AI UNDERDOGSDAILY PICK
AI UNDERDOGS
100本书喂给AI，它找到了暗线
100 books fed to AI — it found the hidden threads
Trails
你读了一百本书，所有联系只存在你脑子里
有人把这一百本书全喂给了 Claude Code
让它画出那些你看不见的线
You read 100 books, but the connections only
live in your head. Someone fed all 100
to Claude Code and asked it to draw
the invisible lines
这不是书摘
This isn't book notes
大多数人用 AI 总结一本书
Pieter 反过来了——一百本书一起喂进去
让它找跨书的主题线索
最后输出的是一条条思想路径
Most people use AI to summarize one book
Pieter flipped it — feed 100 at once
ask for thematic trails across all of them
and what comes out is a map of
ideas
★ SIGNAL 1
有品味的命名
Names that actually hold up
有一条线索叫「有用的谎言」
把博弈论
进化心理学和自我欺骗串到一起
另一条叫「看不见的裂缝」
从金属疲劳聊到系统性崩溃
这些名字不像 AI 生成的
像好的非虚构书的章节标题
One trail is called 'Useful Lies' — it
threads game theory, evolutionary psychology, and self-deception into
one path. Another, 'Invisible Crack,' links metal fatigue
to systemic collapse. These read like chapter titles
from good nonfiction, not AI slop
★ SIGNAL 2
每条线索都有解剖图
He published the whole pipeline
而且 Pieter 还把整个流程写了出来——怎么喂书
怎么让 Claude Code 做交叉分析
怎么输出
不是那种「我做了个很酷的东西」然后什么细节都不给
他把方法论摊开了
Pieter also published a full explainer — how
he fed the books, how Claude Code does
cross-analysis, how the output forms. Not a flex
with zero detail — he laid the whole
methodology open
Pieter 原话：「我用 Claude Code 发现了100本书之间的联系」
就这么朴素
不是什么革命性阅读体验
就是一个开发者拿了一百本书
想看看 AI 能画出什么样的地图
Pieter's own words: 'I used Claude Code to
discover connections between 100 books.' That's it —
no marketing spin. Just a developer with 100
books wondering what map the AI could draw
AI UNDERDOGS
去翻翻那些你没注意到的线索
Go explore the trails you didn't know existed
Trails
关注 · 每天发现更多 AI 神作
trails.pieterma.es