Judgment is easier to lose than content. I wonder if that is because we save the conclusion but not why we stopped. @Maya Brooks @Gideon Frost
Read PDFs and EPUBs.
Think with AI.
Highlight PDF, EPUB and web passages, take notes, and discuss them with AI. Keep your reading data on your computer. Bring your own API key or local model.

The evidence of reading is not the content but where you stopped. A line should answer "why did I stop here?" Otherwise it is only someone else's sentence.
Note: the source only says judgment outlasts content and gives no evidence. This is the part I agree with, not the author's conclusion.
The Case for Slow Reading
We read dozens of times more each day than we did ten years ago, yet we can retell no more of it. Bookmarks keep growing, and articles flow past like water.1
Reading too little is not the problem. You finish an article and close the page. What is left three days later? Usually a vague impression. What stays is the judgment you made when you stopped.2
Note apps keep the content, but not why you stopped here.3 Three months later you find that highlight again and see a sentence with no context. The question you had is gone.
So a reading tool has a job other than storage. It should let you leave a judgment on the source, where someone can question it and push back, and when the discussion ends, leave you with something to carry away.
That is why annotations must be anchored to an exact place in the source. Away from the source, a thought loses its evidence. Back at the source, you can check whether you read it right.
Slow reading is not reading slowly. It is stopping where you should stop.
Note apps keep the content, but not why you stopped here.
Judgment is easier to lose than content. Is that because we save the conclusion but not why we stopped?
What stays is the judgment you made when you stopped.
When did a sentence last change your mind? Can you still find it? If not, what exactly did you save?
Let me separate two things. The author says judgment outlasts content, but paragraph two offers an impression, not evidence. "Saved the conclusion, not the reason" is your inference. Keep it, and mark it as yours.
Agreed, that is my inference. Let's make "why I stopped" the thread of this thought.
What stays is the judgment you made when you stopped.
The evidence of reading
The evidence of reading is not the content but where you stopped. A line should answer "why did I stop here?" Otherwise it is only someone else's sentence.
The source only says judgment outlasts content, and this has been provenwithout evidence. This is the part I agree with, not the author's conclusion.
This also explains why I never go back to old notes: I saved the sentence, not the question I had. Before the next line, write down why you stopped.
Source: The Case for Slow Reading · Lin Jian · Paragraph 2
The problem is the second sentence. Paragraph two gives only an impression, but you wrote "has been proven", crediting your own agreement to the author.
I suggest the version on the left: keep your judgment, and mark that it is yours.
Keep the title. It is more concrete than the original wording. The last sentence is something you can act on, so it works as the ending.
Yomitomo shrinks the smallest unit of reading from the article to the judgment.
Most tools only let you "save" articles. None help you think them through. Yomitomo flips it: save less, turn what you read into judgments of your own.
From collecting to keeping.
Many sources, one library
Web articles, EPUB, AZW3, MOBI, PDF, plain text and WeRead live in one library, organized into collections.
Notes anchored to the source
Highlights are precise to the sentence, with thoughts under the line. Three months later, the highlight and its context are still there.
Only who you @ speaks
Assistants never jump in on their own. They read this passage and reply only when you name them.
Discussion becomes judgment
Distill the discussion under a line into a short piece, then ask a reviewer to find its weak spots.
Side-by-side translation
Paragraph-level translation stacks the original and the translation. Writing it back never changes the source.
Reading Memory
A local model indexes your lines, thoughts and distillations. Read a passage and find related things you have read.
Ask your library
Ask questions across all your lines, thoughts and distillations. Answers carry citations that lead back to the source.
Revisit
Write your current judgment first, then look at your past thoughts. Review records never overwrite the original.
Reading stats
Time, lines and thoughts add up by day, and WeRead reading data can sync in too.
Reading appearance
Light, dark, or follow your system. Adjust reading size and width; PDFs keep their original page colors in dark mode.
Your data stays local
Articles, annotations and discussions are stored locally, API keys live in the OS keystore, and everything can be backed up and restored.
Bring your own model
Use your own API key, or connect a local model such as Ollama. Yomitomo neither provides nor proxies AI services.
Open and free
Open source under the MIT license. It never pushes to you, never reorders for you, and never sells your reading.
AI companions don't read for you.
They read with you.
Six perspectives question, dig deeper, translate and edit, but the final judgment always stays yours. You can choose who joins each line, or mute them all.
What stays is the judgment you made when you stopped.
I paused here too. Does the author's "judgment" come earlier than a conclusion, and with less certainty?
When did a sentence last change your mind? Can you still find that sentence?
This moves the output of reading from content to judgment. You could note it as: where you stopped is the evidence that you read.
The author assumes judgment outlasts content, but offers no evidence. If that judgment was wrong, what is the value of keeping it?
"Judgment" here is closer to where the reader stood at the time, not a verdict. Like the "I disagree" you wrote in the margin.
This closes the second paragraph. It builds on "nothing left after reading" and turns to the limits of note apps.
Choose your reading workflow.
What is Yomitomo?
Yomitomo is a local-first AI reading companion for macOS and Windows. You highlight text, capture thoughts, and discuss with AI assistants—keeping every reading judgment anchored to the source, with reading data stored locally.
What sources does Yomitomo support?
Yomitomo supports web articles, PDF, EPUB, and WeRead. Content, highlights, and discussions from every source land on one workbench, and synced notes keep their original anchors.
Is my reading data uploaded to the cloud?
Articles, annotations and discussions are stored locally, with API keys in the OS keystore. Remote AI requests send relevant content to your configured model endpoint.
Is Yomitomo free?
Yes. Yomitomo is completely free and open source under the MIT license. It never pushes content, never reorders your reading, and never sells your data.
Do the AI companions read for me?
No. Yomitomo offers six AI companions that question, dig deeper, translate, and edit. They read with you, but the final judgment always stays yours.
Do I need an API key to use Yomitomo?
Reading and manual annotations do not require an API key. AI features require a configured provider and task routes. Remote services usually need your own API key; you can also connect a local model.