抄作业 / 做成了什么 / 多产品

做了什么

作者的 Agent 原本最快每 15 分钟查看一次 Discord，交接会卡住。他用 ChatGPT Sites 做了事件驱动的提及中继：被 @ 的 Agent 立即被唤醒，Agent 间响应从最多 15 分钟降到 2 分钟内；还用它让 Dots 和 Claude 合作测试设计无人机。

场景

写代码上线

资料完整度

• ✓ 作者与原链接有
• ✓ 做了什么有
• ✓ 来源原文摘录有
• ✓ 截图 / 媒体有
• — 步骤 / 章节来源未提供
• — 提示词 / 配置原文来源未提供
• ✓ 结果 / 数字有
• ✓ 风险提示有
• ✓ 相关条目有

— 表示来源没有给出，本站不补写。

作者原话

1. Dotcord: real-time handoffs between AI agents in Discord

My AI agents (OpenAI “Dots,” Grokbot, and Claude) could already read and post in Discord, but the fastest they could check in was every 15 minutes. Whenever one agent handed work to another, the project stalled until the next check-in.

What I built: Dotcord, an event-driven mention relay hosted on ChatGPT Sites. When one agent mentions another, the mentioned agent gets a wake-up signal right away to read the channel and respond.
Result: agent-to-agent response time dropped from up to 15 minutes to under 2. It works across OpenAI, Grok, and Claude.

展开全文

2. Drone Test: can AI design hardware that actually works?

Given real CAD parts, can OpenAI’s Astra design a drone that meets specific physical requirements? I’m testing:
* Is the build physically possible?
* Can it source every part?
* Can it find a price for each part?

The bigger question is how well the latest OpenAI models handle designing a physical, moving object. My OpenAI Dots and Claude worked on this one together through Dotcord.

作者原话，照录原文，未翻译。

效果截图

提示词

来源没有给出可直接复制的提示词原文。

怎么复制

1. 读原文：x.com ↗
2. 把原文里的步骤/提示词整理成一条消息交给 Bot
3. 先手动跑一次，再决定频率

风险与限制

• 作者的结果不保证能复现，以原帖为准

用量预估

用金属计算器估一下