WHAT THIS IS

Most summaries fail because they are prose, only shorter. You read a compressed version and retain no more than you would from the blurb. Structure is what helps: the same seven fields on every page, in the same order, so the shape of an argument is visible rather than buried in a smaller version of it. The claim is separated from the evidence, which is what lets you see when a book is one idea and two hundred pages of support.

Every point opens to three depths, and you choose the depth per point rather than per page — because depth is not uniform. You usually want the full argument on the one point you disagree with, and the card on the rest.

These are books requested, not books recommended. A title appears here because it was handed over to be summarised — not because someone read it, enjoyed it and wants to tell you about it. That distinction decides what the ASSESSMENT field is allowed to claim, so it is worth being exact about: an assessment here is a position derived from the argument the book makes, in the edition named on the page. It is not a reader's recommendation, and there are no stars, no score and no five-point scale anywhere on the site, because a scale would imply a judgement this cannot honestly make.

Nothing here picked its own subject, and nothing here ever will. The publishing pipeline is automated — a source goes in, a draft comes out, it is reviewed, and on approval it is committed and rebuilt. What is deliberately absent from that pipeline is any capacity to choose what to write about. There is no keyword research, no gap analysis, and no “suggest the next title” anywhere in it. That absence is the line between a library and a content farm, and it is enforced by what was built rather than by a promise: the tooling has no such function to call.

Facts and ideas are not copyrightable, and a structural summary is transformative. A summary detailed enough to substitute for the book is not, so three limits are built in rather than intended:

No quotation beyond a sentence. Every field is parsed at build time and a quotation running past one sentence fails the build. No chapter-by-chapter structure. The fields are the structure; the format has no syntax for headings or numbered lists, so there is nowhere to write a retelling even by accident, and a chapter reference fails the build. Every page links to buy the book. Driving people to the work is both the honest thing and the strongest position to be in.

Sources are legally obtained. If you hold rights in a book summarised here and think a page has gone too far, write to [email protected] and it will be taken down while it is looked at, not after.

A connection is not a similarity score. Each book states its position on a named idea, with the scope of that position, and a connection exists when two books meet on the same idea — backing it up if they land on the same side, pushing back if they do not. The stance is derived from those positions rather than asserted, which is why the lanes and the connection rows on a summary page cannot disagree.

Every connection carries a scope and a note saying what is agreed or denied. “On means, not ends” is the ordinary case: two books can want the same thing for incompatible reasons, and an edge that does not say so is an opinion.

There is no view of every book at once and no edges are drawn anywhere. The honest answer to “show me the whole graph” is unreadable, so it is not offered.

summarylibrary.com is published by Delta Gamma LLC, which also publishes dlgm.co and several other sites. They share one email list and one disclaimer, and this is said here rather than left to be discovered, because finding it out later feels like a trick.

Generative AI is used to produce the drafts. Nothing publishes without a human moving it out of the drafts directory, and the mistakes these tools make should be assumed present anyway. The library currently holds 4 summaries across 3 ideas.