A manuscript is a forgiving test.
I knew where its argument was going. If a heading was awkward, I knew what it was meant to introduce. If a passage felt hard to find, I could fill in the missing steps from memory. I could mistake familiarity for usability because I was both the writer and the person checking the page.
The shared web reader had answered a narrow question: could a reader and I open the same text from a link? The next question was less comfortable. Could the page carry a book when the book wasn’t mine, didn’t depend on my explanations, and asked to be read for longer than a quick check?
The first real test was a public-domain book.
That was a practical choice. It gave me a real long-form text to work with without making a commercial permission question the center of the first test. I was not trying to make a legal argument about public-domain works. I needed a book that could expose the reader’s ordinary needs before I had to solve the separate question of commercial access.
The book did not know my shortcuts
With my manuscript, I could navigate by memory. I knew which chapter contained the difficult paragraph, which section was still rough, and where a reader would probably pause. The page could be imperfect and still feel familiar.
A public-domain book removed that advantage. The text arrived with its own order and its own expectations. A reader should be able to recognize the title, understand where the book begins, move through chapters, stop, and return without knowing anything about the person who prepared the page.
This changed what I looked for. I was no longer asking whether the reader could reach one passage I had chosen. I was asking whether the book could handle its own basic introductions. A real text had to make sense before any comment or discussion.
With my manuscript, I could forgive an awkward transition because I knew it was temporary. I knew which paragraph was still being revised and which strange choice had a reason I had not written down yet. A reader opening the public-domain book had none of that background. If the first chapter asked for patience, the page had to make that patience possible. If a chapter title disappeared into the rest of the text, the reading experience had to absorb the cost.
That was the first uncomfortable benefit of using a book I did not control. It took away my excuses. I could no longer explain a rough edge by pointing to the work still ahead or by remembering what the passage was supposed to do. The page had to meet the reader as the book stood.
That sounds obvious until you shape a prototype around one manuscript. Then every shortcut has a reason. A familiar chapter can hide weak navigation. A short reading session can hide awkward typography. A writer who already knows the text can mistake the ability to find something for the reader’s ability to find it.
The book brought its own shape
An actual book is more than a column of paragraphs. It has a title page, an opening, chapters that establish a rhythm, sections that end before the reader is ready to stop, and passages that ask to be remembered later. Even a simple novel brings an internal structure that the page has to respect.
The public-domain text made those structures visible. The beginning could not feel like a random excerpt. Chapter boundaries needed to help, not interrupt. Longer paragraphs changed the page’s pace. A heading that looked fine beside a short passage could disappear inside a longer one.
I began to understand that navigation was not a set of controls placed around the book. It was part of how the reader understood the book. A table of contents, a chapter title, and a way to return to a known place were not decoration. They were small promises that the text would still be there when the reader came back.
The manuscript had taught me to care about the relationship between a passage and a response. The real book taught me to care about the relationship between a passage and everything around it: the chapter, the argument, the pause before it, and the reason a reader might continue.
Reading takes longer than a demo
A demo tests the first minute. Reading tests what happens after the first minute.
I could open the page, confirm that the words appeared, and feel that the reader worked. A longer book made that confidence harder to keep. I would read part of a chapter, close the page, return later, and look for the place where I had stopped. If the route back was unclear, I had to reorient before I could resume reading.
That small interruption mattered. The reader does not experience a book as a successful load followed by a series of interface decisions. They experience a stretch of attention. The page has to support the beginning of that attention, the interruption in the middle, and the quiet return.
Typography became part of the test in the same unglamorous way. Lines that were too dense made a long passage tiring. Headings that carried too little contrast made chapters run together. A page that looked acceptable at a glance could feel different after several pages of sustained reading.
I did not need a new philosophy of typography to notice this. I needed to stop checking the reader as if it were a screen and start checking it as a place where someone might spend an evening.
The page had to accept the book
The prototype had been shaped by what I wanted to prove. The public-domain book returned the pressure in the other direction: the page had to accept the book’s shape instead of asking the book to behave like the prototype.
That meant letting the text set the pace. Some chapters were short and invited a quick return. Others were long enough that a reader needed a clear sense of progress. Some passages made sense only after the reader had carried details from earlier pages. The reader surface did not have to explain all of that, but it could not make the structure harder to perceive.
It also made the difference between displaying text and hosting reading more concrete. Displaying text is a first success. Hosting reading means the book remains legible as a whole while the reader moves through particular parts of it.
The shared page had started as an answer to a handoff problem. Now the book was asking it to handle duration, sequence, and return. Those were not separate features to add one by one. They were conditions of the act I was trying to support.
A real book made the next problem visible
The public-domain test did not prove that every book or every reader would behave the same way. It gave me something more useful: a way to see which limitations belonged to my manuscript and which belonged to the reading environment.
Once the page could hold a real book, the original link started to matter differently. It was no longer only a route to a passage I wanted someone to inspect. It was an invitation to read the work as a work, then return to a particular sentence when a thought needed somewhere to land.
That was a gain, but it also created pressure. A note could begin at a sentence and still carry an idea about a chapter, a character, or the book as a whole. The more complete the reading experience became, the less convincing it was to assume that every useful response would stay the size of its starting point.
The sentence could hold a note. It could not always hold the conversation that followed.