① 他说过的话

32 条

最大的原因是模型改进得太快了，我们觉得真的没有其他形态能跟得上它。
The biggest reason was the model is improving so quickly, we felt that there wasn’t really another form factor that could keep up with it.
—— Boris Cherny · [11:06]

指向原始笔记的链接

而且如果你看那条指数曲线，那个时间点由 Claude 编写的代码百分比，如果你只是沿着线画过去，很明显我们在年底就会超过 100%，即使这完全不符合直觉。
If you look at the exponential, the percent of code that was written by Claude at that point, if you just trace the line, it’s pretty obvious we’re going to cross 100% by the end of the year, even if it just does not match intuition at all.
—— Boris Cherny · [14:40]

指向原始笔记的链接

所以，我认为，在这个节点，可以说编码几乎已经解决了。
I think, at this point, it’s safe to say that coding is virtually solved.
—— Boris Cherny · [18:19]

指向原始笔记的链接

至少，对于我做的那些编程类型来说，这只是一个已解决的问题，因为 Claude 能做到。
At least, for the kinds of programming that I do, it’s just a solved problem, because Claude can do it.
—— Boris Cherny · [18:23]

指向原始笔记的链接

但是每个工程师的生产力在拉取请求方面增加了 200%。
But productivity per engineer has increased 200% in terms of pull requests.
—— Boris Cherny · [21:25]

指向原始笔记的链接

当你在所有事情上稍微投入不足时，会发生一件有趣的事情，因为那样人们被迫去 Claude 化。
There’s an interesting thing that happens also when you under-fund everything a little bit, because then people are forced to Claude-ify.
—— Boris Cherny · [24:21]

指向原始笔记的链接

我的看法是，对于今天正在使用 Claude Code，正在使用智能体写代码的人来说，你仍然必须理解底下那层，但是，是的，在一两年后，这就不重要了。
My take is I think for people that are using Claude Code, that are using agents to code today, you still have to understand the layer under, but, yeah, in a year or two, it’s not going to matter.
—— Boris Cherny · [32:27]

指向原始笔记的链接

然后，在某个时候，古腾堡和印刷机出现了，有一个疯狂的统计数据，在印刷机建成后的 50 年里，创造的印刷材料比之前的 1000 年还要多。
And, at some point, Gutenberg and the printing press came along, and there was this crazy stat that in the 50 years after the printing press was built, there was more printed material created than in the thousand years before.
—— Boris Cherny · [33:22]

指向原始笔记的链接

我们说，“产品就是模型。我们想要把它暴露出来。我们想要在它周围搭建最小化的脚手架。给它最小的工具集。“
We said, “The product is the model. We want to expose it. We want to put the minimal scaffolding around it. Give it the minimal set of tools.”
—— Boris Cherny · [51:29]

指向原始笔记的链接

但实际上，如果你只给模型工具，给它一个目标，然后让它自己弄明白，几乎总是能得到更好的结果。
But actually almost always you get better results if you just give the model tools, you give it a goal, and you let it figure it out.
—— Boris Cherny · [64:06]

指向原始笔记的链接

我们押注于为六个月后的模型构建。而不是为今天的模型。
We bet on building for the model six months from now. Not for the model of today.
—— Boris Cherny · [65:56]

指向原始笔记的链接

通常，如果你使用最强大的模型，实际上反而更便宜，且消耗更少 token，因为它可以只用更少的纠错、更少的指导等就把同样的事情做得更快。
Often, it’s actually cheaper and less token-intensive if you use the most capable model, because it can just do the same thing much faster with less correction, less hand holding, and so on.
—— Boris Cherny · [69:39]

指向原始笔记的链接

它可以一次运行几天、几周、几个月,就是不会停下来。
It can go for days, weeks, months at a time. It just won’t stop.
—— Boris Cherny · [01:39]

指向原始笔记的链接

这个模型似乎不再能被提示词注入了。
the model does not seem to be prompt injectable anymore.
—— Boris Cherny · [02:03]

指向原始笔记的链接

有了这三层,我们已经完全无法演示出提示词注入了。
And with these three layers, we just cannot demonstrate prompt injection anymore.
—— Boris Cherny · [03:16]

指向原始笔记的链接

所以是的,我们删掉了 80% 的系统提示词。
So yeah, we deleted 80% of the system prompt.
—— Boris Cherny · [04:30]

指向原始笔记的链接

有趣的是,没有这些提示词时,模型反而会稍微聪明一点。
And what’s interesting is that the model is actually a little bit more intelligent without these prompts.
—— Boris Cherny · [05:05]

指向原始笔记的链接

对于那些不是在构建智能体产品、但在使用 Claude Code 的人:每六个月,删掉你的 Claude MD。
And for people that aren’t building agentic products, but you’re using Claude Code, every six months, delete your Claude MD.
—— Boris Cherny · [06:56]

指向原始笔记的链接

只有当你看到它反复在同样的事情上跌倒时,那才是你把它加回去的时候。
And only when you see it repeatedly stumble on the same thing, that’s when you add it back.
—— Boris Cherny · [08:14]

指向原始笔记的链接

这个概念是:模型用今天的模型——不是未来的模型,而是今天的模型——就能做各种各样的事,只是我们尚未意识到。
And the idea is the model is able to do all sorts of things with today’s models, not a future model, but today’s model that we have not yet realized.
—— Boris Cherny · [11:04]

指向原始笔记的链接

我认为如今用现代模型,有非常多的产品悬置是我没有看到初创公司去捕捉的。
I think that nowadays with modern models, there is so much product overhang that I’m not seeing startups capture.
—— Boris Cherny · [13:34]

指向原始笔记的链接

你应该描述任务、描述护栏、描述退出标准,然后就让模型自己开干,过一会儿再回来。
You want to describe the task, you want to describe the guardrails, you want to describe the exit criteria, and then just go with the model cook and come back in a little bit.
—— Boris Cherny · [15:19]

指向原始笔记的链接

它运行了 11 天,重写了整个代码库。
And it ran for 11 days. And it rewrote the entire code base.
—— Boris Cherny · [17:40]

指向原始笔记的链接

而这个验证,我认为,可能是人们普遍没有做对的最重要的一件事。
And the verification, I think, is probably the single most important thing that people do not get right, largely.
—— Boris Cherny · [20:29]

指向原始笔记的链接

但实际上,你只需要把任务交给模型,给它一种验证自己工作产出的方式,这样它就不会卡住,它就会一直做下去。
But really, all you need is give the model the task, give it a way to verify the output of its work so it doesn’t get stuck, and it’ll just go.
—— Boris Cherny · [22:40]

指向原始笔记的链接

忘掉你在课堂上学到的关于计算机科学理论的一切。
Forget everything that you’ve learned about computer science theory in class.
—— Boris Cherny · [31:49]

指向原始笔记的链接

字面意义上，就是在提示词注入发生时，观察模型大脑中亮起的神经元。
where it’s literally, we’re looking at neurons in the model’s brain that light up when prompt injection happens.
—— Boris Cherny · [03:00]

指向原始笔记的链接

只有当你看到它反复在同一件事上绊倒时，你才应该把它加回去。
And only when you see it repeatedly stumble on the same thing, that’s when you add it back.
—— Boris Cherny · [08:36]

指向原始笔记的链接

通常，我们只是让 eval 饱和，然后我们必须把它扔掉，我们必须想出一个新的 eval。
Very often, we just saturate the eval, and then we have to throw it away, and we have to come up with a new eval.
—— Boris Cherny · [10:36]

指向原始笔记的链接

它运行了11天。它重写了整个代码库。
And it ran for 11 days. And it rewrote the entire code base.
—— Boris Cherny · [18:01]

指向原始笔记的链接

验证，我认为，可能是人们做不对的单一最重要的事情，大体上。
And the verification, I think, is probably the single most important thing that people do not get right, largely.
—— Boris Cherny · [20:51]

指向原始笔记的链接

忘掉你在课堂上学到的关于计算机科学理论的所有东西。
forget everything that you learned about computer science theory in class.
—— Boris Cherny · [32:11]

指向原始笔记的链接

② 出现在这些集

3 集

• 《Claude Code 负责人：写代码已被解决，下一步是什么》 — 作为主持
• 《Claude Code 造物主 Boris:删掉 80% 系统提示词，让模型跑两周不停》 — 作为主持
• 《别再微管理 Claude:Claude Code 造物主的智能体实战心法》 — 作为主持

③ 他谈到的

点进去有真内容 —— 本页主要出口

智能体 · 沙箱 · Claude Code · Opus 5 · Anthropic · 系统提示词 · 评估 · 消融实验 · 产品悬置 · 动态工作流

④ 也在聊「AI 编程」的人