Skip to content
Kioku

Incremental reading

Incremental reading is the practice of reading a Source a little at a time, interleaving it with other Sources and with reviews, promoting the parts worth keeping into Extracts and Items as you go.

It is the reading half of Kioku. The other half — spaced repetition — schedules the flashcards you make. This page explains why reading works this way, and how a passage travels from your screen into your long-term memory.

The problem with reading once

Section titled “The problem with reading once”

A long article read in one sitting is mostly forgotten by the next day. You finish, you feel informed, and a week later almost nothing remains. The act of reading is not the act of remembering.

The usual fixes are worse. Re-reading the whole piece is slow and feels productive without building durable memory. Highlighting everything that seemed important leaves you with a pile of marks you never revisit. And committing a long, complex Source to memory all at once is overwhelming: you cannot tell, in the moment, which sentences will matter in a month.

Incremental reading sidesteps the choice between reading once and grinding through everything. You read a manageable chunk now, decide later what was worth keeping, and let a scheduler decide when to come back.

Reading as a queue

Section titled “Reading as a queue”

In Kioku, reading is not a separate mode you switch into. A Topic — the Element that holds your reading material — is scheduled and surfaced the same way a flashcard is. Both flow through one Queue, and the review session interleaves them: a few Items to recall, then a Topic to read, then more Items.

This interleaving is the point. You are never trapped in one fifty-page article and never buried under a thousand isolated cards. New reading arrives gradually, spread across many short sessions, alongside the recall it eventually feeds.

When a Topic comes up, you do not grade it the way you grade a flashcard. You take a Topic action instead: read if you simply read some of it, extract if you pulled passages out, skip to bring it back sooner, or done to retire it once you have mined everything you want. Each action tells the scheduler what kind of attention the Topic still needs.

From highlight to Extract to Item

Section titled “From highlight to Extract to Item”

The heart of incremental reading is a pipeline that turns prose into recall, one decision at a time. It is a tree: a Topic can spawn Extracts, and an Extract — or the Topic directly — can spawn Items.

  • Highlight. While reading, you mark a span of text. A Highlight is a lightweight annotation; it commits you to nothing.
  • Extract. When a passage is worth keeping, you turn it into an Extract — a smaller reading unit carved out of the Topic and added back to your reading queue at a slightly lower priority. An Extract preserves the surrounding context, so the passage still makes sense on its own later. You can extract from an Extract, narrowing a long quotation down across several passes instead of all at once.
  • Item. When an Extract (or a Topic) contains something you want to recall on demand, you promote it into an Item — a flashcard. The common path is a Cloze deletion: you select the words to hide, and Kioku wraps them in {{c1::…}}. Each distinct cloze number becomes its own Card, so one passage can yield several flashcards.

The structure is what makes this durable. Each child remembers where it came from: an Extract and the Items below it carry a reference back to the original Source and a snippet of the parent passage. Months later, a flashcard that has gone fuzzy still links to the paragraph — and the article — it was born from.

You rarely do all of this in one pass. You read, extract a few passages, move on, and come back later to extract further or make cards. Distilling a Source from a full article down to a handful of flashcards is meant to take several visits, not one.

The Highlight, extract, and cloze guide covers the exact keys and steps for each surface — the static reader, the live webview, and the Reader.

How Kioku schedules re-reading

Section titled “How Kioku schedules re-reading”

Reading and recall flow through one queue, but they are scheduled differently, and the distinction matters.

Items use FSRS, the flashcard scheduler described in Spaced repetition. It models each Item’s memory and predicts when you are about to forget.

Topics and Extracts are scheduled differently — for the rhythm of reading rather than the precision of recall. Rather than modeling memory, Kioku keeps a growing wait between visits that responds to your Topic actions. Reading a Topic lengthens the wait before it returns; making extracts lengthens it more, on the assumption that you have mined that pass and need a longer break before the next. A skip does the opposite: it shrinks the wait to bring the Topic back sooner, for when you are not ready to engage but do not want to lose the thread. Done retires the Topic from the queue entirely while keeping every Extract and Item you made from it. A Topic with no due date set is treated as due now, so freshly imported reading shows up in your next session rather than waiting.

Keeping reading and recall on separate rhythms is deliberate. Forgetting a flashcard is a precise signal FSRS can act on. Reading is fuzzier — there is no single moment you “forget” an article — so it follows a cadence tuned to the rhythm of incremental reading instead.

Section titled “Related”