Build Your Personal Assistant Operating System

4.1

Build a Reading Brain That Gives Ideas Back

Most saved articles become a graveyard of good intentions

You save an article about pricing strategy. You save a podcast episode about leadership transitions. You save a research paper someone recommended. Three weeks later, you cannot find any of them. Six months later, you have forgotten they existed.

The problem is not saving. The problem is that saved content has no second act. It enters a pile and stays there. This gives saved content a pipeline with four stages: capture, summarize, connect, and resurface.

A flat teaching diagram showing saved content moving through capture, summarize, connect, and resurface stages — Four stages turn saved content from a pile into a knowledge base that gives ideas back.

Reading Brain

Promise	A system that captures, connects, and resurfaces ideas from everything you read, so insights become usable rather than forgotten highlights.
Sources	Articles, books, PDFs, bookmarked links, highlighted passages, and your annotations. Reading history from .
Output	Structured reading notes with key ideas, connections to existing knowledge, open questions, and scheduled resurfacing prompts.
Saves to memory	Reading records with source, date, key ideas, themes, connections to other records, and resurfacing schedule.
Approval boundary	Reads and organizes freely. Suggests connections between readings and existing records. Never publishes, shares, or forwards reading notes without your approval.
Failure behavior	If a source is unreadable (paywalled, broken link, unsupported format), notes the failure and records whatever is available (title, author, URL).
Review criteria	Reading notes should capture the core ideas without substituting for the original. Connections to existing knowledge should be genuinely relevant, not superficial keyword matches.
Graduation trigger	When reading notes consistently capture ideas you find useful weeks later, move to automatic processing of bookmarked articles.
Privacy tier	Standard for published material. Notes containing personal reflections or private inherit a higher .

The reading inbox needs triage, just like email

The same classification pattern from the email applies to saved content. Not everything saved should be read. The system categorizes:

Content triage categories

Category	What belongs	What happens
Read today	Short, time-sensitive, directly relevant to something you are working on	Summarized and highlights extracted immediately
Deep read	Long, important, requires focused attention	Scheduled into a reading block on your calendar
Reference	Save for later lookup, do not read now	Summarized and tagged for search
Expired	Time-sensitive content that is now stale	Auto-archived after 30 days unread

Highlights become a searchable knowledge collection over months

When you read something and mark passages (in a reading app, a PDF, or by telling the assistant 'save this quote'), those highlights go into the tagged with source, date, and topic.

After a few months, you can query your own knowledge: 'What have I read about pricing strategy?' 'What is the best argument I have encountered for remote work?' 'Show me everything I saved about negotiation in the last six months.' Your reading becomes a resource you can retrieve, not a river that flows past.

Connection-making is the AI's highest-value contribution here. You save and highlight. The AI connects. 'This article about team dynamics is related to the book chapter on management you highlighted last month, and to the meeting notes from your one-on-one with Alex where you discussed team structure.' Those connections are what the AI sees and you would miss.

keeps ideas alive instead of letting them decay

Highlights fade from memory. The assistant resurfaces them at increasing intervals, starting the day after you save a passage and spacing out over the next month. Each resurfacing is a brief encounter with an idea you valued enough to save.

The concept comes from spaced-repetition research on memory retention. You do not need to study your highlights like flashcards. You just need to see them again at the right intervals so the ideas stay available when you need them.

The morning brief can include a resurfaced highlight section: 'Today's resurfaced highlight: a passage from something you read two weeks ago about stakeholder management, which is relevant because you have a meeting about project governance today.' Reading feeds the day's .

A flat teaching timeline showing the same highlight resurfacing on Day 1, Day 3, Day 7, and Day 30 — makes one saved idea reappear often enough to remain useful.

The synthesis-on-demand turns scattered highlights into a usable brief

The most powerful feature is synthesis on demand. You ask: 'Synthesize everything I have read about pricing strategy into a one-page brief.' The AI pulls all relevant highlights, identifies themes, surfaces contradictions, and produces a document you can use for a meeting, a presentation, or a blog post.

This turns months of reading into a working resource in minutes. The brief is not a summary of one article. It is a synthesis of everything you have read about a topic, drawn from your own curated highlights.

A flat teaching diagram showing multiple highlights compared into a theme, a tension, and a usable brief — Synthesis is useful when it shows both the shared theme and the tension across sources.

An input diet prevents the save-and-forget pile from growing back

Most people save more than they read, which creates guilt and noise. This has a design decision baked in: set a weekly reading budget. Five articles per week, for example. The assistant ranks saved items by relevance to current projects and auto-archives anything unread after 30 days.

The includes a reading section: 'This week you read four articles, saved seven, and archived three. Your reading focused on pricing and team management. Nothing saved this week connects to your current project on client onboarding. Consider searching for content about onboarding specifically.'

How this module breaks

What goes wrong	How you notice	What to correct once	What rule to save
The assistant's summary substitutes for reading the original, and you lose the nuance.	You reference an idea in a meeting and realize your understanding is shallow because you only read the summary, not the source.	Change the reading note format: lead with open questions and surprising claims rather than a comprehensive summary. The note should make you want to read the original, not replace it.	Reading notes lead with two or three questions the source raises and one claim that challenges existing assumptions. Full summaries are optional appendices, not the primary output.
The assistant connects every new reading to the same few topics because those topics have the most records.	Every reading note says 'this connects to productivity' or 'this relates to leadership' regardless of the actual content.	Add a diversity rule: connections must reference a specific existing record, not a general topic. If no specific record is genuinely relevant, the note says 'no strong connections to existing records.'	Connections reference specific existing records by name and date, not general topic labels. Absent a genuine connection, the note states that none were found.
Reading notes accumulate without resurfacing, becoming a graveyard of forgotten ideas.	You have 200 reading notes and never look at them. The system captures ideas but never returns them to your attention.	Add : each reading note gets a review date. The includes two or three reading notes due for resurfacing.	Each reading note gets a resurfacing date (default: 30 days after creation). The surfaces reading notes due for revisiting.

Your turn

Build your reading pipeline

You'll set up a reading pipeline with triage categories, highlight extraction, cross-referencing, and a resurfacing schedule.

Why this exercise matters

A reading pipeline turns saved content from a graveyard of good intentions into a knowledge base that gives ideas back when you need them.

You’ll leave with

A reading inbox with triage categories.
Highlights extracted from five real sources.
Cross-references between highlights that share a theme.
A resurfacing schedule for your most valuable highlights.
A weekly reading budget.

Use the prompt in order

1
Import your saved content and apply triage
Paste or connect your saved items. The assistant classifies each one: read today, deep read, reference, or expired. Review the classification and correct any mistakes.
2
Extract highlights from the items you have already read
For content you have already read, tell the assistant what stood out. It will extract highlights, tag them with source and topic, and store them in the .
3
Review the cross-references the assistant finds
The assistant looks for connections between your highlights and between your highlights and other modules (meeting notes, journal entries, tasks). Confirm or reject the connections.
4
Set your resurfacing schedule and reading budget
Choose how often highlights resurface (the default 1-3-7-30 day schedule works for most people). Set a weekly reading budget (how many new articles per week). Both can be adjusted after a month of use.

Prompt details

Each detail will be inferred, or you will be asked to clarify in the chat.

2/2 details

2 details will be inferred, or you will be asked to clarify in the chat

Starter prompt text

Open the full text if you want to check what will be copied.

Show starter prompt

Sort my recently saved articles and highlights into a prioritized reading queue.

Here are five items I have saved recently:
Infer from what you know about me or ask if it is unclear

For each item: 1. Mark it as: read today / deep read / reference / expired. 2. If I have already read it, pull out the key passages worth keeping. 3. Label each passage with its source, date, and topic. 4. Note any overlap between passages and between passages and my
   other modules (meetings, tasks, journal).


Show me saved passages again at 1, 3, 7, and 30 day intervals so the important ones resurface.

My weekly reading budget: Infer from what you know about me or ask if it is unclear articles per week. Archive anything unread after 30 days.

What the answer should give you

A reading pipeline with five items classified, highlights extracted from three that were already read, two cross-references identified ('this pricing article connects to the negotiation chapter you highlighted last month'), and a resurfacing schedule set to 1-3-7-30 days. Auto-archive rule active for unread items after 30 days.