About & methodology

Disclaimer

An unofficial fan index. Not affiliated with or endorsed by Scott Alexander. Posts are © their respective authors; this site links to the originals and publishes only its own summaries and scores.

Site author’s view

The world would be much better if we all read more Scott Alexander. Yet, 1) we may not be convinced Scott Alexander is worth reading yet. 2) We don’t have the time to read his entire corpus.

It is therefore helpful to have a guide that can point us to his most important works, and if there is an area of specific interest, explore that.

Goal

There is too much for any of us (unless we’re some historical scholar), to figure out where we should start reading Scott’s works. How do we find the best of over 1,800 posts?

Use Claude Opus 4.8 to create a compendium. An independent and neutral review and ranking, indexed, sortable and filterable, will help new readers with that endeavor.

What it is not

This is not designed to be a critique of his work. Although Claude, based on the site author’s parameters, has given values to the quality of the writing, its paradigm shift to Claude (no proof that’s even remotely correct introspection), and real-world impact, don’t put too much stock in it. We humans need some heuristic to easily make our next reading decision.

What I hope it is

#3 is probably, on the whole, going to impact you more than #23, and almost certainly more than #230. This is for interest’s sake, and to help you more quickly find great essays! To state it again, I’m not trying to be a critic, I’m trying to help readers find the best Scott has to offer within a huge corpus of writing. This requires us to in some way say this one is better than that one and you should read it.

Scope

Nothing here is breaking news, by design: articles do not appear in this index until at least three months after publication. The point of the ranking is finding the historical gems, not hot takes. A new post deserves to be read and argued about on its own terms (and contest entries voted on) before a score sits next to it. Posts join automatically once they’ve aged in.

1853 posts have been scored so far (Astral Codex Ten 786 · Tumblr 40 · Slate Star Codex 894 · LessWrong 113 · LiveJournal 17 · ai-2027.com 1 · Unsong 1 · raikoth.net 1). This is a work in progress: the index grows as more of the archive is scored, and is not yet a complete catalogue.

Inclusion & exclusion

The aim is to cover Scott’s substantive standalone writing across his venues, listed here from his most current to his earliest: Astral Codex Ten (his active blog since 2020); his Tumblr, Slate Star Scratchpad, which still sees the occasional post; Slate Star Codex (2013–2021), the main body of the work; his earlier LessWrong essays, written under the pen name Yvain; and his earliest LiveJournal posts — plus a few one-off pieces published elsewhere. In practice that means his original essays, book reviews, fiction, and long “much more than you wanted to know” investigations.

The Tumblr needs a word of its own. It is mostly a conversational space — reblogged threads and answers to reader questions — rather than a place for finished pieces, so it is not covered exhaustively. Instead, a curated selection of its genuinely standalone posts is included: short fiction and parables, sustained humor pieces, self-contained essays, and a handful of memorable personal ones. Reblog chains and ask-box replies are left out.

Deliberately left out everywhere are posts that aren’t standalone works: open threads, meetup and housekeeping announcements, and link round-ups. A small number of guest posts (for instance, Book Review Contest entries) are included and clearly flagged as guest.

Coverage is a work in progress and is not yet the complete archive; more is added as it is scored. The selection was carried out by Claude following these criteria, so a post may occasionally be wrongly included or overlooked.

How scoring works

Every post is judged on four axes, each against fixed anchor bands: two 1–100 scores (Quality and Claude’s paradigm shift at publication) plus real-world impact (RWI, 0–10) and a light Humor rating (0–5). The scores are never blended into one number; the ranking sorts by quality, and Humor is purely descriptive, so it never affects the order.

Quality

Impact, insight, reframing and explanatory power, as the piece reads today, including its real-world significance.

Score	Label
90–100	Landmark
75–89	Excellent
60–74	Strong
40–59	Solid
20–39	Minor
1–19	Negligible

Claude’s paradigm shift at publication

How novel and disruptive the ideas were in their historical context to Claude, judged in the post’s own time. The goal is to measure how much the article would have shifted Claude’s beliefs, taking into account that much of it is already in Claude’s training data. Does this actually work? Who knows; Claude said it was reasonable. LOL. Early landmarks whose ideas are now mainstream still score high here; they are not penalised for having won.

Score	Label
90–100	Paradigm-defining
75–89	Major shift
60–74	Notable shift
40–59	Moderate
20–39	Slight
1–19	None

RWI: real-world impact (0–10)

A separate measure of concrete change in the material world and in ordinary lives (laws, institutions, professional practice, widely-adopted tools). Whether that change reached a broad population rather than influence confined to a narrow elite was a consideration that adjusted the values. Shown where assessed.

Score	Label	What it means
9–10	Transformational	Large-scale, verifiable change to laws/institutions/practice and everyday lives.
7–8	Major	Built durable institutions or moved real policy/practice in actual jurisdictions, and/or affects the material conditions of many ordinary people. (e.g. Georgism = 8)
5–6	Substantial	Broad discourse adoption, or strong influence across a large professional field, with some concrete spillover. (e.g. Moloch = 6)
3–4	Moderate	Influential within a niche or professional sphere; vocabulary used by an educated subculture.
2	Minor	Small community / within-blog influence.
0–1	Negligible	Little/no real-world effect, or too recent to assess.

Humor (0–5)

An escalating ladder of five Scott Alexander in-jokes, ordered by how hard you laughed, with 0 reserved for posts that never set out to be funny. Like RWI, it sits outside the ranking — a purely descriptive axis. Only significant, intentional comedy scores above 0; Scott’s habitual incidental wit doesn’t count. Each level is named for the SSC/ACX post it riffs on:

Badge	Label	What it means
—	Not Intended to Be Funny	The article had no intention of being funny — humor not applicable.
	Lizardman’s Constant	~4%, statistical noise. You’re not actually sure you laughed.Lizardman’s Constant Is 4%
	Man of One Study	A single chuckle. Could be a fluke; needs replication.Beware The Man Of One Study
	Scissor Statement	Funny enough that half your friends are now fighting about whether it was funny.Sort By Controversial
	Moloch	You are laughing against your will, swept up by forces beyond any individual’s control.Meditations On Moloch
	Cactus Person	Transcendent, DMT-tier, you have left your body.Universal Love, Said The Cactus Person

Contest winners

Scott has run two reader contests whose top three finishers, as announced in his own results posts, carry a gold, silver or bronze podium badge next to their entry — a purely descriptive mark that, like Humor, sits outside the ranking. Entries that placed below the top three are still flagged as guest posts but carry no badge.

The Review Contest (usually books; the 2025 edition was open to any topic) uses an open-book badge:

Badge	Place	What it means
	First Place	The winning entry — gold.
	Second Place	Runner-up — silver.
	Third Place	Third place — bronze.

The earlier Adversarial Collaboration Contest, in which two people who disagreed co-wrote a single piece, uses a converging-arrows badge — two opposing sides meeting on common ground:

Badge	Place	What it means
	First Place	The winning entry — gold.
	Second Place	Runner-up — silver.
	Third Place	Third place — bronze.

Whose judgement is this?

I chose the categories and the framework for valuing the articles. Within that framework, I asked Claude to do its own analysis (avoiding others’ reviews and commentary), though it could follow data links and search for evidence to strengthen or refute Scott’s statements.

Claude was required to read each article in full, never truncating it. For a sense of scale, the evaluation phase took on the order of thirty-three (33) context windows run to 800–850k tokens of Opus 4.8. Each new session reinitialises the whole project’s parameters, and the evaluation rules and anchors are reloaded before every article.

There was a good deal more to it than this short summary, and I’m not interested in nitpicking the details. Is it actually Claude’s own analysis, without any input from web commentary or critique? I’m not betting it is. Could I have done better prompt engineering? Almost certainly. Would it have changed the rankings much? Probably not.

One deliberate exception: given its length, I didn’t require Claude to read all of Unsong (Scott’s novel), and I let it draw on outside reviews and critique for that single entry.

The results seem reasonably consistent with general chatter online, and quite effective for the purpose; a quick heuristic for the next thing to read.

Privacy & cookies

This site uses Google Analytics to understand, in aggregate, how visitors use it. It loads only after you accept on the cookie banner — nothing is loaded and no cookies are set unless you opt in. If you accept, Google Analytics sets its own cookies (for example _ga) to measure visits; if you decline, it is never loaded. You can change or withdraw your choice at any time via the Cookie settings link in the footer. No accounts, comments, or personal data are collected by this site itself.

Reuse & licensing

The summaries and scores on this site are licensed CC BY 4.0. Reuse them freely with attribution. Scott Alexander’s original writing remains his; this site only links to it.

Data last generated 2026-08-20.