow. That is a big cut.

Host

是的。所以这些就是我们想要推进的事情。

Yeah. And so those are the sorts of things that we want to push on.

Adit

明白了。

Gotcha.

Host

就像真正推动整个领域的 Purto 前沿。

Like really driving that Purto frontier across the board.

Adit

明白了。你之前谈到了成本和性能。所以,这主要是关注成本,还是只是降低成本同时保持性能在之前的水平?

Gotcha. And you were talking before about cost and performance. So, is this mainly a cost focus or just like bring it down while keeping performance at the level that was before?

Host

实际上,幸运的是,这甚至不是因为我们试图降低成本本身。我们只是发现,我们可以获得更高的每美元智能,而恰好只是专注于性能本身。就像我们的主要 Nordstar 将始终是最准确的文档处理平台。

Actually, fortunately, this is not even from us trying to drive cost in and of itself. We're just finding that we can get to better intelligence per dollar while happening to just focus on the performance itself. Like our main Nordstar is always going to be the most accurate document processing platform out there.

Adit

嗯。而且正如你所说,也是模型无关的。

Mhm. and being model agnostic as well like you said.

Host

好的,完美。

Okay, perfect.

Adit

好的,太好了。

Okay, great.

制作牛排三明治 Making a Steak Sandwich

Host

好的,你想放多少?你想再放一些吗?

Okay, how much do you want to put on? You want to put some more?

Adit

我们应该做两层。

We should do like two layers.

Host

好的,这里我们可以做,你想,我可以移到这边。也可以加一些。

Okay, here we can do you want to I could move on this side. Can also add some.

Adit

太好了。

Great.

Host

太好了。好的。你感觉怎么样?

Great. All right. How are you feeling?

Adit

很好。我在想,我忘了他的用户名。但我想确保我这样分层的方式是

Good. I'm thinking about the I'm forgetting his username. But I want to make sure I layer this in a way that's

Host

是的,你不会再被比例吐槽了。

Yeah, you don't get ratio again.

Adit

不,不是那个。这些也是非常小的切块。所以,是的。

No, not that. These are also very small cuts. So, yeah.

Host

应该能很好地分开。

Should pull apart well.

Adit

太好了。

Great.

Host

太好了。它们看起来很棒。是的。

Great. They look great. Yeah.

Adit

把 bar 放上去。

Put the bar on.

Host

是的。你想放 bara,我来放 teacher?

Yeah. Do you want to put the bara and I'll put on the teacher?

Adit

好的。哦,哇。那很好。你把它放在牛排上还是面包上?

Okay. Oh, wow. That is good. You put it on top of the steak or on top of the bread?

Host

嗯,可能放在牛排上。好的。是的。然后我们就把它盖上去。是的。

Um, probably the steak. Okay. Yeah. And then we just top it off. Yeah.

Adit

涂开。是的。

Spread it. Yeah.

Host

哦,太好了。看起来太好了。

Oh, great. That looks so good.

Adit

太好了。

So good.

Host

是的。那是 PGs。

Yeah. That's PGs.

Adit

好的。然后你想做 Chimmy Cherry 吗?

Okay. Then do you want to do Chimmy Cherry?

Host

当然。是的。混合一下,然后放在上面。哦,我的天。

Sure. Yeah. Mix it and then just put it on top. Oh my goodness.

Adit

这些动作镜头会非常疯狂。

These action shots are going to be crazy.

Host

是的,看起来很冷。

Yeah, looks really cold.

Adit

你也想在上面做吗?你可以把面包放在上面。让我们来,是的,很好。

Do you also want to do it on top? You could put the bread on top action. Let's get the Yeah, nice.

Host

好的,完美。让我们切开它。

Okay, perfect. Let's cut it.

Adit

嘿,我们能回应一下批评者吗?

Yo, can we address the critics?

Host

是的。

Yeah.

Adit

是的。我觉得我们应该数一下你咬的时候出来的牛排片数。据称会是三片。

Yes. I think we should count the number of steak slices that come out as you bite. It's alleged that it's going to be three.

Host

嗯。

Mhm.

Adit

来吧。

Go for it.

Host

一。

One.

Adit

是的。简单。好的。

Yeah. Easy. Okay.

Host

味道怎么样?

How's it taste?

Adit

太棒了。

Amazing.

Host

很好。好的。让我们干杯。

Good. Okay. Let's cheers.

Adit

嗯。

M.

Host

哦,还有辣味。

Oh, spicy, too.

Adit

嗯。太好了。哦,我的天。哇。牛排三明治永远不会出错。

Mhm. That's so good. Oh my goodness. Wow. You can never go wrong with a steak sandwich.

Host

是的,桃子。

Yeah, the peaches.

Adit

嗯。伙计,这太棒了。我们做得很好。

Mhm. Man, this is great. We did a great job.

Host

谢谢你邀请我们来。

Thanks for having us over.

Reducto的隐喻 Metaphor for Reducto

Host

那么,这怎么

So, how does this

Adit

那么,这个三明治里 Reducto 的隐喻是什么?

So, what's the metaphor for reducto in this sandwich?

Host

你想让我做一个与文档相关的评论吗?

You want me to make a document related comment on

互动版:逐字朗读 + 针对本期提问 →