Research研究
We publish what we learn. Plain language, no gatekeeping, no email wall.我们把学到的东西公开发布。人话,不设门槛,不用留邮箱。
We write about what we're actually working on and reading. Every piece is meant to be useful to someone running a business, not only to someone building models. If a piece needs jargon, the jargon gets defined on the page.我们只写自己正在做、正在读的东西。每一篇都力求对经营企业的人有用,而不只是对做模型的人有用。如果一篇文章绕不开术语,术语就会在页面上被讲明白。
An identity for something that isn't anyone给一个「谁都不是」的东西一个身份
How do you trust a piece of software you've never met? On the early standards for agent identity, reputation, and accountability.怎么信任一段素未谋面的软件?谈智能体身份、信誉与追责的早期标准。
2026.08.14 · 6 min6 分钟 Agent Economics智能体经济学When software starts paying for things当软件开始自己付钱
Cards and invoices assume a human is present. What machine-to-machine payment actually looks like — and whether you need it (probably not yet).银行卡和发票都默认背后站着一个人。机器间支付真实的样子——以及你是否需要它(多半还不)。
2026.08.06 · 6 min6 分钟 Plain Language人话解释What is an agent harness?什么是智能体系统(agent harness)?
The model is the engine. The harness is the rest of the car — and it decides whether any of this works in your business.模型是发动机,harness 是车的其余部分——这套东西在你的业务里能不能用,由它决定。
2026.07.28 · 5 min5 分钟Get new research by email用邮箱接收新研究
Occasional. No marketing.偶尔发送,没有营销。