① 他说过的话

15 条

所以我认为这个结构、这个架构其实归结为一个问题，那就是：谁为智能体犯错买单？
And so I think that structure, that architecture really comes down to one question and that is who pays for the agent being wrong?
—— Abhi Arya · [02:35]

指向原始笔记的链接

而当它搞砸的时候——它肯定会搞砸——你完全不会察觉，因为它对自己在做的事情非常自信，而你要为它的错误买单，对你的客户、你的用户等等。
And when it messes up, which it will, you won’t have any idea because it’s really confident about what it’s doing, and you’re the one that’s paying for it being wrong to your customers, to your users, et cetera.
—— Abhi Arya · [02:55]

指向原始笔记的链接

最后一个桶，也是我认为会胜出的桶，是服务用户的智能体体验。
And lastly, the bucket that I think wins here is agent experience in service of the user.
—— Abhi Arya · [03:27]

指向原始笔记的链接

这个框架使智能体体验不再是一个关于模型的问题，而更多是一个关于上下文、能力和整体自我提升的问题。
This framing is what turns agent experience into less of a question of the model and more of a question of context, capability, and overall self-improvement.
—— Abhi Arya · [03:47]

指向原始笔记的链接

现在智能体没法在我们已知为真的部分上自由发挥了。
Now the agent doesn’t get to freelance the parts that we know are true.
—— Abhi Arya · [07:29]

指向原始笔记的链接

智能体不再猜测，开始引用。
The agents stopped guessing and started referencing.
—— Abhi Arya · [08:00]

指向原始笔记的链接

智能体本身非常自信地错了，而智能体所拥有的不确定性完全不可见。
The agent itself is very confidently wrong and the uncertainty that the agent has is completely invisible.
—— Abhi Arya · [09:30]

指向原始笔记的链接

在我们的系统里，或者说在 Reducto，失败携带着信息。
In our system, or with Reducto, failures carry information.
—— Abhi Arya · [09:46]

指向原始笔记的链接

所以我们奖励的是提问，而不是猜测。
So we rewarded asking instead of guessing.
—— Abhi Arya · [10:47]

指向原始笔记的链接

智能体基本上满足的是你所写的东西，而不是你想要的结果。
The agent basically satisfies what you wrote and not the outcome that you wanted.
—— Abhi Arya · [11:14]

指向原始笔记的链接

而在底层，它们基本上是同一个形状：智能体拥有的自主权和访问权限超出了人类所能验证的范围。
And underneath, they’re basically one shape where the agent had more autonomy and more access than the human could verify.
—— Abhi Arya · [12:02]

指向原始笔记的链接

智能体只是盒子里的智能，而你围绕它们构建的 harness 才是赋予它们能力、在下游交付这些成果的东西。
Agents are just intelligence in a box and the harness that you surround them with is what provides them with the ability to deliver these outcomes downstream.
—— Abhi Arya · [12:59]

指向原始笔记的链接

在我们不用手工重写的情况下，这就变成了每个组织或每个用户构建这些流水线时的一个自我改进循环。
Without us rewriting by hand, this then became a self-improving loop per org or per user that was building these pipelines.
—— Abhi Arya · [13:45]

指向原始笔记的链接

但除此之外，关键在于：当你的智能体自信地出错时，谁会发现？他们多快会发现？
But beyond that, it’s when your agent is really confidently wrong, who finds out and how fast do they find out?
—— Abhi Arya · [15:54]

指向原始笔记的链接

如果你为此而构建，那么你就能构建出人们真正愿意托付实际工作的东西。
If you build for that, then you can build for something that people will actually trust with real work.
—— Abhi Arya · [16:03]

指向原始笔记的链接

② 出现在这些集

1 集

• 《智能体出错的账，该谁付？Reducto 删掉重做 MCP 的教训》 — 作为嘉宾

③ 他谈到的

点进去有真内容 —— 本页主要出口

Reducto · 智能体 · 智能体体验 · MCP · Claude Code · 销售管线 · 不确定性

④ 也在聊「智能体」的人