Bring a question you are still working out. Grayy plans a book around it, writes it from a library of 2,835 real books, papers and case studies, and cites them on the page.
Scroll. Thirteen sources become one chapter.China is especially important because it complicates any account that treats sustained growth as a direct consequence of early technological superiority. Mokyr presents China as a leader in several technical fields before 1400, including metallurgy, printing, and navigation, while arguing that its progress slowed after 1400 as centralized bureaucracy became more conservative and political competition weakened.[7] In this account, China demonstrates that an early lead can be lost. Technical capability is therefore an initial condition, not a guarantee of continuing dynamism.
Nathan Rosenberg and L. E. Birdzell offer a related comparison through political structure. They argue that centralized empires such as China could possess advanced technology while still giving political authorities strong reasons to preserve stability rather than accommodate disruptive economic change.[8] Their claim is not that Chinese society lacked ingenuity. It is that a single authority could apply a restrictive decision across a large political territory, whereas a European innovator or merchant might find another jurisdiction willing to accept the activity.
Grayy asks a question or two back, then proposes a read. Before it writes a word, you set how it speaks, how deep it goes and how far it ranges.
The conversation is a demo. The three reads it proposes are real ones Grayy has written.
The status line in the app is the real order of work. Six stages sit between your question and the first chapter. Here is each one, using the read from the top of this page.
Your brief, your profile and the three settings become one contract for the read.
Grayy maps the library against your question and decides what belongs. It also decides what does not, and says why.
Chapters are planned as one argument. Each gets an objective and the sources it may draw on.
Each chapter is written from cited passages, in parts, then joined into a single piece.
Attribution and source accuracy are checked on every chapter. What fails is repaired and checked again.
The finished read lands in your library. You can leave while it is written; it keeps going.
What you asked for, who it is for and what to avoid travel with the read through every later stage.
Grayy writes only from its own library. It does not browse the web while it writes.
The plan Grayy actually made for this read. Chapter eight exists to say where the evidence runs thin.
A paragraph from chapter four, exactly as it was written.
If too much cannot be verified, the read is held back and you are told why.
The order of checks every chapter goes through. The states shown are an illustration.
A book takes a while to write. You can close the tab; the read is in your library when it is finished.
Book type, a column you set, and nothing else on screen. This is chapter four of the same read. Change the settings and the page follows.
Joel Mokyr’s answer begins by separating two forms of useful knowledge that are often compressed into the single word technology. Propositional knowledge, or Q-knowledge, concerns what people believe about natural phenomena and their regularities. Prescriptive knowledge, or A-knowledge, concerns the techniques, instructions, and recipes used to manipulate those phenomena in the production of goods and services. The distinction is analytical rather than absolute. A technique may exist without a well-developed explanation of why it works, while a proposition about nature may have no immediate practical use. Yet sustained innovation depends on their interaction.[3]
Q-knowledge therefore matters less as a stock of abstract propositions than as a source of controlled variation. When practitioners understand why a process works, they can alter its inputs, sequence, scale, or application with some expectation of what will follow. The resulting knowledge is still imperfect, but it is more portable than a recipe tied to one setting. A prescriptive rule can be copied. A proposition can help explain when the rule should be copied, when it should be modified, and when it should be abandoned.
The relationship also runs in the other direction. Practice does not merely wait for science to deliver useful applications. New techniques can expose anomalies, generate measurements, and create problems that existing explanations cannot resolve. In Mokyr’s account, this friction between what people can do and what they can explain is productive. The result is a positive feedback loop rather than a one-way pipeline from science to industry.[3]
Nothing marked in this chapter yet. Select text to highlight it or add a note.
A read does not scroll away when you finish it. You can question it, mark it, and come back to the argument without reopening the whole thing.
The Great Divergence is often presented as a question about when Europe became richer than China. That formulation is too narrow.
Grayy is written by AI models from sources you can see. This is what stands between a first draft and your library, with the record of the read from the top of this page.
Every read in Grayy is written by AI models from the sources in Grayy’s library, with the passages it drew on cited inline. The model that writes a read is never the one that judges it.
How Grayy uses AI.
Before anything is planned, Grayy maps the library against your question. A topic the library cannot carry stops there, and you are told why.
Each one has to resolve to a real passage in the library. A citation that points at nothing blocks the read.
It comes from a different provider than the writer. It checks that the cited passages support what the chapter says and that the chapter does its job. What it rejects is rewritten and reviewed again.
All eight chapters together, for what a chapter-by-chapter review cannot see: whether the argument holds from start to finish, whether chapters repeat each other, and whether each link between fields carries weight.
A read that falls short is rewritten from the first chapter affected and judged again. If it still does not pass, it is not published.
Each read records what it left out and why, and its last chapter says where the evidence runs thin.
We don’t yet have enough reliable source material in the library for this topic. We’re actively adding sources, so try a related topic or check back soon.
Try againWe held this read back because 3 passages couldn’t be matched closely enough to their sources, so it isn’t published as it stands. We’re reviewing it.
And when a read does not pass, these are the exact messages a reader sees instead. Nothing is written around a gap.
Loading sign-up…
Create a free account and ask for the book you need. A free account gets one read a month, and a read takes a while to write.