READER — CALM STATE

Chrome at rest: a 54px rail, a hairline status strip, and the annotation margin. No toolbar over the page, no floating buttons. The narrow strip between page and margin is the connection gutter — one tick per anchored thing, positioned at the depth it refers to, so the page can tell you it is already wired into your vault without a single badge on the text.

Ólafsdóttir, Bush & Barry (2018) — The role of hippocampal replay in memory and planning
⌘K
⌂
?
▤
◈
⌖
◻
⚙
HIPPOCAMPAL REPLAY · CURRENT BIOLOGY 28 · R37–R50
3.2 Replay outside of sleep
Sequential reactivation is not confined to sleep. During quiet wakefulness, place-cell sequences recur at compressed timescales, and in rodents these waking events are at least as frequent as those observed in slow-wave sleep. Whether the same phenomenon is measurable in humans with non-invasive recordings remains an open methodological question, and the few attempts to date have relied on decoding approaches whose sensitivity is difficult to establish independently of the hypothesis being tested.
The functional significance of waking replay is likewise unsettled. Proposals range from immediate stabilisation of the just-encoded trace to prospective simulation of routes not yet taken, and these accounts are rarely distinguished experimentally. Where they have been contrasted, the designs have generally lacked the statistical power to separate them, a limitation acknowledged more often in discussion sections than in abstracts.
A further complication concerns the direction of replay. Forward sequences predominate before movement and reverse sequences after reward, which has been read as evidence for planning and for credit assignment respectively. Both readings are compatible with a more conservative account in which the sequence direction reflects the order of recent experience rather than any prospective computation.
We consider the methodological requirements for adjudicating between these positions in Section 5, where we argue that the critical evidence will come from tasks in which the animal's future trajectory is experimentally constrained.
11
MARGIN · PAGE 11 6 anchored
h·7→ question14:22
“…measurable in humans with non-invasive recordings…”
Would this hold for sparse inputs, or does decoding sensitivity collapse?
⧉ copy ref[[olafsdottir2018#h7]]
h·8ingested · 12 Sep
“…lacked the statistical power to separate them…”
Cite this in aim 2 — it is the cleanest admission of the power problem I have found.
linked from 2 notes
h·9last read · Mar
“…reverse sequences after reward…”
THIS SOURCE, ELSEWHERE
4 notes · 9 questions · 1 research question
⌘⇧C — open connections
page 11 / 14read 47 min today37 highlights · 12 notes
⌘' capture

SELECTION — THREE ACTIONS, NO DIALOG

The popover appears at the selection and nowhere else. Question is the default action and takes the Return key; the other two need a letter. Provenance is already written along the bottom of the popover before you type — source, page, highlight id, position — so capture is one keystroke to open, one to commit.

3.2 Replay outside of sleep
Sequential reactivation is not confined to sleep. During quiet wakefulness, place-cell sequences recur at compressed timescales, and in rodents these waking events are at least as frequent as those observed in slow-wave sleep. Whether the same phenomenon is measurable in humans with non-invasive recordings remains an open methodological question, and the few attempts to date have relied on decoding approaches whose sensitivity is difficult to establish.
The functional significance of waking replay is likewise unsettled. Proposals range from immediate stabilisation of the just-encoded trace to prospective simulation of routes not yet taken, and these accounts are rarely distinguished experimentally.
A further complication concerns the direction of replay. Forward sequences predominate before movement and reverse sequences after reward.
? Make a question provenance attached ⏎
✎ Make a note seeded from selection N
⇥ Link to existing note, question or source L
highlight only ⧉ ref
olafsdottir2018 · p.11 · ¶1 · chars 214–318 → will become h·7
AFTER ⏎ — WHAT GETS WRITTEN
questions/2026-09-19-1422.md
? Would this hold for sparse inputs,
  or does decoding sensitivity collapse?
source: olafsdottir2018
anchor: h·7 (p.11, ¶1, 214–318)
quote: “…measurable in humans with
  non-invasive recordings…”
captured: 2026-09-19T14:22
status: open
Everything below the question text was resolved from where the cursor already was. The only thing typed is the sentence itself — and you are returned to the page, at the same scroll position, the moment you commit.
ADDRESSABILITY
A highlight is an object with an id, not a mark on a file. [[olafsdottir2018#h7]] resolves anywhere in the vault and opens the page scrolled to the passage. ⧉ in the margin copies that reference; typing #h in any note autocompletes from the highlights of sources already cited there.

CONNECTIONS — WHAT THIS PAPER ALREADY TOUCHES

⌘⇧C replaces the margin with the tie-back panel. Nothing about it interrupts the page: the text does not reflow, no inline badges appear, and the connection gutter's ticks are the only trace while it is closed.

3.2 Replay outside of sleep
Sequential reactivation is not confined to sleep. During quiet wakefulness, place-cell sequences recur at compressed timescales, and in rodents these waking events are at least as frequent as those observed in slow-wave sleep. Whether the same phenomenon is measurable in humans with non-invasive recordings remains an open methodological question, and the few attempts to date have relied on decoding approaches whose sensitivity is difficult to establish.
The functional significance of waking replay is likewise unsettled. Proposals range from immediate stabilisation of the just-encoded trace to prospective simulation of routes not yet taken.
CONNECTIONS Ólafsdóttir 2018 esc
all 14 questions 9 notes 4 rq 1
■
Is awake replay measurable non-invasively in humans?
research question · anchored to h·7 · 2 for / 0 against
◆
Why does the replay literature so rarely address awake replay in humans?
question · p.11 · 3 weeks ago
◆
How would I detect replay without assuming the sequence in advance?
question · p.14 · 7 months ago
✎
Dissociation designs — power
note · cites h·8 · edited 3d ago
✎
Grant aim 2 — draft
note · cites h·8, h·12 · edited Aug
◈
Wamsley 2019 — Memory consolidation during waking rest
source · co-cited in 3 notes
WHILE CLOSED
The gutter shows one tick per anchored object at the depth it points to — amber for an open question, teal for a note, grey for a highlight with no links. That is the entire ambient display.

iPAD — FULL SCREEN, PENCIL AND TOUCH

1366×1024. The same app, not a reduced one: the margin becomes a swipe-in sheet, every control is at least 44pt, and the pencil palette sits under the hand rather than in a toolbar. Selection is a long-press or a pencil double-tap; the three actions become 56pt rows so they can be hit without looking. Capture from the pencil writes exactly the same file as ⌘'.

14:22Ólafsdóttir 2018 · p.11/14100%
3.2 Replay outside of sleep
Sequential reactivation is not confined to sleep. During quiet wakefulness, place-cell sequences recur at compressed timescales, and in rodents these waking events are at least as frequent as those observed in slow-wave sleep. Whether the same phenomenon is measurable in humans with non-invasive recordings remains an open methodological question, and the few attempts to date have relied on decoding approaches whose sensitivity is difficult to establish independently of the hypothesis being tested.
The functional significance of waking replay is likewise unsettled. Proposals range from immediate stabilisation of the just-encoded trace to prospective simulation of routes not yet taken, and these accounts are rarely distinguished experimentally. Where they have been contrasted, the designs have generally lacked the statistical power to separate them.
A further complication concerns the direction of replay. Forward sequences predominate before movement and reverse sequences after reward, which has been read as evidence for planning and for credit assignment respectively.
? Make a question provenance attached
✎ Make a note
⇥ Link to existing
olafsdottir2018 · p.11 · ¶1 → h·7
✎
⌫
‹
›
⌕
? capture
◫ margin

INGEST REVIEW — A PDF COMES BACK FROM THE TABLET

Phase 1: the file is annotated in Preview on iPadOS and syncs back through the vault folder. The app re-derives every annotation's identity from its quoted text and geometry, which can fail — so the review is organised by confidence, not by page. Nothing is written until the bottom row is empty.

Klinzing, Niethard & Born (2019) returned from iPad
19 Sep 14:48 · 22 annotations read · 14 new · 6 unchanged · {{ pendingLabel }}
discard all
{{ commitLabel }}
MATCHED CLEANLY · 9 NEW HIGHLIGHTS no decision needed
{{ c.id }}
{{ c.quote }}
{{ c.meta }}
{{ c.conf }}
? BECAME QUESTIONS · 3 note text began with Q:
{{ q.id }}
{{ q.text }}
{{ q.quote }}
{{ q.meta }}
→ inbox
! COULD NOT RE-MATCH · {{ unmatchedCount }} decide before committing
{{ u.id }} {{ u.reason }} {{ u.state }}
WAS · {{ u.wasId }}
{{ u.was }}
{{ u.wasLinks }}
CAME BACK AS
{{ u.now }}
{{ u.nowMeta }}
{{ a.label }}
WHY THIS SCREEN EXISTS
Identity is reconstructed from quoted text plus geometry. Reflow, a re-exported file, or an extended selection can break the match. If the app guessed silently, [[klinzing2019#h12]] in four notes would quietly start pointing at nothing. So a failed match is never resolved by the app — it waits here, and the commit stays disabled until it does not.

THE DESK — TABLET READS, MAC ANSWERS

How the two halves are actually used at the same time, and where a question goes between them.

iPAD · READING, 3 HOURS
14:22p.11/14
? capture
Pencil in hand, app full-screen, no keyboard. Phase 1: Preview on iPadOS writing into the synced folder. Phase 2: this same view in the browser.
1Pencil double-tap on a passage → three actions → Make a question. Written into the vault folder as Markdown with the anchor.
2Folder syncs. In phase 1 the PDF also carries Preview's own annotations back.
3Mac ingests: clean matches land silently, Q: notes become Questions, failures wait in the review.
4The question appears in the Inbox with the page you were on — and the Research Question you are writing on the Mac picks it up as a feeder capture.
MAC · WRITING THE ANSWER
Is awake replay measurable non-invasively in humans?
FOR 2
AGAINST 0
FEEDER CAPTURES · 5
Would this hold for sparse inputs…
arrived from iPad · 6 min ago
The Research Question does not need you to file anything: a capture anchored to a source it already cites shows up as a feeder the moment it lands.
iPad · Reader ──── vault folder ────▸ Mac · ingest review ──── clean matches + Q: notes ────▸ Question Inbox ──── promote ────▸ Research Question the only thing that crosses the gap is a Markdown file