Datasets:

yoonholee
/

style-eval-corpus

author stringclasses 11 values	title stringlengths 2 899 ⌀	url stringlengths 18 120	text stringlengths 9 1.26M	word_count int64 3 199k
Paul Graham	How to Do Great Work	http://paulgraham.com/greatwork.html	July 2023 If you collected lists of techniques for doing great work in a lot of different fields, what would the intersection look like? I decided to find out by making it. Partly my goal was to create a guide that could be used by someone working in any field. But I was also curious about the shape of the in...	11,827
Paul Graham	Having Kids	http://paulgraham.com/kids.html	December 2019 Before I had kids, I was afraid of having kids. Up to that point I felt about kids the way the young Augustine felt about living virtuously. I'd have been sad to think I'd never have children. But did I want them now? No. If I had kids, I'd become a parent, and parents, as I'd known since I was ...	1,537
Paul Graham	How to Lose Time and Money	http://paulgraham.com/selfindulgence.html	July 2010 When we sold our startup in 1998 I suddenly got a lot of money. I now had to think about something I hadn't had to think about before: how not to lose it. I knew it was possible to go from rich to poor, just as it was possible to go from poor to rich. But while I'd spent a lot of the past several years s...	678
Paul Graham	The Brand Age	http://paulgraham.com/brandage.html	March 2026 In the early 1970s disaster struck the Swiss watch industry. Now people call it the quartz crisis, but in fact it was a compound of three separate disasters that all happened at about the same time. The first was competition from Japan. The Swiss had been watching the Japanese in the rear view mirr...	7,663
Paul Graham	The Shape of the Essay Field	http://paulgraham.com/field.html	June 2025 An essay has to tell people something they don't already know. But there are three different reasons people might not know something, and they yield three very different kinds of essays. One reason people won't know something is if it's not important to know. That doesn't mean it will make a bad ess...	769
Paul Graham	Good Writing	http://paulgraham.com/goodwriting.html	May 2025 There are two senses in which writing can be good: it can sound good, and the ideas can be right. It can have nice, flowing sentences, and it can draw correct conclusions about important things. It might seem as if these two kinds of good would be unrelated, like the speed of a car and the color it's pain...	1,645
Paul Graham	What to Do	http://paulgraham.com/do.html	March 2025 What should one do? That may seem a strange question, but it's not meaningless or unanswerable. It's the sort of question kids ask before they learn not to ask big questions. I only came across it myself in the process of investigating something else. But once I did, I thought I should at least try to a...	1,590
Paul Graham	The Origins of Wokeness	http://paulgraham.com/woke.html	January 2025 The word "prig" isn't very common now, but if you look up the definition, it will sound familiar. Google's isn't bad: > A self-righteously moralistic person who behaves as if superior to others. This sense of the word originated in the 18th century, and its age is an important clue: it shows t...	6,168
Paul Graham	Writes and Write-Nots	http://paulgraham.com/writes.html	October 2024 I'm usually reluctant to make predictions about technology, but I feel fairly confident about this one: in a couple decades there won't be many people who can write. One of the strangest things you learn if you're a writer is how many people have trouble writing. Doctors know how many people have...	555
Paul Graham	When To Do What You Love	http://paulgraham.com/when.html	"September 2024 \n \nThere's some debate about whether it's a good idea to \"follow your passion.\(...TRUNCATED)	1,545

End of preview. Expand in Data Studio

Style Eval Corpus

Writing from 11 internet writers with instantly recognizable but distinct styles. Built for style-as-reward-inference experiments: given a writer's body of work, infer their implicit reward function and encode it as eval components (rubrics, classifiers, probes).

Author	Pieces	Words	Register	Source
Paul Graham	229	562,990	Contrarian startup essays	paulgraham.com
Andrej Karpathy	34	101,337	Tutorial-as-thinking-aloud	karpathy.github.io + bearblog
Gwern Branwen	220	2,800,357	Quantitative empiricist	gwern.net (CC-0)
dril	7,494	150,557	Absurdist authority	crumb/dril-tweets
Donald Trump	46,683	924,881	Superlative combative	fschlatt/trump-tweets
Derek Sivers	546	231,786	Zen-minimalist aphorisms	sive.rs
Maciej Ceglowski	353	524,614	Sardonic literary tech critic	idlewords.com
Scott Alexander	137	560,874	Dense calibrated reasoning	astralcodexten.com
Naval Ravikant	52	46,538	Compressed aphoristic wisdom	navalmanack.com (CC)
Joel Spolsky	209	123,984	Narrative tech management	joelonsoftware.com
Eliezer Yudkowsky	1,496	4,753,270	Rationalist pedagogy	lesswrong.com
Total	57,453	10,781,188

Design

The corpus spans several deliberate contrasts:

Long-form analytical (PG, Karpathy, Gwern, Scott Alexander, Ceglowski, Joel, Eliezer) vs short-form (dril, Trump, Sivers, Naval). Same eval framework, different registers.
Same genre, different rewards: PG / Karpathy / Gwern / Scott Alexander / Eliezer all write tech/rationalist essays but with radically different implicit reward functions. Within-genre discrimination is the hard task.
Corpus size varies deliberately: Karpathy (34 posts) and Naval (52 chapters) are the few-shot setting. Trump (46K tweets) and Eliezer (1.5K posts) are data-rich. Tests generalization from thin vs thick reference sets.

Schema

Column	Type	Description
`author`	string	Writer name
`title`	string or null	Title (null for tweets)
`url`	string	Source URL
`text`	string	Full text (markdown for essays, plain for tweets)
`word_count`	int	Word count

License

Gwern Branwen: CC-0 (public domain).
Naval Ravikant: Almanack is CC-licensed, explicitly free.
Eliezer Yudkowsky: LessWrong content is CC BY 4.0.
dril / Trump tweets: sourced from existing HF datasets.
All others: publicly posted writing, research use under fair use.

Curation/parsing released under CC0.

Downloads last month: 53

Style Eval Corpus

Contents

Design

Schema

License