How an AI agent can create a source-linked synthesis with disagreements preserved
This workflow gives an AI agent a defined job, a bounded set of records, and a result a person can review. The agent reads the relevant Readwise context, applies the rules in the prompt, and keeps the source behind every recommendation. It returns a proposed handoff rather than taking consequential actions on its own.
Can an AI agent create a source-linked synthesis with disagreements preserved?
Yes. Start with the scope, date range, decision rules, and fields that identify the right records. The agent can collect the evidence, compare states or sources, mark conflicts and missing data, and organize the result around the outcome above. A reviewer then checks the matches and judgment calls before approving messages, record updates, bookings, purchases, publishing, or other write actions. The guide below shows the records, boundaries, prompt, and handoff needed for this specific workflow.
Keep citations beside themes
A Readwise note-review agent starts with a defined source set and research question. Give it the books, articles, tags, or date range in scope and the citation fields the team requires. It can cluster related highlights, identify repeated concepts, and surface passages that disagree, while retaining author, title, location, and original note.
The agent should distinguish source language from the reader’s annotation and its own synthesis. If a short highlight lacks enough context, mark it for rereading rather than extending the author’s claim.
Example starter prompt
Review Readwise highlights from [sources, tags, or date range] to answer [research question].
Group related passages by theme. For every point, include source title, author, location or URL, the relevant highlight, and any reader note. Preserve conflicting evidence and flag highlights that need more context.
Do not invent quotations or merge different authors into one claim. Return a synthesis and source register.Read against the outliers
Check whether the largest cluster reflects repeated imports from one source rather than broad agreement. Give minority and contradictory passages their own space.
Questions this workflow answers
What do my saved highlights say about this topic, and where do the authors disagree?
The agent works from a defined reading set instead of searching every saved note for a familiar word. It records each source’s title, author, publication, location or URL, highlight, reader annotation, and import date where available. The research question determines relevance. A passage that mentions “trust” in a different domain does not belong merely because it shares the term.
Themes preserve the source structure. The agent can group passages that make the same claim, distinguish examples from arguments, and show when one author qualifies a point another states broadly. It counts sources as well as highlights, because twenty passages imported from one book do not represent twenty independent views. Contradictory and minority passages receive their own treatment rather than being edited into artificial consensus.
Short highlights need restraint. A sentence clipped from the middle of an argument may not support the conclusion the reader wants to draw. The agent flags it for rereading and records the missing chapter or surrounding page instead of extending the quotation. Personal notes are labeled separately from author language, and paraphrases never appear inside quotation marks.
The output includes a synthesis, source-linked claims, disagreements, unanswered questions, and a rereading list. A reviewer can inspect every important passage and correct a theme without reconstructing the search. The agent does not invent quotations or treat a reading note as settled policy; it prepares evidence for the next piece of writing, discussion, or research decision.
The final handoff should include open questions and a rereading list, not just polished themes. That keeps the next research step tied to the source material.
Disagreement should retain the terms each author uses. Two passages may appear to conflict because one discusses short-term motivation and the other measures long-term retention, or because they study different populations. The agent can place the quotes, source locations, definitions, and surrounding note side by side before describing the tension. When a highlight lacks enough context, it belongs on the rereading list with the chapter or page to inspect. That prevents a tidy theme from combining claims the sources never intended to answer together.