① 他说过的话

14 条

早期,我们的延迟从两秒漂移到了大约两秒半,而那半秒钟几乎让产品里的所有指标都崩了。
Early on, our latency drifted from two seconds to about two and a half seconds, and that half second tanked basically every metric in the product.
—— Paula Dozsa · [03:39]

指向原始笔记的链接

所以我们构建了智能话轮转换,它会读取你的语音模式来判断一个打断是否是真实的,我们把最糟糕的过早中断削减了一半以上。
So we built smart turn-taking that reads your speech pattern to decide whether an interruption is real, and we cut the worst early aborts by more than half.
—— Paula Dozsa · [04:38]

指向原始笔记的链接

延迟不只是你最后检查的一个数字,它实际上就是产品,我们把管线的每个阶段分开来测量,因为「感觉慢」是没有用的。
Latency isn’t just a number you check at the end, it’s actually the product, and we measure every stage of the pipeline separately because it feels slow is useless.
—— Paula Dozsa · [04:54]

指向原始笔记的链接

我们的主要原则是,我们基于利害关系而非成本来路由。
And our main principle is that we route based on stakes, not on cost.
—— Paula Dozsa · [06:08]

指向原始笔记的链接

举例来说,我们有危机类或治疗师风格的语气,我们在这些方面从不省钱。
For example, we have crisis or therapist-style tones, and we never cheap out on those.
—— Paula Dozsa · [06:24]

指向原始笔记的链接

我们做了很多 A-B 实验,在那里我们发现的令人惊讶的结果是,把三分之一的话轮路由到小模型,对留存率几乎没有可测量的影响。
And we do a bunch of A-B experiments, and the surprising result we found there is that routing a third of our turns to the small model has almost no measurable effect on retention.
—— Paula Dozsa · [06:57]

指向原始笔记的链接

具体来说,你应该重建上下文,而不是对抗漂移。
Specifically, you should rebuild context and not fight drift.
—— Paula Dozsa · [08:15]

指向原始笔记的链接

但在易变的语音对话里,这是个陷阱,因为一旦用户转移话题,你复用的上下文就立刻错了。
But in a volatile voice conversation, it’s a trap because the second the user pivots, your reuse context is actively wrong.
—— Paula Dozsa · [08:25]

指向原始笔记的链接

主要是因为没有现实世界的参照物可以锚定,这意味着用户可以把自我投射上去,它会变成用户所需要的样子。
Mostly because there’s no real-world reference to anchor on, which means that users can project onto it, and it becomes what they need.
—— Paula Dozsa · [09:06]

指向原始笔记的链接

第二,我们也知道,如果人格漂移了,人格就一文不值。
Second, we also know that personality is worthless if it drifts.
—— Paula Dozsa · [09:39]

指向原始笔记的链接

我相信现在对在座的各位大多数人来说都是如此,基本上从去年年底开始,在我们 iOS 应用中,Claude 共同编写的代码已经超过了团队里的任何一位工程师。
So I’m sure this is the case for most of you in the room now, but basically as of late last year, Claude has co-authored more code in our iOS app than any individual engineer in the team.
—— Paula Dozsa · [10:05]

指向原始笔记的链接

构建这套系统的最大教训是,智能体的上下文主要来自代码库本身,而不是那么来自 Claude MD 文件。
And the biggest lesson in building that system is that an agent’s context comes mostly from the code base itself, not so much from the Cloud MD file.
—— Paula Dozsa · [10:33]

指向原始笔记的链接

所以当界面是你的声音、而另一端的东西记得你并有自己的人格时,它就不再只是软件,而是变成了一段真实的关系,这也是为什么负责任地构建它、把它构建好,是值得为之痴迷的事。
So when the interface is your voice and the thing on the other side remembers you and has a personality, it stops being just software and starts being an actual relationship, which is why building it responsibly and building it well is worth obsessing over.
—— Paula Dozsa · [12:26]

指向原始笔记的链接

你基本上必须把问题分解、带着检查点去委派、给出快速反馈、认真审查工作成果,并且知道什么时候该亲自介入。
You basically have to decompose the problem, delegate it with checkpoints, give fast feedback, review the work seriously, and know when exactly to jump in.
—— Paula Dozsa · [14:23]

指向原始笔记的链接

② 出现在这些集

1 集

• 《Tolan 如何做语音优先的 AI 陪伴体》 — 作为嘉宾

③ 他谈到的

点进去有真内容 —— 本页主要出口

Tolan · 智能体 · 延迟 · 记忆 · 上下文 · Claude · 检索 · LLM

④ 也在聊「智能体」的人