AI 构建者日报 | 6月7日

I think what's most exciting about this time is that I think it's gonna be, like, a billion, maybe even more than that. I think there's this incredible time that we're moving through of product development and really the democratization of technology.

a smarter alternative to "always use plan mode": always frame your task as a question, so that the model is invited to push back and rate the quality of the idea/suggest alternatives, rather than blindly execute what you SAID to do (which is often not precisely what you MEANT)

把任务框架成一个问题，让模型有机会反驳你、评价想法质量、建议替代方案，而不是盲目执行你说的每个字（往往不是你真正想要的）。

We've doubled Claude Cowork usage limits for the next month. Delegate bigger, more complex tasks to Claude.

Claude Cowork 使用限额翻倍，持续一个月。把更大、更复杂的任务交给 Claude 吧。

Better memory = Shorter prompts = More utility per token

更好的记忆 = 更短的提示词 = 每个 token 更多实用价值。

How to build AI skills that check their own work and improve over time: Give it context. Make it easy to trigger. Add evals. Add memory. Build a skill to edit skills.

如何构建能自我检查、持续改进的 AI skill：给上下文、降低触发门槛、加入评估机制、加入记忆能力、再做一个"修改 skill 的 skill"。

One of the most common mistakes I see enterprise AI teams make is building for today's model capabilities and price points. Think 6 months out. Models will be way smarter and cheaper. Scaffold around today's model weaknesses to push the frontier. Build for the slope.

企业 AI 团队最常犯的错：为今天的模型能力和定价而构建。往前看 6 个月，模型会更聪明、更便宜。围绕今天的模型弱点搭建脚手架来推动前沿，为"斜率"而建。

Replit x Shopify!

Replit 与 Shopify 达成合作！

Agent filesystem state can now be read, written and mounted independently of Sandbox lifecycle. Storage that's decoupled, but attachable to Builds, Functions, Sandboxes & more.

Agent 文件系统状态现在可以独立于沙箱生命周期进行读写和挂载。存储与计算解耦，但可以附加到 Builds、Functions、Sandboxes 等。

Coding is basically the pinnacle of what you could reasonably automate with AI, and yet we still need human engineers to oversee agents for them to be effective. So if with all of that, engineers still remain in very high demand, the risks are going to be less than what's perceived for other areas of knowledge work.

编程基本上是 AI 能合理自动化的巅峰案例，但即使这样我们仍需要人类工程师监督 Agent 才能有效。既然连这种情况工程师都仍然高需求，那其他知识工作领域的风险可能比感知的要小。

designing in code is now as easy as: click, chat, hold shift to multi-select. works best with Composer 2.5

在代码里做设计现在变得像这样简单：点击、聊天、按住 Shift 多选。配合 Composer 2.5 效果最佳。

You can now finally try one of those big projects I was teasing that I started working on a few months ago. We want this to over time get better at helping you learn the best techniques to build better software, faster.

你现在终于可以试用我几个月前开始做的那个大项目了。我们希望它能随着时间推移，越来越擅长帮你学习最好的技术，让你更快地构建更好的软件。

Ok bad stories about VCs are spreading on X right now, but VCs have horror stories about founders too. Like, that one time when a founder decided to take another term sheet with a higher valuation despite our obvious ability to add value, thought leadership and vendor discounts.

现在 X 上到处是关于 VC 的负面故事，但 VC 也有关于创始人的恐怖故事。比如那次，创始人明明知道我们能加值、有思想领导力、能提供供应商折扣，却还是选了估值更高的另一份 term sheet。

We've doubled usage limits in Claude Cowork for the next month. Delegate bigger, more complex tasks to Claude.

Claude Cowork 使用限额翻倍，持续一个月。把更大、更复杂的任务交给 Claude。