D
437 posts


Naval 说过:
你是你接触的信息环境的平均值。
关注谁、读什么、和谁聊——这三件事加起来,就是你大脑的操作系统。
但问题来了:这个平均值是多少,你测过吗?
于是写了个 prompt,2 分钟测出你的信息环境质量。
复制粘贴到任意 AI 对话框,填空就行
——
我想做一次信息环境体检。
我回答 3 个问题,请你给出诊断。
Q1 我每天花最多时间看的 3 个信息源:___
Q2 最近一周读过的东西里,我现在还能说出来的有:___
Q3 遇到问题时最常找谁聊:___
请输出:
① 一句话诊断(用比喻)
② 最大优势
③ 最大漏洞
④ 一个今天就能做的改变
——
我自己测完才发现:技术阅读质量挺高,但和谁聊那栏几乎空白——搁那天天跟 AI 对话呢。
聪明人不问我怎么更努力,只问我应该把自己放在哪里。
先测,再调。2 分钟,可能改变你下个月的信息消费方式。
中文
D retweetledi

🦞 ClawLibrary: The Generated Asset Index & Monitor for OpenClaw
My project ClawLibrary is officially open-source! 🥰 It's a retro 2D pixel-art style asset management UI for OpenClaw.
The backstory: As my openclaw gets smarter, the generated assets are piling up. Finding things became very time-consuming. Inspired by @ring_hyacinth and @simonxxoo 's Star-Office-UI, I built it a dedicated library!
Current features:
🗂️ Full Asset Index & Preview: Seamlessly browse & preview almost everything OpenClaw generates (docs, code, images, Skills, memories, cron jobs, logs, etc.) right in the UI.
📊 Real-time Dashboard: Keep an eye on the lobster's config, task queues, active models, health status, and error tracking at a glance.
👀 Dynamic Tracking: Visually track real-time access and modification history of your resources.
🎨 Highly Customizable: Swap out animated characters, plus seamless English/Chinese support.
It's completely open-source now! If you like it, a little Star 🌟 would mean a lot! Issues and PRs are super welcome to help improve it together! 🥹
🔗 github.com/shengyu-meng/C…
#openclaw
Simon Meng@meng_shengyu
Inspired by @ring_hyacinth & @simonxxoo, I overhauled their "Lobster House" into a "🦞 Lobster Library"! I was drowning in scattered generated code, images, and docs, plus its memories and logs. So I built this unified hub to track what Lobster is reading in real-time, browse recent gens, and index local files. It's still a rough WIP but if y'all are interested, I'll polish it up and open-source it! #openclaw
English

突然有点理解马斯克总说一些不着边际的话了,因为他是尼采口中的超人,不被旧世界束缚。
我问chatgpt我是超人吗,他说虽然我也有超人的发动机,但还是处于既想超越旧世界又被旧世界牵动
Long Chen@LongChenNotes
中文
D retweetledi

@big_duca Someone has to prompt the Claudes, talk to customers, coordinate with other teams, decide what to build next. Engineering is changing and great engineers are more important than ever.
English

地狱难度的餐饮业
之前看到国内餐饮业 三年倒闭率高达70% ,平时也会想到,写几点感受:
1. 供给过剩
因为经济下行,传统制造业和服务业面临挑战,大量劳动力和资金需要寻找新的出路。餐饮业因其“低门槛”假象,成为一个“蓄水池”,加剧了供给过剩。
中国餐饮门店数量超过 900 万家,而美国只有不到 65 万家,中国人均餐厅数量是美国的 3 倍多。
2.内卷导致业级补贴
餐饮业实际上正在通过从业者的高投入和价格内卷,变相补贴消费者和房东,陷入了“毛利覆盖成本但净利润为负”,形成了一种不可持续的“虚假繁荣”。
以前外卖大战是平台补贴消费者,现在是餐饮创业者,在用自有资金或借贷补贴消费者,形成了一种餐饮行业级的补贴。
3.沉没成本阻碍理性退出
市场出清机制失效,创业者因前期投入(装修、设备、房租)过大,形成沉没成本依赖,即使连续亏损仍选择续租硬撑,进一步延缓市场出清,使供给过剩周期被人为拉长。
勇哥连线的时候,有太多人不服输,想继续硬撑,继续赔,比如"生蚝哥"...
基于我的理解,因为餐饮行业的竞争是完全互斥的,国内的餐饮行业,已经是地狱难度了。
只要你有利润能赚钱,就有人想着先不赚钱来抢你生意。
在很长一段时间内,大体只有这两个选择了:
1.愿意接受夫妻店式的薄利多销,赚个辛苦钱;
2.有机会几十万投个蜜雪冰城、瑞幸这类强势品牌,准备长线运营个几年(或10年)
餐饮业的暴利时代,已经结束了。
如果你想搞餐饮,建议多想想,多看看勇哥的连线
千万不要用你的存款,来开店补贴大家点外卖。
中文
D retweetledi

I have a theory: the amount of information provided in the context differs significantly for these two types of agent scaffolds.
And this causes the reasoning model like o1 to perform differently.
Reasoning models are trained to THINK hard, e.g., by solving extremely challenging math and/or leetcode problems. One commonality for these problems is that they are usually self-contained (e.g., all the information required to solve the problem is given beforehand). And because all the information is available, you can likely get better at solving the task simply by thinking.
But real-world problems are different. To solve a Github issue, you need to gather context (e.g., relevant code, documentation, etc.) because it is very likely that the Github issue description is not self-contained. This not only requires the model to (1) solve a problem well when complete information is provided, but it also needs (2) the ability to proactively act to gather context.
Scaffolds like Agentless use decomposed prompt + workflow to fix (2) context gathering for these reasoning models, so the model can just focus on performing what they are trained to do the best -- (1) thinking.
For a generic scaffold like OpenHands/CodeAct, when the agent fails to do (2) context gathering, no matter how strong its ability to think, it is likely not going anywhere: You can't solve a problem by thinking/imagining IF you don't even know what exactly the problem is.
English


Can anyone's @cursor_ai agent modify python notebook cells? mine always tries and keeps apologizing that it can't and creates python files for me to copy from 😅
Like bro, if I wanted to copy paste I'd use the chat


English

Just tested DeepSeek v3 in AG2 (seamlessly supported). It's insanely good considering the price🤯! A quick instruction: github.com/ag2ai/ag2/issu…
English










