Hi X! I'm building Knowledge Delta, and here's my 30-second introduction. I'm a researcher in applied economics and rural development, exploring how AI is changing research, education, and the way we think. I'm also building AI Skills and Agents for real research workflows. Even the hand-drawn animation in this video was created using a Skill I developed, after quite a few rounds of trial and error. Now I'm curious: If you could have an AI Agent handle one part of your research workflow, what would it be? Reading papers? Cleaning data? Running analyses? Writing? Teaching? I'd love to hear your ideas. I might build the most requested one next. Ask better. Use AI better. Judge better.

A thought on the people-centered view of history in the age of AI: As AI becomes increasingly capable of creating economic value, will people still have the power to shape the direction of technological progress and share in the benefits it creates?

I still need to set aside real time for reading, even if it's just ten or twenty minutes. I used to think I could sneak in a few pages while waiting for the AI to respond. Turns out attention doesn't switch that easily. Book in hand, but part of my mind is always wondering if the AI's almost done. Sometimes it's fast, sometimes it's slow, and the reading never really sinks in.

Rereading Classics 𝚫 Vol.1 #003 Schumpeter, Capitalism, Socialism and Democracy, Chapter 1: Marx the Prophet The Original Schumpeter reads Marxism as a kind of religion: it explains the discontent of the masses and promises paradise on this side of the grave. He also notes that before the 20th century, Marxism was largely absent from the American labor movement and from American intellectual life. American workers struck and fought, yet no such faith took shape. In Schumpeter’s account, for discontent to become doctrine, someone has to interpret it as a worldview. The 𝚫 In recent weeks, France has seen nationwide high school protests. They began over run-down schools, crowded classrooms, teacher shortages, and an education budget rising slower than inflation. Street clashes and mass arrests followed. The government calls it rioting. Protesters call it demands. Parties accuse each other of fanning the flames. The same discontent is being interpreted by many voices at once. In Schumpeter’s time, thinkers and parties interpreted discontent. Today, media and algorithmic feeds join them. As more people come to understand the world through AI, who will interpret discontent next?

现在经常反复看以前的老电影老电视剧，《大明王朝1566》已经看了不知道多少遍，《大决战》三部也刷了好几遍。 新剧点开没几分钟就关了，老的反倒越看越有味道。 有时候会想：这是不是说明我老了，接受不了新东西了？😂😂😂

冷知识：爱因斯坦也被审稿人气到过。 1936 年他给《Physical Review》投稿，编辑把稿子送去外审，审稿人提了一堆意见。爱因斯坦很不高兴，回信说他没同意在发表前给别人看，直接撤稿，转投了别的期刊。

如果你上大学那会儿就有现在的 AI， 你觉得自己会变得更厉害，还是更懒？说实话。

网站做了快两周，还没上线。每次觉得快好了，就想再加一个功能。我开始理解为什么我的论文也总是"快好了，再等等"。拖延癌晚期！😂

碎碎念：在 AI 时代，我反而比以往更想读老书。也许是因为，写它们的人有时间慢慢想。

Stata 什么时候能直接内置一个 AI？不求多聪明，能少复制几次就行。

写稿子、写项目，能一鼓作气就一鼓作气吧。我这篇初稿中间耽误了几天，今天重新捡回来，发现得从头再理一遍思路，当时想到哪儿、为什么这么写，全忘了。所以留痕真的很重要。

Lately I've had Codex, Claude and Grok all running at once, plus Obsidian on my laptop, and the speed of it all feels a bit unreal. So I've been sticking to handwritten task lists. The few minutes while the AI runs? That's my writing time. A little old-fashioned, but it works for me.

Lately I've had Codex, Claude and Grok all running at once, plus Obsidian on my laptop, and the speed of it all feels a bit unreal. So I've been sticking to handwritten task lists. The few minutes while the AI runs? That's my writing time. A little old-fashioned, but it works for me.

Diminishing returns on land: keep adding fertilizer, gains keep shrinking. Same with letting AI polish the same paper. After round five, it starts getting worse.

Cochrane's agricultural treadmill: early adopters profit, then everyone adopts, prices drop, nobody's better off. AI paper-writing: early users published faster. Now everyone uses it, and journals just cut acceptance rates.

Engel's law: the richer you get, the smaller the share you spend on food. My food share is impressively low. Mostly because my AI subscriptions are eating the budget. 恩格尔定律：收入越高，吃饭的开销占比越低。我的恩格尔系数很低。主要是因为 AI 会员费占比太高了。

Dunning–Kruger: the less you know, the more confident you are. AI is the perfect mascot: it sounds most certain exactly when it knows least.