为什么大家都讨厌软件开发人员？

很有见地。你是个聪明人，达克斯。

It's good insight. You're smart guy, Dax.

别说了。别说了。停下。你知道我喜欢做什么吗？我敢肯定大家都喜欢这个，但我特别喜欢拿到商品的零售包装盒。

Stop talking. Stop talking. Stop it. You know what I love doing? I'm sure everyone loves this, but I love getting like the retail box of a thing.

是啊。是啊。

Yeah. Yeah.

对，这样感觉就像把它带回家了，你懂吧，

Yeah. So then it feels like I have it at home, you know,

嗯哼。

like Uh-huh.

我觉得它

I think it

是啊。就在店里这里。

was Yeah. Here at the store.

嗯，是的。我拿到了

Uh-huh. Yeah. I got

我经常这么做。

I do this a lot.

对，那边那些LARABARS也是这样的。

Yeah. That's the thing of LARABARS over there too.

是啊。哦，你现在还吃LARABARS吗？我已经很久没吃了

Yeah. Oh, do you still do Larabars? I haven't done

我有一阵子没吃了，但后来Liz又突然买了一些。

them I a haven't for a while, but then Liz randomly bought them again.

它们真的很好吃。

They're so good.

她早上需要吃点东西。

She's needed something in the morning.

是啊，尽管他们没有赞助我们，但他们的产品真的很棒。

Yeah. Despite them not sponsoring us, they're so good.

我每次健身后都会喝这种电解质水。

I've been drinking this electrolyte water every time

它叫什么名字？

What's it called?

从健身房回来时喝的。Element？

To get back from the gym. Element?

对，我喝过Element。跟我说说Element吧，我们可是老交情了。

Yeah. I took Element. Tell me an Element. We go way back.

不错。Liz买这个有一阵子了，我喝了之后发现自己只喜欢青柠味的，我还以为他们不可能单卖青柠味的整箱呢。结果他们真的有青柠整箱卖。

Nice. So Liz has been getting them for a while and I would drink them and I realized I just like the lime ones and I was like, man, there's no way they sell a lime box. But they do just They a lime box. Yeah.

我买的是葡萄柚味的整箱，强烈推荐葡萄柚口味。好的，我会试试那个味道。确实很好喝。

I buy boxes of the grapefruit, which I highly recommend the grapefruit flavor. Okay. I'll try that one. It's delicious. Yeah.

我真不敢相信你喝Element饮料却不听播客。因为我觉得每个喝Element的人都是从播客广告里听说的，他们到处打广告。

They I can't believe you drink Element and you don't listen to podcasts. Cause I feel like everyone who drinks Element has just heard it on all the podcast ads because they advertise

哦，它经常打广告吗？

Oh, it's like advertised?

哦，是的。

Oh, yeah.

好吧，人们正在这个播客上听到它。

Well, people are hearing on this podcast.

嗯，这倒是真的。但区别在于我们播客和大多数播客不同——我们没拿广告费。那些大播客都在疯狂推广Element。

Well, that's true. Yeah. But we're not getting paid for it, which is the difference between our podcast and most podcasts. No. Like all the big podcasts, they all advertise Element heavily.

哦，有意思。不过它确实不错。神奇的是我喝了一段时间后觉得这就像糖水，但其实里面根本没糖，全是人工甜味剂。

Oh, interesting. Well, it's good. What's crazy is I was drinking it for a while and I was like, oh, this is just like sugar water. But then there's no sugar in it. It's all it's like fake sweetener.

对，对。盐水。

Yes. Yes. Salt water.

所以它不像佳得乐那种饮料。

And so it's like not like Gatorade.

是啊。好吧。人工甜味剂。这个话题有意思。我觉得科技在不断进步。

Yeah. Okay. Fake sweeteners. Here's a topic. I feel like science continues to advance.

市面上有太多不同的人工甜味剂了。除了阿洛酮糖，其他都是垃圾。你对这个有什么看法吗？

There are so many different fake sweeteners. They're all garbage except for Allulose. Do you have an opinion on this?

我认为关键不在于甜味剂本身，而在于如何使用。比如我用罗汉果做的食物味道就不错。有些用甜菊糖的东西...这个是什么来着？这个是甜菊糖，我通常...

So I think it's less about the sweetener and more about how it's used because I've had things with like monk fruit that's tasted good. I've had things with So that that tastes that's tasted so what is this? This So this is stevia and I usually

最讨厌那种...比如电解质饮料，咸鲜味的饮料本来就不需要太甜。如果是甜点这类需要甜味的东西，木糖醇或赤藓糖醇之类的简直难以下咽。一尝就知道是假甜，就像健怡可乐——至少是我小时候最后一次喝的那个味道。

hate If they just if they just like So element drinks, salty, delicious drink, doesn't need to be that sweet. If it's something that needs to be sweet, like it's a dessert or something, then like xylitol or erythritol, whatever, awful. Like, I just you can tell it's just fake sweet. It's like Diet Coke, at least what Diet Coke used to be when I had it last time when I was a kid. The yeah.

阿洛酮糖应该是这个领域最新的发现，因为科技进步嘛。据我所知它没有任何副作用，不会影响消化。像糖醇类就会让人肠胃不适，我就有点受不了。

The the Allulose stuff is the most recent, I think, most recent thing in this world because science probably figured it out. It doesn't, like, it doesn't have any side effects that I know of. Like, doesn't affect your digestion. Like, I think the sugar alcohols can mess with people. Kinda messes with me.

阿洛酮糖完全没有营养成分。你的身体直接忽略它，这挺神奇的。但它的甜度确实很高。

Allulose, like, has no nutritional profile. Your body just ignores it. Just kinda cool. But it's quite sweet.

它会有那种假甜味剂的后味吗？不，通常

Does it have like that fake sweetener aftertaste? No. That's usually

就像我说的。它的口感就像他们甚至把它做成粉末状出售，你可以像撒糖一样撒在食物上，感觉就像真的在吃糖

what saying. It's got the profile of like they even sell it in like a powder that you like sprinkle on stuff like sugar and it just feels like you have sugar

罗汉果甜味剂也是以同样的方式出售的。我应该...我应该

on it. They sell monk fruit sweetener in that same way. I should I should

是的，罗汉果是另一种不错的选择。

Yeah. Monk fruit's another good one.

我可能吃过一些含有它的东西。

I've probably had some I've probably had something with it in there.

好吧。人工甜味剂等级排行。开始吧。开玩笑的。

Okay. Artificial sweetener tier list. Let's go. I'm just kidding.

但这听起来像是阿洛酮糖是最好的，而罗汉果也相当不错

But that means It sounds like aliolus is the best and it sounds like monk fruit is pretty good

说到罗汉果，那可是S级的。其他的都是F级，简直垃圾。

by the I was in monk fruit are up there, s tier. The rest of them, f tier, dog water.

你吃过那种——哦我忘了牌子——基本上就是仿制普通糖果，用代糖做的，价格却贵十倍的东西吗？就是那种仿瑞典鱼糖的假货。对。

Have you had those it's like oh, I forgot the name of the brand, but they basically copy normal candy but it's made with fake sweetener and costs like 10 times more. Is that what that is? They have like the Swedish Fish, like the fake Swedish Fish. Yeah.

哦不，那是Smart Sweets。哇，这个我不太了解。嗯。

Oh, no. That's Smart Sweets. Wow. Don't know much about this. Yeah.

Smart Sweets啊。我们给孩子买过。其实我不太喜欢水果软糖，也不是假装不吃，就是真的不爱。

Smart sweets. Yeah. We get those for the kids. I don't really I'm not really into fruit gummies. I'm not like playing it coy like I don't eat them.

我只是不太喜欢...我是个巧克力控。巧克力焦糖味的。嗯。

I just don't love I thought I'm a chocolate guy. Chocolate caramel Yeah.

是啊，我喜欢换着吃。我不挑食。

Yeah. I like to mix it up. I don't disjuvenate. The

Joyride就像...有那种仿Twizzlers的假货

Joyride is like the have like Joyride. Fake Twizzlers and fake

它们好吃吗？

Are they good?

天啊，这些是真的糖。不像假糖，所以我也不确定。

Oh my god. These are real sugar. These are not like a fake sugar, so I don't even know.

哦，所以它们只是山寨货？不是健康食品？不，它们是

Oh, so they're just bootleg? Not like a health thing? No. They're

它们宣传说是含糖量减少了80%。

they're marketed as like 80% less sugar.

好吧。

Okay.

我觉得它们确实添加了其他甜味剂，但里面确实有真糖，而Smart Sweets我认为一点都没有。

And I think they do supplement with some other sweetener, but they do have real sugar in them, whereas Smart Sweets I think don't have any.

我是说，减少80%意味着我能多吃80%的量。

I mean, 80% less means I can eat 80% more.

我知道。这正是我想说的。什么？所以他们做了这些像绳子一样的糖果袋。它们像是樱桃味的绳子糖。

I know. That's what I'm saying. What? So they make these bags of these like ropes. They're like cherry ropes.

他们可能还做了其他口味。这些糖非常好吃，整袋都是。是的。

They make other flavors probably. They are so good and the entire bag Yeah.

我

I

热量很高，吃完感觉就像吃了一整袋扭扭糖。我是说，它们非常令人满足。你应该试试Joyride，它们真的很好吃。

got calories and you feel like you just ate an entire bag of like Twizzlers. I mean, it's like they're very satisfying. You should get some Joyride, they're they're delicious.

Joyride糖果。我没听说过这个牌子。

Joyride candy. I haven't heard of this one.

说到营养极度匮乏的糖果，这个词很贴切。按它的热量来看，简直难以置信你吃下去的能量其实没那么多。有种巧克力棒叫Gatsby，你可以查查。

While we're talking about candy that is like crazy nutritionally sparse is the word. Like, calorie for what it is. Like, you can't believe when you're eating it that it's not that much energy in it. There's these chocolate bars, Gatsby. Look them up.

Gatsby巧克力棒。好的。我最近...太热了没法户外跑步，所以我一直在用Liz健身房的访客通行证，那是家Equinox。他们前台有精选的蛋白棒。

Gatsby chocolate bars. Okay. So I just Delicious. I I've been it's gotten even too hot for me to run outside, so I've been going to I've been using the guest passes at Liz's Gym, which is an Equinox. And they have like a selection of protein bars in the front.

有一个看起来特别像马戏团风格的，简直就像是MrBeast会卖的那种产品。

And there's one that looks like it just looks so, like, circus y and, like, over the it kinda looks like something that MrBeast would sell.

TrueBar？所以我明白了。是TrueBar吗？

TrueBar? So I got it. TrueBar?

不，不是TrueBar。好吧，我尝过了，味道特别好，完全就像直接吃糖果棒一样。

No. It's not TrueBar. Okay. I got it and it tastes so good. It tastes just like a straight up candy bar.

这是我见过最滑稽的质地了，我从来没吃过这样的食物。很明显这不是正常食物——它像橡胶一样有弹性，感觉像融化的橡皮擦。但看了成分表，其实对身体无害。只是无论如何，食物的质地不该是这样的，你懂吗？

It is the most hilarious texture I've ever like, I've never eaten a food. It's, like, so clear that this is not food because it's, like, it's, like, rubbery and it stretches. It feels like like a melted eraser or something. But I looked at the ingredients and they're like, it's not like it's, like, bad for you. It's just, for whatever reason, it's just it's just it's like it it just there's no way food was supposed to be this consistency, you know?

他们在不断学习。比如蛋白棒就是个好例子——他们正在研究如何让蛋白棒不再像纸板一样难吃，也不再是那种可怕的质地。所以我之前提到科学——感觉现在食品科学正在发展，他们不断找到方法让食物更美味，同时减少对健康的危害。

They're learning stuff. I like they're learning how so protein bars are a good example. They're learning how to not have protein bars taste like cardboard and not just like have that terrible texture. That's why I said science earlier. Like, I feel like there is like a food science thing going on where they just keep finding ways to like make food tastier, but not quite as bad for you.

问题是，我们根本不知道。我们的身体从未接触过这些东西。人类以前不需要消化阿洛酮糖，我们究竟了解多少呢？

Well, the like, we don't even know. Our bodies have never had this stuff in our body. You know what I mean? Like, humans haven't had to digest allulose before. What do we really know?

对，这款蛋白棒叫Bilt，品牌名就是Bilt。

Yeah. So this protein bar is called a it's Bilt. Name of Bilt.

那好吧。

Then Okay.

我买了这个布朗尼面糊味的，还自己订了一整盒。对。不过，成分表第一位就是胶原蛋白混合物。里面最奇怪的成分大概是棕榈油，但它在配料表靠后的位置，而且很多食品都含棕榈油。主要是胶原蛋白，然后我觉得是明胶让它有这种诡异的——

I I get this brownie batter one and I just ordered like my own box. Yeah. But yeah, the ingredient the first ingredient is just collagen protein blend. The weirdest thing in it is, like, palm oil, but like, it's closer to the bottom and like so many things have palm oil. So collagen and then like, I guess a gelatin is what gives it this like crazy like

像橡胶一样的。

Like a rubbery.

对，但它是软的。所以我也说不清。对我来说，虽然知道这很不天然，但我就是超爱。超喜欢这种怪异口感，而且味道超棒。

Yeah. But it's like soft. So I don't know. For me, it's like I know it's so unnatural, but I just love it. Like, I love how weird it feels and it tastes delicious.

嗯。

Yeah.

是啊，真有趣，

Yeah. It's funny how

竟然会这样。

that works. With it.

我注意力很不集中。凯西正在我头顶上吸尘。就是

I'm very distracted. Casey is vacuuming right above me. Like

听声音根本不像在吸尘，我几乎听不见。几乎听不见。听起来像是背景里轻微的刮擦声。

The way it sounds, it doesn't sound like vacuuming and I I barely hear it. I barely hear it. It just sounds like low scraping in the background.

是啊。她拿着那种小型手持吸尘器，就是那种快速吸尘的。你们家有四台吸尘器吗？我们家的吸尘器多到离谱。

Yeah. She's got like the little hand vac just like it's the the quick the quick vac. Do you have like four vacuums? We have like so many vacuums.

我们家房子比较小，所以通常没法...嗯...存放太多重复的东西。不过这些年来我们用过好多吸尘器，现在用的美诺（Miele）很满意，就是忘记怎么发音了。

We've I mean, our house is smaller so we usually can't Yeah. Store too many of the same thing. Yeah. We've through so many vacuums over the years though, but we're really happy with what we have now, which is Miele. I forgot how you pronounce it.

M-I-e-l-e。读起来像是

M I e l e. It's like

我都不知道他们还生产吸尘器。我记得他们做的是厨房电器之类的吧。

I don't know they made vacuums. They make like kitchen appliances, I think.

没错。所以我们试过好多戴森...戴森根本是诈骗品牌。我现在非常确信这一点。

Yeah. Exactly. It's not like so we went through so many Dyson Dyson is a scam brand. I I'm I'm like convinced.

你听到了吗？

Did you hear that?

嗯。

Yeah.

我觉得那是吸尘器掉下来的声音。

I think that was the vacuum falling.

是的

It was

是戴森的。所以也许在戴森坏掉后，我就只能吃一顿饭了。

a Dyson. So maybe I'll just get a meal meal after a Dyson breaks.

我们经历过太多戴森吸尘器了，因为它们总是坏。它们看起来很酷，你会觉得，哦，它是无线的，还有各种功能，但我认为它们就是不行。我觉得我们都只是被营销骗了。后来我们换成了这款带集尘袋的，一开始觉得‘啊，还要用袋子’，但实际上我们很少需要更换。就算要换也超级简单，比戴森那些破事好多了。

The the I we've been through so many Dyson vacuums because they just constantly break. They look cool and you're like, oh, it's cordless and, like, you know, all this stuff, but I think they just suck. I think that we all just fall for the marketing. And then we switch to this one, and it has a bag, which seems like, oh, no, a bag, but, like, we rarely ever have to change it. And when we do, it's just like, it's super easy and it's way better than whatever shit happens with the Dyson.

是啊，我不喜欢戴森。感觉每次用的时候都像要散架一样。对，就是个大塑料玩意儿。

Yeah. I don't like the Dyson. I mean, it's just like a giant it just feels like it's gonna break every time I use it. Yeah. It's this giant plastic thing.

是啊，还有它的内部构造。但就像我们这儿有人打翻了东西，地板已经清理过了。我们用的是中央吸尘系统，可以打扫整个房子。

Yeah. And like the inner workings of it. But it's just like our like somebody spilled something right there and we've already done floors. We have like a central vac that we do the whole house with.

你喜欢中央吸尘系统吗？我记得小时候父母家里就有。等等，你家的中央吸尘是怎么运作的？我父母家那个很烦人，墙上到处是插槽，还有根巨长的软管

Do you like the central vac? I remember my parents had it in our house growing up and Okay, wait, how does your central vac work? The one in my parents house was annoying because it just had the slots in the wall everywhere and then there was like this giant like cord

可以插进去的那种。

that you can plug in.

对，可以插进去的。

Yeah. They can plug in.

没错，但我们家的不是那样运作的。这可是生活小窍门——如果你要建房子，就该这么装中央吸尘系统。我们装过两次了。

Yeah. So that's not how ours works. And this is the the life hack. If you ever build a house, this is how you do central vac. We've done it twice.

他们有这种藏在墙内的软管系统，实际上是铺在地板龙骨里的。软管嵌在墙里，我们家有四个不同位置的接口，你只要接上吸头，软管就会从墙里拉出来。

They they have these hide a hose that go in the walls. They actually go in the floor joists. So the hose is in the wall. There's like four different points in our house that have a hose and you just like bring the attachment and connect the hose and the hose drags out of the wall.

太棒了，这设计真不错。是啊，虽然中央吸尘概念很合理，但这种软管设计太糟糕了。

It's amazing. That is really nice. Yeah. Because I was the central vet concept makes a lot of sense, this hose thing sucks.

是啊是啊。如果管道在墙里，我确实对吸尘器很了解。我只是在想我这辈子在这件事上花了多少心思，简直多得离谱。就像

Yeah. Yeah. If the hoses are in the wall, I know I know a lot about vacuums. I'm just thinking about how much I've thought about this in my life and it's way too GD much. Like

亚当就喜欢吸东西。

Adam loves to suck.

就那两次，我是说我们建造中央吸尘系统的两次，那就像是个重大决策点，还有各种不同的...我也不清楚。但我们还有两个机器人吸尘器，嗯哼，从来没用过。我不明白既然有中央吸尘系统为什么还要买它们。

The two just the the two I mean, the two times we built, like, the central vac was like this big decision point and like all the different I don't know. But we also have the two robot vacuums that are like Mhmm. Never used. And I don't know why we have those if we have central vac.

我们也有个机器人吸尘器。它就放在我架子底下，我也从来不用。以前养宠物时经常用，因为宠物掉毛太严重了。在纽约时它每天早上都工作，对减少地上的毛发起了很大作用。但现在我们每周都请人来家里打扫，所以...

We have a robot vacuum too. It just lives under my shelf there and I never use it also. We used to use it a lot with pets because the pets the pets are a huge thing. Like in New York, it ran every morning and it was like a huge it made a huge difference with how much just stuff was on the ground. But now we just we have someone come to our house and clean every week, so Yeah.

这就不是什么大问题了。

It just hasn't been much of an issue.

我觉得大家装中央吸尘系统或买机器人吸尘器时都想着'再也不用自己吸尘了'。但不知怎么的，用着用着就闲置了。

I feel like everybody gets a central vac or gets a a robot vacuum with like the idea that I'll never have to vacuum again. And then at some point, you just never use it again. I don't know what it is.

你试过那种带拖地功能的吗？

Did you you try the ones that mop also?

不，我们已经看过、讨论过无数次了，我就知道我们会买一个，结果肯定会让人失望，然后就像以前一样再也不会用了

No. We've see, we've talked about this so many times and I've just known we'll get one, it will disappoint and then it'll never be used again like

大家都在谈论那个新款。就是那个立方体形状、前面带吸尘口的。大家都说‘哇这个真好’。其实很多人根本没用过。我记得...好吧

There's that new one everyone's been talking about. The one that's a cube with like the vacuum in the front. Everyone's been like, oh, this is so good. A bunch of people don't really have one. I remember Okay.

它能拖地吗？我想...我觉得它应该也能拖地。大家都说它很棒。好吧

Was it Does Does it mop? I think. I think it also mops. Everyone says it's great. Okay.

我看到它的原因是它有个很酷的开箱设计——当你打开盒子时，它已经知道你的名字了。他们不知怎么预编了订单姓名，它会打招呼，然后自己从箱子里滚出来

The reason I saw it was it's got this cool onboarding thing where when you open the box, it like knows your name already. Somehow they like program your order name and it says like it says hi and it like rolls out of the box on its own.

哇哦，你只要打开那个小门然后

Oh, wow. You just open that little door and

它就

it just

自己开出来了。简直像

drives on out. That's like

一个非常精致的细节。

a really nice detail.

是啊，如果它吸尘好用，拖地也好用的话——我特别喜欢干净的地板。刚拖完地时赤脚踩上去的感觉太棒了，我简直爱死这种感觉了。我可能清洁剂喷太多了。

Yeah. Well, if it's good at vac if it's good at the mopping part, I like, I love clean floors. I love the feeling on my bare feet when you've just mopped. Like, I love it. I use like too much spray.

因为我总是喷过量，所以更喜欢自己动手拖，这样地面还能保持微微湿润的状态。

Like I like to do it myself because I overspray so it's still a little damp.

就是迷恋那种感觉。

Just love that feeling.

噢，可能我有点怪吧。

Oh, maybe I'm weird.

这太好笑了。

That's so funny.

今天有些事我想聊聊，你可以把它归类为...毕竟这是我们的播客，想聊什么都可以。我本来犹豫要不要说这个，因为不知道会不会显得像是在背后说人坏话...但管他呢，我们就要聊这个，因为我就是...

There's stuff I wanna talk about today and it's you're gonna file it under like, it's our podcast. We can do whatever we want. Like, I thought maybe I shouldn't talk about this because I don't know. I don't wanna come off as like throwing people under the bus or like Just shitting our other We're gonna do it. Gonna do it because it's I just have a

强烈的感觉。

strong feeling.

是啊。好吧。你提起这事是因为刚在推特上吵了一架。对。当然。

Yeah. Okay. You were gonna bring it up because you just got in a Twitter fight. Yeah. Of course.

当然，这事迟早会被提起。所以我才会想到。比如，我们该不该讨论这个？对，我们应该谈谈。

Of course, this would have come up. That's why I thought of it even. Like, should we talk about that? Yeah, we should.

你先开始。

You start.

你先说。不。不。你你才是挑起话题的人

You go first. No. No. You you were the one picking

提起这事。

it up.

好吧。我要这么说：通常在推特上，我会直接转发引用这类内容，这样大家就不知道我在说谁。我感觉好受些，因为我说出了让我恼火的事，但又不必指名道姓地让人难堪。我经常发这类推文。

Okay. Well, here's what I'm gonna say. Normally, like on Twitter, I I would just quote tweet something like this and no everyone can just like not know who I'm talking about. I feel a little better because I got to say the thing that's annoying to me, but I didn't have to really call anybody out and be like, everybody look at them. And like I do a lot of these tweets.

事实上，我觉得过去几周我的推特全是这样——我对某人感到非常不满，但我不想这样。

In fact, I I think like the last few weeks, this is all my Twitter has been where it's just like, I'm really like upset about somebody and I don't want

变成这样，但是

to be this way, but

我只是被某些事惹恼了，发推只是为了自我宣泄。

I'm just annoyed by something and I'm just tweeting for my own sake to like get something out.

就像，发泄一下。

Like, vent.

是的。在播客里我觉得有更多上下文。我能说更多话，你们也能看出我是活生生的人，而不只是个对别人恶语相向的混蛋。言归正传，我们得聊聊AMP，他们快把我逼疯了。我们私下确实讨论过这事。

Yeah. On the podcast, I feel like there's more context. Like, I can say more words, and you can see that I'm a human being and I'm not just an asshole saying mean things about other people. So with all that said, we gotta talk about AMP because they drive me crazy. And we I mean, we've definitely talked about it in private.

现在有些内容已经公开了，你和AMP的联合创始人（或Sourcegraph创始人）在推特上你来我往。我就直说了吧。我也不知道。可能我已经后悔了。德克斯，你后悔了吗？

And now some of it has made it into public where there's tweets flying back and forth between you and the AMP co founder or founder Sourcegraph. Here here's I'm just gonna say my piece. And I don't know. I maybe already regret this. Do you already regret this, Dex?

不，我不后悔。

No, I don't.

好吧。你没有。当然

Okay. You don't. Of course

你没有。

you don't.

你是德克斯。来自中西部，所以我非常遗憾。我相信他们都是可爱的人，推特上那75个人都是。因为我发誓，我从没见过一家公司有这么多人工作。是不是我只看到了其中一个

You're Dex. From the Midwest, so I very regret it. I'm sure they're all lovely people, all 75 of them on Twitter. Because I swear to God, I've never seen more people work at one company. Is it just that I saw one of

每一天，我都能发现一个新的人在那里工作。

the Every single day, I see a new I discover a new person that works there.

这太疯狂了。是的。所以第一个观察是，他们人数众多，而且对自己的方式非常直言不讳。这就是让我困扰的地方。如果你是任何领域的联合创始人，不仅仅是AI工具领域，但在AI工具领域，这是最让我印象深刻的一点。

It's insane. Yeah. So first observation, there are many of them and they're very vocal about their way. And this is what bothers me. If you're a cofounder in any, like, area, not just AI tools, but in the AI tool space, this is the one that stands out for me.

如果你正在构建某个东西，或者在一家早期初创公司工作，团队很小，你对正在构建的东西非常兴奋，就像明白你可能是错的。我认为这是我在研究这些AI东西时最大的体会。我们根本不知道发生了什么。没有人知道。五年后，一切都会与现在大不相同。

If you're building something or you work in an early stage startup and there's just a small team and you're very excited about what you're building, just like understand you could be wrong. And I think that's the biggest thing for me working on this AI stuff. We don't have any idea what's going on. Nobody has any idea. In five years, it's all gonna look very different than this.

其中部分原因是技术进步或其他什么。但很大程度上只是因为我们现在在构建时并不知道自己在做什么。我对Ant人的问题是，他们如此确信自己会找到最佳方案，就让他们有片刻时间去探索，他们会从山上下来，将AI辅助的神圣礼物赐予开发者——这就是整个氛围，渗透在整个公司里。我确信是有几个个性鲜明的人真的觉得自己能解决问题，然后等他们搞明白了就会告诉大家。所以他们发推的方式非常权威，好像他们知道一切。

And some of that will be because the technology advances or whatever. A lot of it will just be because we didn't know what we were doing when we were building right now. And the problem I have with the Ant people is that they're so certain that they will figure out the best way and just let them have their moment to figure it out and they'll come back to they'll come down from the mountain and bestow upon developers God's gift to AI assisted That is the whole vibe and it permeates the entire company. So I'm sure it's like a couple personalities that really have that feeling that they can figure it out, and then they'll let everyone know when they figure it out. So they tweet in such a way that's like very authoritative and like, we know what's going on.

然后当你说些什么的时候，如果你想讨论推特那些事，他们就会试图往回找补，但他们说的每句话都带着一种'我们最懂'的态度，简直让我抓狂。其实我还有很多想说的，但我不确定该不该说。

And then when you say something, if you wanna get into the Twitter stuff, it like, they try to kinda backtrack, but it's like everything they say is just in that light of we know best and it drives me crazy. And now there's a lot more I could say, but I don't know.

也许这就够了。不，这个总结很到位。我是说，我想向他们学习的是，他们是如何在整个公司建立起如此一致且紧密的企业文化的。嗯。

Maybe that's enough. No. That that is a good summary. It is a a it it's I mean, the thing I would like to learn from them is how did they get such a consistent and tight culture across their company. Mhmm.

因为每次我看到——真的每天都能发现新员工——他们所有人说的话、表达的方式都如出一辙，

Because all every time I see again, every every day, I discover a new person that works there, and they're all they're always saying the exact same thing and they're talking the exact same way,

他们就是那样...

which They got that like

这不是一种好的文化，但非常...非常紧密。

It's it's not a good culture, but it's a it's very it's very tight.

确实很一致。我见过几家这样的公司。比如丹·吉尔伯特，底特律的亿万富翁，拥有体育球队之类的资产，他旗下有Quicken Loans等上百家企业。

It's consistent. Yeah. So I've seen a couple companies like that. Dan Gilbert is like a Detroit owned he's a billionaire that owns like sports teams and stuff. He has the Quicken Loans and like a 100 other companies.

我记得参观Quicken Loans园区时，每个员工都能向你背诵他们所谓的'信条'。那种高度同步的文化令人震惊——所有人思维完全一致，完全理解丹的人生和商业哲学。是的，他们能对你倒背如流。这让我想起了那种感觉。

And I remember visiting the Quicken Loans campus and like getting a tour. Every single one of those employees like could tell you their little isms or whatever they call them. There's like this crazy crazy synced culture where they all think exactly the same or they just like all understand Dan's philosophy on life and in business Yeah. And they can recite to you. It reminds me of that.

但是，你知道吗，疯狂程度更甚。几周前我发了个帖子，这是我和Liz讨论过、意识到的一件事。我提到有种气场，你可以表现得非常放松自信，但我花了很长时间才明白这其实是新手标志。我之所以理解这点，是因为我终于想通——嘿，我以前就这样。记得早期和Liz合作时，我们和客户交谈就是这种状态。

But, you It's know, way worse crazy. So there's a thing I posted a few weeks ago and I've been this is like something that me and Liz talked about, something that we realized. So I posted something about there's this vibe that you can put out there where it's like really chill and confident that it's taken me a very long time to understand that's actually a sign of an amateur. And the reason that I understood this is because I I finally put together that, hey, I used to do this. I remember like when me and Liz were working on stuff together like very early on and we would talk to customers.

我们当时就是这种满不在乎的态度：'这领域很疯狂，但我们擅长这个，对我们来说轻而易举'。现在仔细回想，那不过是在自我催眠我们会成功。

We had this exact same like laid back, like, oh yeah, this space is crazy but like, know, we're good at this stuff and like we're we're coming in and it's like easy for us and Yeah. Whatever. And I and if I really think hard about that, I'm like, that was just us convincing ourselves that we were gonna succeed

嗯。

Mhmm.

尽管当时还没任何实际成果。所以现在看到这种状态，我立刻能识别出来：'哦，这是你惯用的把戏'。当然身处其中时你毫无察觉，只觉得是理所当然的自信。

Despite, like, having nothing to really show yet. So now when I see that, I, like, recognize that I'm like, oh, this is a thing you do. And of course, you're in it, you don't feel like that's what you're doing. Feel like you are confident and you are and you are, you know, whatever. Sure.

但现在我完全不是这样了。我只觉得这领域太疯狂，成功窗口转瞬即逝。我像疯了一样四处搜寻机会，知道每周甚至每天情况都会不同。

But now, I don't operate that way at all. I am just like, this space is crazy. There's such a small window to get something in that works and, like, I'm, like, just, like, spastically, like, looking for it and chasing it. I know that it's gonna look different every single week and every single day. Yeah.

所以我没法再装淡定。现在站在旁观者角度看同行公司：'你们凭什么这么从容？现在应该急得跳脚才对'。

So I can't I just I just can't do this comm thing. So but now I'm on the other side of it. I'm like looking at the other companies in the space. I'm like, why are you so chill? Like, you should be going insane right now.

看看你们的处境、市场机会的规模、涌入的资金量——根本容不下所谓的从容自信，这压根不现实。回到你刚才说的，这种做派确实让人恼火，因为这不是...我是说，单就这种企业文化本身就很令人反感。

Like, look looking at, like, where you are, the size of the opportunity, how much money is going into this space. Like, there is no room for, like, calm confidence. Like, it's just not a real thing. And going back to the same thing you were saying, it's yeah. I also find them annoying because it's not like they're just I want mean, on its own, it's annoying just to, like, that to be your culture.

但他们也在向同领域的其他公司说三道四。当时有个在那儿工作的人回复了克莱因公司的某人，他说，'我觉得你们那套行不通'，基本上就是在说'你们的东西不会成功的'。嗯。有趣的是，AMP和Cline都是VS Code的扩展插件。

But it's they're also like telling other companies in the space stuff. So there was interaction with someone that worked there replying to someone at Klein and he was like, yeah, I don't think you guys your guys' thing will work. Like like, I think you're the way your position is not gonna work, like, you're and and, like, he was basically telling them, like, it's your stuff is not gonna work. Mhmm. And what's funny is AMP is a Versus Code extension and Cline is a Versus Code extension.

我看到那番对话后就想，'哦，让我去看看下载量如何'。一看AMP，3.2万次下载；再看Cline，200万次下载。我当时的反应是：'搞什么鬼？你凭什么对这些人指手画脚？'

So I saw that interaction, I'm like, oh, like, let me go check to see what the downloads are. Looked at AMP, 32,000 downloads. Okay. Looked at Cline, 2,000,000 downloads. I'm like, what the how can you go and tell these people anything at all?

我明白他们资历更老什么的，但是——

Like, I get they've been around longer or whatever, but

是啊。

Yeah.

——比如，闭上你的臭嘴。

Like, shut the fuck up.

比如，你的头发应该

Like, your hair should be on

烧起来才对，想着'他们怎么把我们碾压成这样'，懂吗？对。然后还跑去对他们说教。你根本不是给他们建议，只是在自我安慰你们的东西能成罢了。

fire being like, how are they crushing us so bad, you know? Yeah. Like, to like then go and like try to tell them something. You're not really telling them anything. You're just trying to tell yourself that your thing is gonna work.

是啊。既然你这么一说，我完全明白了，这更像是他们在自我说服。就像一种虚假的自信，也许能帮他们撑过一天。我也不确定。

So Yeah. So now that you framed it that way, it all makes a lot more sense to me that it's more like convincing themselves. It's like a yeah. It's a false confidence that helps them get through the day maybe. I don't know.

我从来不是个自信的人，一直都很焦虑。在我经验不足、本该更淡定的阶段，这种问题反而没出现过——我这辈子就没淡定过。所以虽然无法感同身受，但我就假设事情就是这样吧。

I've never been a a confident person. I've always been very anxious. Like, this has not been a problem of mine when I was less experienced where I was like calm. Just never been calm in my life. So I can't relate to any of that, but I'm just gonna assume that's what that's what's going on.

对啊，就是...挺滑稽的。本来以为只是少数人这样，没想到有这么多人。

Yeah. It's yeah. It's it's just funny. And I think it was just it was just like a few people whatever, but it's just there's so many of them.

确实太多了。

There's so many of them.

没错！就像你说的，我前几天终于憋不住怼了他们——好像是昨天？记不清具体时间了。最搞笑的是他们CEO回复后，我用下载账号回怼，完全碾压了他。结果AMP员工还在转发CEO那条被群嘲的推文。

Yeah. So so, yeah, like you mentioned, I finally, like, couldn't hold it anymore and I, like, finally dunked on them a couple days ago or what was yesterday? I don't remember when when it was. And I thought it was so funny because their CEO replied and then I replied with their download account and we're just like completely ratioing him. And then I kept seeing employees from AMP retweeting their CEO's post.

简直笑死我了。招这么多人难道就为了凑转发量？但连这都搞砸了——他们转发的正是老板被全网嘲讽的帖子，等于是放大了打脸现场。这操作真是...绝了。

And I'm like, this is so funny. Maybe maybe this is why you hire so many people so you can get, like, more retweets. But then even that was done in a stupid way because they were retweeting a post where in the reply he was getting ratioed. So they was kind of amplifying the dunk. So like this this is just like Oh.

这种事总让我震惊，因为见过太多次了

This stuff blows my mind because I've seen this happen

很多

a lot

在那些人们年龄大致相仿的地方。比如，我们都属于同一代人，你知道的。是的，上下差不了几岁。在我脑海里，我觉得我们这代人都懂互联网，懂社交媒体，因为我们是伴随着它们长大的。但后来我总看到这样的情况，我就想，哦，原来这里还有个亚文化群体，他们就是搞不懂这些。

where there are people that are all roughly the same age. Like, we're all in the same generation, you know Yeah. You know, give or take. And in my head, I'm like, everyone from our generation, like, understands the Internet and, like, understands social media because we grew up with it. But then I always see stuff like this and I'm just like, oh, I guess there's actually a subculture here where people just don't get it.

是啊。你还记得Drizzle和Prisma的那些互动吗？比如Prisma的首席运营官总是那样——他也就比我大个五到十岁，不是五六十岁那种。但他就是理解不了Joseph开的玩笑。

Yeah. Do you remember all the Drizzle and Prisma interactions? Like, the the Prisma COO would always just like he and he's like it's not like he's like a guy in his fifties or sixties who's maybe like five or ten years older than me. But like he just could not understand when Joseph was making the joke.

对，对。不是所有人都这样。我想千禧一代里也不是每个人都...我是说，Z世代的人可能在笑话我们不懂互联网。但从千禧一代的角度来说，我们其实是懂的。

Yeah. Yeah. It's not everybody. I guess not everybody to in the millennial generation just I mean, I guess Gen Z people are laughing at us and saying we don't get the Internet. But I mean, like, we get the Internet from a millennial standpoint.

我们懂的部分够用了。但确实存在鲜明对比。我们和这个特定领域的几家公司打过交道，他们明显不懂...是的，完全不了解互联网或社交媒体。

We get enough of it. There's definitely a stark contrast. We've interacted with a few different companies in this very specific space that clearly don't get Yeah. And understand the internet or social media. Yeah.

也许是这个领域的问题？可能搞开发的人就是...我也不知道。

Maybe it's maybe it's this space. Maybe there's something about developers working in I don't know.

但你看——我还见过其他例子。比如PlanetScale的人。他们的CEO Sam在推特上就特别有趣。就算卷入这种小争论时，他的反应也很有幽默感。

But like so you take the I've seen some other things. Like you see PlanetScale people. Like the PlanetScale CEO, like like Sam is really funny on Twitter. And like Yeah. Even when he gets into like little engagements like this, he's like funny in this way.

然后我看到他的竞争对手们完全无法理解这一点。你知道，我就是不知道该怎么处理这种情况。我就想，你们这些人到底在干嘛？你们是刚通网吗？我实在搞不懂。

And then I see his competitors just like not be able to process that at all. There's like, you know, it's just I just don't know what to do with it. I'm like, what what were you guys doing? Like, did you just get the internet recently? I don't understand.

是啊。确实存在某种特质——虽然我脑子转不过来想不明白——但肯定有某种特质在网上就是行不通。这些人可能自以为很受欢迎，但实际上...我不知道，就是他们社交方面的某些东西...

Yeah. There's there's some trait and I I'm trying to like, I don't have the brain to like figure this out, but there is definitely some trait that just does not resonate online. And lots of these people, they probably feel like they're resonating, but like, it just it's just it's the I don't know. Something about their social

媒体形象

media presence

就是不对劲。就像恐怖谷效应，你看起来不像是真人。我不知道你是什么，但这听起来...算了。反正是我们的播客，想聊什么都可以。

is just like it's off. It's just like uncanny valley. Like, you're not actually a person. I don't know what you are, but that sounds I don't know. I just feel like this is it's our podcast, so we can talk about it where we want.

不过在播客上单方面谈论这些互动确实感觉怪怪的，毕竟对方没法为自己辩护。所以...想辩护的话，或许可以...

But it does feel weird to be on a podcast, like, saying our side of things and talking about interactions when they can't defend themselves. So I don't know. If you want to defend yourselves, maybe take

他们可以自己做播客啊，随便他们。

They me can do their own podcast, whatever.

他们确实有自己的播客。对，他们完全可以...

They do their own they do their own podcast. Yeah. They can they

可以说

can say

他们想怎么说我们都行。但问题是我们是不是混蛋？我开始怀疑了，我们和很多其他AI助手公司都有不少争执。是吗？我是说，感觉我们之间有点历史渊源。

what they want about us. And are we the asshole? I'm starting to wonder, like, we're getting in a lot of fights with a lot of other AI assistant companies. Are we? I mean, I feel like we've got a little history.

OpenCode的创立始于

The beginning of OpenCode started with

好吧，那次不算数。

well, that that one doesn't count.

不算数对吧？好吧，我们不是。

It doesn't count, does it? Okay. We're not.

太酷了。没关系。其实Liz之前就跟我指出过这点。她说，你和你的朋友们特别擅长互联网那套。她还解释说，大多数人其实不是这样的。

So cool. It's fine. I mean, Liz did point this out to me. She was like, oh, you and your friends are like very good at internet stuff. And she was explaining to me like, yeah, most people aren't that way.

因为在我脑海里，我们都曾经活跃在AAM和即时通讯软件上，所有事情似乎都源于那个起点。但我想对很多人来说，那只是他们接触互联网的起点和终点，他们后来就逐渐淡出了，而我们却一直沉浸其中。

Because in my head I'm like, oh, we all were like on AAM and instant messaging and like everything kind of flows from that origin. But I guess for a lot of people that's kind of where it started and stopped and they kind of grew out of it and we just never grew out of it.

如果他们从中成长了，那确实是好事。这种说法挺有意思的。就像他们休息了一段时间后又回到了互联网。大概就是这种感觉。对。

If they grew out of it, that's a good yeah. That's an interesting way to put it. Like, they took a break and came back to the Internet. That's kind of what it feels like. Yeah.

感觉这些人回来是因为他们有理由需要上网推广自己的东西。就像...每次我上Reddit时——我不是Reddit用户，只有想炫耀自己做了什么时才会去。

It feels like people who came back because they had a reason they need to be on the Internet to like promote their thing. It's like Yeah. Every time I've ever gone to Reddit. I'm not a Reddit user. I only go when I wanna like post something I did.

比如'看啊，我做了件很酷的事'。而在他们眼里，他们大概觉得'这混蛋是谁啊？'这就是我们对这些人的感受——

Like, look, I did a cool thing. So like to them, they're like, who is this asshole? That's how we feel about these people on

右派人士。

the Right.

他们就是不懂这里的氛围，就像'先生，这里是温迪快餐店'这种感觉。

They just don't get like the vibe here, like this is a Wendy's, sir. It's something like that.

对。还有个有趣的事，我最近一直在想这个，我觉得我要对此采取强硬立场：当有人当面表现得咄咄逼人、刻薄无礼时，这显然会激起我的反应。嗯。但当有人对我假客气时，我觉得应该用十倍力度回击。

Yeah. The other thing that's funny is so I was thinking about this so much and I'm like, I think I'm I'm gonna take a hard position on this. When someone is in my face aggressive, mean, rude to me, that obviously pulls out a reaction. Mhmm. When someone is fake polite to me, I think that deserves a 10 x reaction.

嗯。因为我觉得虚伪的礼貌更恶劣。在我看来，这种行为就像霓虹灯一样刺眼——表面礼貌但充满被动攻击性，每个词都精心挑选过，带着政治正确的味道。对。

Mhmm. Because I think that's just like way worse And to me, it's like it's like it like radiates in neon. It's obvious that this is like technically polite but passive aggressive and words chosen. Like, the words were chosen carefully chosen carefully and there's like a political correctness to it. Yeah.

对我来说，那种感觉简直扑面而来。我能看出来，然后我就觉得，去他的吧。对其他人来说，他们根本察觉不到。他们会觉得，哦，这就像是

For me, that just radiates. I can see that and I'm just like, fuck this. Yeah. For other people, they're they don't register it. They're like, oh, this is like

这就是我们在领英上的说话方式。

This is how we speak on LinkedIn.

没错，正是如此。

Yeah, exactly.

这就是我们的作风。我们会说，我想成为你的竞争对手，这很有趣。

This is what we do. We say, I'd like to be your competitor, this is fun.

有人回复我的评论说，哦，他试图表现得礼貌。我心想，你个白痴。这根本不是礼貌。这是硅谷那种愚蠢的说辞。写他妈的三段话长篇大论，真正想说的内容却藏在里面。

Someone replied to like my my reply being like, oh, he tried to be polite. I'm like, you dumbass. Like, this is not politeness. This is that like stupid Silicon Valley like saying Yeah. You know, writing a fucking like three paragraph thing and like the thing they actually want to say is hidden in there.

所以对我来说，我觉得这种状况必须结束。我们不能仅仅因为它在技术层面上算礼貌，而用不礼貌的方式回击会让你看起来像个混蛋——这才是问题所在。嗯。所以总会有人利用这点，而你却束手无策，因为你不想让别人觉得你粗鲁。但那种做法本身就很无礼，对吧？

Yeah. So yeah, to me I'm just like, got this has got to end, you know, we can't just because it's technically polite and you look like a dick replying to it in a non polite way, that's the problem. Mhmm. So then someone can always employ that and you're kind of paralyzed because you don't want other people to think you're being rude. But that thing is rude in the first place, right?

嗯哼。没错。要么一开始就直截了当地粗鲁，要么干脆什么也别说。这是另一个选择

Uh-huh. Yeah. It's better to say Yeah, be be rude in the first place or just say nothing. That's another

是啊。只是我觉得这是我第一次在AI领域工作，感觉这个领域尤其充斥着各种江湖骗术、低劣和诡异的营销废话。

option Yeah. There it's just I feel like it this is the first time I've worked in AI. I feel like this space is even extra like snake oil, sleazy, weird marketing BS

假知识分子，我觉得可能是这样。

Fake intellectual, I think is probably

假知识分子，没错。这个方面。我们太聪明了，因为我们知道怎么调用LLM的API，闭嘴吧。

Fake intellectual, yeah. Aspect. We're so smart because we, you know, how to call an LLM API like, shut up.

是啊。

Yeah.

停。

Stop.

另外一点是，我试图...因为我能感觉到这里有些似曾相识的暗示，因为上一波浪潮我参与的是那些前端技术框架之类的东西。嗯。我得说相似之处在于当时也有很多'我们知道最佳实践，你应该按我们的方式来做'。但区别在于，当他们说'我们知道最佳实践'时，确实存在一种具体可查、可理解、可批评和指摘的方式，这让它更容易忍受，因为你可以据理力争。而现阶段，感觉就像'我们知道最佳实践，但这是个秘密'。

It's a the yeah. The other thing is I'm like trying to because I can there's some hints here that feel familiar because the previous wave I was a part of was like all this like front end y tech framework stuff. Mhmm. And I will say it's similar in that there's a there was a lot of like, we know the best way, you should do it our way. But I think the difference was when they say we know the best way, there was a literal way you could look at and understand and criticize and point to, which made it a lot more tolerable because you can like argue In against this phase, it just feels like we know the best way and it's a secret.

但相信我，效果真的很好。只不过这是个秘密，我们不能告诉你具体是什么。

But trust me, it works really well. But it's a secret. We can't show you what it is,

但它确实运行得很好。

but it works really well.

所以这就像是，它非常相似，但实际上没有什么可看的。是的，没有办法

So there's like, it's it's it's very similar, but there's nothing to actually look at. Yeah. There's no way

去衡量，比如说

to measure either, like

是的，没办法衡量。这非常令人困惑。

Yeah. There's no way to measure. It's very confusing.

如果有人确实有最好的方法，假设现在确实有人拥有最好的方法。也无法证明或知道这一点。就像，没有标准，对吧？现在没有任何基准，全是凭感觉。就像，我喜欢Cloud Code因为Cloud Code漂亮且感觉良好。

If someone did have the best way, let's say someone actually does right now have the best way. There's no way of proving that or knowing that. Like, there's no, right? There's no like benchmark right now that's like, it's all vibes. It's like, I like Cloud Code because Cloud Code pretty and it feels good.

这就是人们仅仅因为这个原因使用Cloud Code。并不是说有人能衡量出区别。

That's what people just use Cloud Code for that reason. It's not like anyone can measure the difference.

是的。而且，你知道，确实有基准测试之类的，但是，这个领域太复杂了。你改变一个变量，它就会以如此疯狂的方式产生连锁反应，比如，谁知道呢，使用短函数名可能完全改变整个系统的行为。任何小变量都可能影响，嗯。你使用的语言，你编写的模式，代码库的成熟度。

Yeah. And and and, you know, there there are benchmarks and stuff but like, yeah, it's this is such a complex space. You change one variable and it cascades in such a crazy way like, you know, for all you know, all we know, like using short function names across your code base might totally change the behavior of the whole thing. Like any like little variable like that Mhmm. What language you're using, like the patterns that you write with, like the maturity of your code base.

最让人难以接受的是，有人说‘哦，这个工具用起来很棒，比那个工具好多了’。你几乎总能找到持完全相反观点的人。是啊，这简直像个愚蠢的竞技场。还有件有趣的事，最初我刚接触时，大家都在说‘我们团队不眠不休花了一周时间为新模型做优化’。

It's just it's so hard to take anyone saying, oh, had a good experience with this tool and it's better than that tool. You can find someone saying the exact opposite almost Yep. Almost always. So yeah, it's like a dumb space. And the other thing that's funny is initially, you know, when I was new to it, was like, okay, everyone's saying, oh, our team spent a week without sleeping to do all these optimizations for this new model.

我当时就想，哇，这也太疯狂了。让我也研究下——我也想让我们的产品优化。比如我去看看能做什么，结果发现：哦，我只需要改两个变量。

So I'm like, oh, wow, that's crazy. Let me go look into I I want I want our stuff to be optimized too. Like, let me go look into it. What can I do? Oh, I can change two variables.

哦，我还可以调整这个提示词，居然就奏效了。就像编辑工具有时不起作用，那就忽略空白字符呗。这些‘优化’花了我一周时间随便捣鼓就搞定了。难道我漏掉了什么吗？

Oh, I can fuck with this as some prompt and it like kind of works. Oh, like, you know, like I guess the edit tool sometimes doesn't work so ignore white space. Yeah. That took a week of me casually poking into it And that's it. Like, am I missing something?

我到底... 我到底错过了哪些优化？更讽刺的是，在我做这些之前，用户已经在夸‘这比Clockwork强太多了’。

What am I Yeah. What are these optimizations that I'm missing? And then even before I implemented that, people were already using our stuff being like, wow, this is so much better than Clockwork.

因为全靠感觉——只要有人用你的产品体验好，他们就会成为死忠粉，并坚信其他工具都是垃圾。

Because it's all vibes and if somebody has a good experience with your thing, then they're sold and they can sit they're convinced that every other tool is garbage.

是啊，但我...

Yeah. But which I'm

发现这种情况太多了

finding there's a lot

营造良好氛围并让产品给人愉悦感其实很简单。

It's easy to create good vibes and make a product feel good.

是的。我们正在努力做的就是打造一款有趣的好产品。好了，我说完了。不。

Yeah. That's what we're trying to do is just make a good product, which is fun. Okay. I'm done. No.

关于放大器的事我还有一点想说，这让我想起他们的处理方式。但我觉得我已经说得够多了。没关系。如果放大器那边的人想，我不知道，回应这个...不知道。我们需要吗？不需要。

There's one other thing I want to say on the amp thing that reminded me of just the the way they approach it. But I think I've said enough. It's fine. If you guys if the amp people wanna like, I don't know, respond to this don't know. Do we need Don't.

就像我不需要给他们发邀请函一样。

To like I don't need to give them an invitation.

好的。没关系。如果你回应这个，我可能会倾向于回应。而且我们已经讨论过我更擅长网络互动，所以大概...

Okay. That's fine. If you respond to this, I'm gonna be inclined to respond to this. And we've already been over how I'm better at internet, so it's probably

感觉怪怪的。在播客里点名别人确实感觉很奇怪。不知道，就是觉得有点不对劲。

It feels weird. It feels weird. It does feel weird to call people out on a podcast. I don't know. Something about it just kind of feels weird.

是啊。我可能已经这样做了上百次，我应该...

Yeah. I've probably done it like a 100 times and I should just get

离开后我们为此付出了代价。

away And we've suffered consequences for it.

是吗？

Have we?

或者说我为此付出了代价。有一次我收到一封措辞含糊的邮件。哦，

Or I've suffered consequences for it. Once I got a partially worded email. Oh,

不。你还查邮件吗？我已经不再查看邮件了。就在最近几周，也许是几个月前，我突然意识到自己已经不再查看邮件了。现在我甚至完全不看邮件，这让我吃过几次亏。

no. Do you check email? I don't check my email anymore. I just realized at some point in the last few weeks, I stopped checking email, maybe in the last few months. Like I just don't even look at it anymore and that's bit me a couple times.

我不得不...我反而增加了查看邮件的频率，因为

I had to I have like I'm like increasing my email checkage if anything because

哦，真的吗？

Oh, really?

只是因为现在有更多事情找上门来，所以总的来说...有时我还能在邮件里发现一些好东西。

Just because there's more stuff coming in at us, so like, just generally. So sometimes I find some good stuff in there.

我收不到邮件是因为我只是个普通员工。问题就在这里。根本没人有理由给我发邮件。谁来给我发封邮件吧。我希望收件箱里能有点值得一读的东西。

I don't get email because I'm an individual contributor. That's the problem. I just there's nothing that anybody would email me for. Someone send me an email, please. I'd like to get something, I don't know, in my inbox that was worth reading.

不知道为什么，这让我想起了弗兰克。弗兰克有趣的地方在于，你给他一个明确的任务，他就会直接去做。不问任何问题，直到完成。我们曾对他说：'嘿，弗兰克，你应该去和中国的所有AI公司合个影。'

I don't know why, but it reminded me of of Frank for some reason. So what's funny about Frank is you give him a clear task, he will just do it. No questions asked. He will just do it into completion. And we're like, hey, Frank, you should, like, get a picture with all the AI companies in China.

每一天都...如果换成是我，我大概会说'好啊，等我记得的话可能会去做'。但弗兰克是每一天都...

And every single day and, like, if I that that was me. I'm like, yeah. Like, maybe I'll do that, you know, if I remember. But every single day,

砰的一声，新照片又来了。又是新的合影，

like, boom. New a New picture. New new picture of

弗兰克·德克斯特的标志。现在我们已经收集到三张了。太棒了。天啊。没错。

Frank Dexter logo. And like, we're up to three now. That's awesome. Oh, man. Yeah.

你还好吗？说到中国AI模型，你还在用Quin3吗？那是你的日常主力模型吗？

Are you still okay. So getting into the Chinese AI models. Are you still using Quin three? Is that your daily driver?

我又换回Sonnet4了，不过正在考虑转用GLM系列。听说GLM这周要发布新版本。嗯。嗯。

I switched back to Sonnet four, but I'm I'm thinking about switching to the GLM stuff. Admit GLM? Putting out some new releases this week. Mhmm. Mhmm.

我看了我们的数据，GLM现在是我们最受欢迎的模型，就在上周。

And I was looking at our numbers and like, GLM's our most popular model now, like Yeah. Just last week.

至少在OpenRouter上是这样。

On OpenRouter anyway.

所以，是的，看来人们确实对它感兴趣。我认为目前唯一的问题是它只支持128k的上下文窗口。对我来说没问题，但大多数人可能会觉得有点局限。

So, yeah. So it seems like people are are into it. Yeah, I think the only issue with it is it's they only serve a one twenty eight k context window right now. For me, not a problem, but I think generally people will find that tight. Yeah.

是的，我在想等我们在美国部署后，访问应该会方便很多。因为现在服务器是在新加坡。

Yeah. And I think I'm waiting for like once we do our like US deployment of it, it'll be a lot more accessible I think. Because right now it's hosted in in Singapore.

我们还没在播客里聊过GBD5吧？

We we haven't talked about GBD five, have we, on the podcast?

我觉得没有。

I don't think so.

首先，那些提示词最后怎么样了？是野兽模式提示词，还是算了用Codex来回调试？现在到底是什么情况

Where first of all, where did the prompts land? It was beast mode prompt, it was never mind use the codex prompt back and forth. Where where what is the current

我试过很多不同的版本。基本上我每隔一天就会在开源代码中替换默认设置，看看大家的反应。最新的是VS Code Copilot里的野兽模式。我觉得它在几个方面达到了不错的平衡。但问题是它很容易进入一个状态，就是每一步都要你确认。

I tried I tried so many different ones. I tried them basically I would swap out default in open code every other day and see what people thought. The latest one is the newest beast mode that's in Versus Code Copilot. I think it landed in an okay balance of a few things. So the the struggle with it was it's very easy to put it into a state where it just asks you for confirmation for every single step.

但你可以做到，只是会一直问'我该做这个吗？该做这个吗？'你只能不停地说'是，是，是'。嗯。或者你可以让它进入一个状态，连续运行两小时处理所有这些东西。

But you can do it, it just it's just like, should I do this? Should I do this? You just go, yes, yes, yes. Mhmm. Or you can put it into a state where it is just gonna go for two hours and just do, like, all this stuff.

嗯。最新版本似乎较好地把握了这个平衡。虽然还不完美，我还是觉得它要求确认的次数太多，但总比它完全失控要好。不过现在其他人...不知道你是否又看到了，这又回到了没人能真正...我向来是轶事证据的忠实粉丝，在这方面依然如此，但考虑到这些东西的非确定性，轶事证据尤其不可靠。

Mhmm. So the newest one is, like, it seems to thread that balance decently. It's still not perfect. I still think it asks too much for confirmation, but that's better than it, like, running Doing too wild. But, yeah, now there's other people like don't know if you saw again, this goes back to no one really being able to I am generally a huge fan of anecdotal evidence, and I I still am for this stuff, but anecdotal evidence is especially weak given how nondeterministic this stuff is.

但现在所有人都在说，Cursor用户都说JPD5很糟糕。Theo说他最初用Cursor时体验惊艳，现在再试同样的功能却很糟糕。所以，说实话我不确定自己改变了哪些变量。

But now everyone's saying that, for Cursor users are saying that JPD five was really bad. Theo was saying that what he tried originally with Cursor was amazing and now he's trying the same stuff again and it's terrible. So, yeah, it's kind of unclear what variables I'm changing.

OpenAI有前科。我听说他们每次发布都这样——嗯...第一周表现极佳，之后就很糟糕。他们是不是真的在削减成本，搞什么量化处理？

There's a history with OpenAI. I've heard this rhymes with every other release they've had where it's like Mhmm. Really good for the first week, really bad after that. Is it just are they literally just cutting costs and like quantization or whatever? Quantitizing it?

'量化'？我不...知道这个词

Quantitizing? I don't So, know word

好的。我前几天和Claude长谈过，试图真正理解这些技术的实用性。因为我在研究GLM时发现他们用32位精度训练。所以如果我们部署32位精度的模型，需要什么硬件配置？它确实告诉了我硬件要求。

okay. So I I had a long conversation with Claude the other day trying to really understand the practicality of some of these things. Because I was looking at GLM and I was like, okay, they train it at 32 bit precision. So if we deploy a 32 bit precision, like, what does this take? And it, like, told me the hardware requirements.

我当时就觉得，这也太夸张了。然后我就想，批处理模式下每秒能支持多少token？结果大概是每秒1到200个。我不确定这个数字对不对，但我觉得对于这么庞大的硬件配置来说，这速度也太慢了，成本高得离谱。

I'm like, that's pretty huge. And I'm like, how many tokens can this support per second with batch? And it was like one to 200 per second. I I don't know if this is this is correct. But I'm like, that is so slow for that much hardware, like, that is so expensive.

是啊，我还真没听说过有人用32位运行的。

Yeah. I've I've not heard of people running 32 bit.

对，所以我就想，OpenAI真的会这么做吗？因为他们不仅面临同样的问题，还有数亿用户。

Yeah. So I'm like, does does OpenAI even do that? Because they have not only do they have that same problem, they have hundreds of millions of users.

确实。

Right.

他们的用户量比其他所有公司都多得多。

Like, so many more users than everyone else.

我们了解这类信息吗？比如知道他们用什么精度训练？是16位吗？

Do we know things like that? Like, do we know, like, what they're training at? Like, what like, is it 16?

基本上大家都在用混合精度训练，16位和32位混合。具体会根据操作动态调整。因为当前这代NVIDIA硬件就是针对这个优化的。下一代好像能支持FP4训练，也可能是FP8，我记不清了。

Pretty much everyone trains at at at so it's a mix of sixteen and thirty two. It's like it'll blend depending on operation. The reason is because the current generation of NVIDIA hardware is optimized for that. There's an there's like, the next generation basically can let you train at FP four, I think, or FP I think FP four, maybe FP eight. I forgot what it was.

所有推理服务提供商都希望下一代模型原生就是这种较小尺寸，这样性能更好、运行更轻量。但我在想，OpenAI的模型肯定巨大无比，比这些开源模型还要大得多，而且用户量也大得多。Anthropic也是同样情况。所以我就纳闷，这些数字怎么对不上。

So all the all the inference providers are hoping that next generation models are natively at this smaller size so that they can perform better and and run smaller. But, yeah, I'm thinking like, man, OpenAI's models must be giant, even bigger than these open source ones. And they have way more users. Same thing with Anthropic. So I'm like, how the numbers aren't adding up.

他们不可能不这么做，必须得这么做。除非有其他我不知道的优化方法，也许有，但他们绝不可能满足得了这种需求。

Like, there's no way they could like they have to do this. They have to. Yeah. Unless there's like other ways to optimize it I'm not aware of, maybe there are, but there's no way that they can possibly serve this demand.

是啊，有趣的是这又回到了那个老问题——这些事根本无法量化。作为商业模式，他们完全可以通过稍微降低智能程度来赚更多钱，而用户真的能证明他们这么做了吗？你懂我意思吧？

Yeah. And it's it's interesting that goes back to the whole like there's no way to measure this stuff. As a business model, like they can just make more money by making it a little dumber and can people really prove that they did? You know what I mean?

所以OpenCode的问题也是这样。总有人跟我说做这个改进能在基准测试上提升15%，但除非用户能感知到，否则我根本不在乎。用户感受不到的改变，对我们增长没帮助，实际上毫无意义。

So yeah, that's the other thing when I'm thinking about stuff for OpenCode. Because people are always telling me, oh, do this and like it gets like 15% better on the benchmarks. I actually don't care unless a user can tell. If the user can't feel it, it actually doesn't help us grow. It doesn't really help us do better.

嗯，你可以说做100个这种小优化累积起来效果显著。行吧。但实际上根本不存在100个这种小优化。

Mhmm. And you can argue that doing a 100 of these little optimizations add up. Sure. Fine. But there really aren't a 100 of these little optimizations.

所以我对所谓20%的质量提升并不感冒——天知道那是什么意思。同理，他们可能很乐意接受20%的质量下降，反正用户察觉不到。网上还有各种阴谋论，比如『我有证据他们在做量化，因为这周模型明显变差了』。

So I'm, like, not that compelled by a 20% increase in quality, whatever that means. So similarly, they're probably very compelled by a 20% decrease in quality because people can't really tell. Mhmm. And there's so many conspiracy theories about like, oh, I think they're quantizing. Like, I have proof that they're quantizing because like, you know, this week the model's been worse.

但Anthropic又会否认，这就很迷惑，或许他们采用了某种技术性话术不算说谎。现在终于有个新基准测试专门衡量这个，他们正在运行...

And but then Anthropic will deny it, which is kind of confusing, but maybe they have to go with different definitions that are technically not lying. There is a new benchmark finally that is focused on measuring this. So they are running

有意思。

Interesting.

他们正在对不同供应商运行相同的模型。嗯。不知道你看到没有，Bedrock简直烂透了。

They're running the same model against different providers. Mhmm. I don't if you saw this, but Bedrock is complete ass.

我看到了。真遗憾。是啊。大概低了4%左右吧？我是说，那些供应商基准测试的差距非常小。

I did see that. That's unfortunate. Yeah. Well, it was like 4% lower or something, right? I mean, the the spread on those provider benchmarks was very tight.

不过那是个很宽泛的基准。对我们这种智能体工具调用场景来说...哦有道理。整个事情都...对。确实。

Well, but that that's like that's like a very broad benchmark. So for our use like the agentic tool calling Oh, it makes sense. Like all the thing. Yeah. Sure.

而且我听说...我不想点名具体公司，因为我觉得我不该知道这事。有家大公司和Bedrock签了合同现在想解约，因为他们体验很差。

And I've heard I don't want to say specific company names because I don't think I'm supposed to know this. There is a large company that is stuck with a contract with Bedrock that they're trying to get out of because they're not having a good experience with it.

因为亚马逊有自己的芯片对吧？专用芯片？对。所以他们没法快速解决问题。

Cause Amazon has their own chips, right? Like chips? Yeah. Yeah. So it's not a quick fix for them.

管它叫什么来着。

Or whatever it's called.

那大概就是训练过程吧。嗯，我也不确定。

That probably that that's the training. Yeah, I'm not sure.

是啊。所以我觉得谷歌直接用的是英伟达的硬件。所以通过谷歌使用Sonnet，体验可能会更好。有人在测量这个，所以他们应该能看到，好吧，同一服务商内部是否存在性能波动？他们和

Yeah. And so I think Google just has straight up NVIDIA stuff. So yeah, if you use Sonnet through Google, it's probably a better experience. So someone's measuring this. So I think they'll be able to see, Okay, are there like, within the same provider, are there like dips and increases They and

应该就做那个所有开发者都想知道的关键测试——当美国开发者醒来开始工作时，系统真的会变差吗？因为大家似乎都这么认为。这很有意思。

should just do the one test that every developer wants to know, which is like when The US developers wake up and start working, does it really get worse? Because everybody seems to think so. It's really interesting.

哦，有意思。你觉得那些...对，我猜时区重叠的人会感觉系统变差？

Oh, interesting. You think the people that have oh, yeah. I guess the people that have like overlapping time zones, like feel it get worse?

嗯，这是Anthropic公司的情况，人们抱怨Sonic在美国工作时间会变得特别笨。然后清晨时段

Well, This is the anthropic that people complain that Sonic gets way dumber during US working hours. And then like early morning

我不清楚，因为我的工作时间就是美国时段。

I wouldn't know because I work US opening hours.

是啊。我工作的时间是清晨之前

Yeah. Work I work early morning before

哦，是吗

Oh, do you

想想大多数人。我不知道。我其实没怎么感觉到。我是说，我确实有过糟糕的会话，但我从不会觉得是因为下午的缘故。有时候你就是会遇到那种怎么都搞不对的会话。

think think most people. I don't know. I've not really felt it. I mean, I definitely have bad sessions, but I don't ever think like, oh, it's because it's the afternoon. Like, it it does feel like sometimes you just get a session that was like, it was never gonna get this right.

就像，我可能尝试了十次。但通常只会觉得是自己把问题框架搞错了。我得先解决部分问题，然后分解出它能处理的部分。我一般会直接认定这是个棘手问题，而不是他们在量化。我觉得量化有点像替罪羊，你会开始把所有问题都归咎于它。

Like, I could have tried 10 times. But usually, just think like, that's I'm just framing the problem wrong. Like, I gotta take on part of the problem and then, like, break out the part it can do. Like, I generally just assume it was an intractable problem, not they're quantizing. I think, like, it is kind of a boogeyman thing where you can just you can start blaming everything on quantizing.

但这确实是个有趣的局面，对我来说有两点：第一，他们有充分的动机却几乎不用承担后果；第二，而且这无法

But it it is an interesting situation where here's the two things for me. One, they have every motivation and very little consequence, I guess, if And it can't

被证实。考虑到数量问题，我认为他们实际上必须

be proved to. I think they literally have to given the numbers of

满足需求。是啊是啊。而且他们已经在崩溃边缘了。特别是Anthropic，已经出现太多宕机和过载错误，显然他们有理由需要这么做。

support the demand. Yeah. Yeah. And they're already melting. Like, they're already like, especially Anthropic has so many outages, so many overloaded errors that like clearly they have a reason they need to to do this.

第二点是推特上那个Shrek头像的家伙x j d r。嗯，他相信他们确实这么做了。如果他这么认为的话，

And then two, the Shrek guy on Twitter, x j d r. Yeah. He believes that they do. And if he does,

这不是什么阴谋论，而是事实。

it's not like a conspiracy theory. It's fact.

他看起来非常聪明。

He seems very smart.

他确实看起来很聪明，是的。这很奇怪，因为想象一下你付费使用，这有点让人困惑，我理解为什么他们那边会这样设计，因为他们有消费者市场，比如Claude和ChatGPT之类的产品。是的，然后他们的业务范围从消费者市场一直延伸到企业级B2B领域。

He does seem smart, yeah. Yeah. It's it's so weird because like, imagine you pay because this is this is a it's kind of confusing and I get why it's carved for them from them on their end because they have a consumer market which is their Claude and ChatGPT whatever. Yeah. And then they range from that all the way to like enterprise b to b.

是的。如果我是一家企业公司，为底层基础服务付费。嗯。这就像付钱买一个CPU，结果他们随机决定说‘实际上我们只给你0.1个CPU’之类的。

Yeah. And if I am an enterprise company paying you for a low level primitive Mhmm. It's like paying for one CPU and they're randomly like, oh, we're actually just gonna give you point one CPU for like this

小时。嗯。

hour. Mhmm.

所以这真的很奇怪，我实在想不通。

So that's really weird and I can't really think.

确实如此。

It is.

我知道亚马逊有所谓的'吵闹邻居'问题，云计算领域普遍存在这种现象，但这不算什么大问题，影响其实不大。而在这里，他们某种程度上是在坑你。

Like, I know Amazon has like the noisy neighbor or like cloud generally can have like the noisy neighbor thing, but it's not that big of a deal. Doesn't really impact things that much. Whereas here, I'm like, they're kind of ripping you off in a way.

确实。我是说，这种事情按理说应该受到监管才对。但我看到SigKitten和XJDR都提到过，从经济学角度看，你基本上是在为某种概念买单，而他们可以在你不知情的情况下彻底改变规则。

It is. I mean, there there has to be some kind of like this is kind of thing that I feel like gets regulated, right, or something. But I've seen, like, SigKitten and I've seen XJDR say things about this. Like, the economics of it are like, you're basically, like, paying for this idea of something and they can just totally change that without your knowing.

行不通。对。

Not work. Yeah.

是啊，所以你实际上没得到等价的服务。不知道，这种时候应该有消费者保护法介入才对。

Yeah. So you're not really getting what you paid for. I don't know. Seems like there's consumer protection stuff that would come into play there. I

不知道。我是说这对那些构建上层应用的人特别难，比如Bolt、Lovable、V0这些团队。一方面他们无法控制底层，但处境却最糟糕——他们向用户收费的工具本应实现某些功能。

don't know. Yeah. Mean, this gets really hard with the people building the next layer of stuff, like let's say, like Bolt, Lovable, V0, etcetera. So one, they're not in control of this, but they're kind of in the worst position because they're charging users for a tool that's supposed to do something. Yeah.

如果工具失效而你已支付代币，这算怎么回事？就像他们收钱只给你一个可能成功的机会。但另一方面，确实也没有其他可行的商业模式。总不能只在成功时才收费吧？那样经济上完全不可行。

And if the tool doesn't work and you pay for the tokens and, like, what the heck? It's like they charge you for a chance of it working. But at the same time, there's no other model that can work. Like, what's the alternative? Like, they can't only charge you when it works because then it's totally unfeasible economically.

现在这样经济上已经不可行了。咱们实话实说——

It's already unfeasible economically. Let's be

诚实。是的。正确。是的。一路到底。

honest. Yes. Correct. Yes. All the way down to the bottom.

我感觉这些内容不断涌现，越来越多的文章发布出来。比如昨天那个光标问题，就像个Word文档，为什么有人会在Twitter上分享Word文档？他们是想追踪有多少人阅读吗？我不清楚。但你看到那个光标问题了吗？

I feel like that's coming out like, there's more articles coming out. Like, there the cursors the cursor thing yesterday, it was like a Word document, which why why did somebody share a Word document on Twitter? Do they just wanna, like, track how many people read it? I don't know. But did you see the, like, cursors problem?

是的。但那来自领投Klein这轮融资的风投机构。所以我

Yeah. But that came from the VC who led the round into Klein. So I

我之前不知道这事。你读过了吗？我是说，看起来

didn't know that. Did you read it? I mean, like, it seemed

逻辑上这是事实。确实如此。这种要么收取固定费用然后任由用户随意使用的模式，显然行不通。嗯。这种提供工具本应实现某种功能，却要为失败场景付费的情况，也挺奇怪的。

logically It's it's true. It's it's true. Like, this whole model where you just charge either a fixed price and then, like, people can use whatever usage, that's obviously not going to work. Mhmm. This whole thing where, like, you're delivering a tool that's supposed to do something and you're paying for a situation where it fails, that's, like, also kind of weird.

所以这些都是需要解决的问题。我认为大家对此的看法有个误区——如果他们的商业模式错了，而有人发现了正确模式，他们自然会转向新模式。是的。事实就是如此。人们总在强调'我们与众不同，这是我们的差异化优势'。

So these are problems to figure out. I think what is wrong about everyone's take on this is if their model is wrong and there is a correct model that is discovered, they will just move to it. Yeah. It's true. It's this it's this thing where everyone harps on, oh, we're different because this is our differentiator.

这就是我们的差异化优势。

This is our differentiator.

是啊。

Yeah.

问题在于，一旦市场意识到你是对的，那些比你规模更大的公司就会直接采纳你的做法。这样一来，差异化优势就消失了。

The problem is the moment the market realizes you're right, the guys bigger than you just adopt that. And now there's no differentiator.

比如说，全球约半数人知道Cursor是什么，但只有32,000人下载了AMP。所以祝你的定价策略好运吧。就像

What are Like, you half the world half the world knows what cursor is and like 32,000 people downloaded AMP. So good luck with your pricing model. Like

确实。这根本...这根本不可持续。说实话，你讨论得越多，写这类内容越多，他们可能就越快意识到你是对的。明白吗？

Yeah. It's just it's just not it's just not it's it's not sustainable. To be honest, the more you talk about it, the more you write these things, the faster maybe they consider that maybe you're right. You know?

没错。他们会转向。但我想他们已经尝试过调整了，对吧？就像Kerser案例中，他们试图调整定价却遭到反弹，最后不得不撤回。

Yeah. And they shift. But I guess they have tried shifting. Right? Like, in the Kerser case, they've tried shifting their pricing and gotten some backlash, and then they had to pull it back.