2Terminal-adjacent — the two models you kept

Same structure as 1a and 1b, recast near-black with hairline rules and monospace carrying the signal. Turn 1 stays below for comparison; the warm palette card in turn 0 is superseded by 2p.

2pThe revised language — near-black, hairlines, monospace carrying the signal.
TYPE
Questions and quotations stay serif
The only humane thing on the screen. Mono for a three-line question would be hostile at hour four.
Nav, labels, counts, provenance → IBM Plex Mono
p.4 · ann#12 · 09:41 · 94d · ⌘K · 14 untriaged
Plex Sans survives only in prose UI
Sentences the interface speaks. Everything enumerable moved to mono.
SURFACE & SIGNAL
well
#020713
page
#09111D
chrome
#121A25
rule
#2C333E
muted
#707885
open Q
#E5B64A
scout
#48DBA2
opposing
#F5688A
Four greys and three hues, nothing else. Amber is an open question, teal is the machine acting, rust is evidence against; a screen with no amber on it means nothing is waiting for you. Separation is 1px rules at #2C333E plus a 4-point step in surface value — never a shadow, never a card. Radii drop to 3px; rows tighten to 30–36px. Highlights on the page are the only warm fill in the app, so a marked passage reads as a physical mark rather than a UI state.
Dropped from the warm version: inner card shadows, the paper cream, the 6px radius. Kept: serif for anything a human wrote, and provenance always rendered as machine text.
2aSidebar — surfaces as places. Conventional, legible, teaches the model.
consolidation-vault
⌘K
SURFACES
Home
Question Inbox14
Reader
Research Questions9
Scout Queue6
Vault
THREADS
Replay & awake reactivation
TMR dose–response
Slow-wave density metrics
Methods · EEG artefacts
3 scouts idle · synced 4m
Question Inbox
14 untriaged · 3 today
All
Unlinked
Stale > 30d
TODAY
Does slow-wave density predict recall gain, or is it just a proxy for encoding strength at learning?
◆Klinzing 2019 · p.4 · highlight #1209:41
Why does the replay literature so rarely address awake replay in humans?
◆Ólafsdóttir 2018 · p.1109:12
Is targeted memory reactivation dose-dependent, or is one cue enough?
◇captured from note · Cueing methods08:30
EARLIER THIS WEEK
What counts as a "reactivation event" across the three labs that report it?
◆Schreiner 2021 · p.3Tue
Has anyone replicated the 2016 spindle-coupling effect with high-density EEG?
◇no source attachedMon
Is there a principled reason to prefer overnight over nap designs?
◆Lab meeting noteMon
TRIAGE
Does slow-wave density predict recall gain, or is it just a proxy for encoding strength at learning?
PROVENANCE
Captured 19 Sep, 09:41 while reading
Klinzing, Niethard & Born (2019), p.4
"…slow-wave activity has consistently been associated with overnight retention benefits…"
Promote to Research Question⏎
Attach to existing question…A
File as noteN
Answered — resolve with a sourceR
Dismiss⌫
NEARBY
2 open questions already cite Klinzing 2019 →
Scout swa-metrics proposed 3 papers on this →
· Nav is a place-list; surface = destination. Zero learning cost.
· Counts on Inbox / RQ / Scout are the only numerals in the chrome.
· Triage pane is keyboard-first — every action has a single key.
· Cost: Reader and Vault both want the full width and don't get it.
2bIcon rail + contextual pane — reading gets the window. Shown mid-capture.
Klinzing, Niethard & Born (2019) — Mechanisms of systems memory consolidation during sleep
⌘K
⌂
?
▤
◈
⌖
◻
⚙
OUTLINE
1 Introduction
2 Slow oscillations
2.1 SO–spindle coupling
2.2 Density measures
3 Replay
4 Open problems
QUESTIONS HERE4
Does slow-wave density predict recall gain…
p.4 · open
Which density metric is standard?
p.5 · resolved
Are nap designs comparable at all?
p.9 · open
37 highlights · 12 notes
2 Slow oscillations and the density hypothesis
Across three decades of work, slow-wave activity has consistently been associated with overnight retention benefits in declarative tasks. The magnitude of that association, however, varies by almost an order of magnitude between laboratories, and the reasons for this dispersion have not been systematically examined. Where authors report density rather than power, the definitions used are rarely stated explicitly.
A second difficulty concerns the direction of the inference. Participants who encode more material during learning also tend to exhibit greater subsequent slow-wave activity, which makes the overnight benefit difficult to attribute to consolidation as such rather than to prior encoding strength.
We return to this problem in Section 4, where the available dissociation designs are reviewed and their statistical power is assessed.
4
NEW QUESTIONfrom highlight · p.4
Is the between-lab dispersion an artefact of density definition, or a real moderator?
#methods
#swa
+ tag
Capture ⏎
Promote ⇧⏎
esc
ANNOTATIONS · P.4
#12 · 09:41 · linkable
"…varies by almost an order of magnitude between laboratories…"
→ 1 question captured here
#13 · 09:44
"prior encoding strength"
This is the confound Schreiner tries to design out. Cross-ref §4.
#14 · 09:52
"available dissociation designs are reviewed"
SCOUT · THIS SOURCE
Watching citations since Mar. 3 new proposals →
· Rail is 54px and iconic; the second pane changes meaning per surface.
· Capture popover is anchored to the highlight — provenance is automatic.
· Questions-in-this-source sit in the outline pane, not a separate trip.
· Cost: six icons is six things to learn; needs tooltips + ⌘K as the real nav.

Try next: "merge them — rail from 2b, question list from 2a" · "show Home and the Scout Queue in this language" · "one notch denser" · "now the iPad reading mode"

1Three navigation models for the main window

Same six surfaces, same visual language, three different theories of what the window is for. Each is drawn at 1180×780 — a modest laptop window, not a maximised one, because that is where a tool like this actually lives. Context, palette and the surface map are in turn 0 below.

1aSidebar — surfaces as places. Conventional, legible, teaches the model.
consolidation-vault
⌘K
SURFACES
Home
Question Inbox14
Reader
Research Questions9
Scout Queue6
Vault
THREADS
Replay & awake reactivation
TMR dose–response
Slow-wave density metrics
Methods · EEG artefacts
3 scouts idle · synced 4m
Question Inbox
14 untriaged · 3 today
All
Unlinked
Stale > 30d
TODAY
Does slow-wave density predict recall gain, or is it just a proxy for encoding strength at learning?
◆Klinzing 2019 · p.4 · highlight #1209:41
Why does the replay literature so rarely address awake replay in humans?
◆Ólafsdóttir 2018 · p.1109:12
Is targeted memory reactivation dose-dependent, or is one cue enough?
◇captured from note · Cueing methods08:30
EARLIER THIS WEEK
What counts as a "reactivation event" across the three labs that report it?
◆Schreiner 2021 · p.3Tue
Has anyone replicated the 2016 spindle-coupling effect with high-density EEG?
◇no source attachedMon
Is there a principled reason to prefer overnight over nap designs?
◆Lab meeting noteMon
TRIAGE
Does slow-wave density predict recall gain, or is it just a proxy for encoding strength at learning?
PROVENANCE
Captured 19 Sep, 09:41 while reading
Klinzing, Niethard & Born (2019), p.4
"…slow-wave activity has consistently been associated with overnight retention benefits…"
Promote to Research Question⏎
Attach to existing question…A
File as noteN
Answered — resolve with a sourceR
Dismiss⌫
NEARBY
2 open questions already cite Klinzing 2019 →
Scout swa-metrics proposed 3 papers on this →
· Nav is a place-list; surface = destination. Zero learning cost.
· Counts on Inbox / RQ / Scout are the only numerals in the chrome.
· Triage pane is keyboard-first — every action has a single key.
· Cost: Reader and Vault both want the full width and don't get it.
1bIcon rail + contextual pane — reading gets the window. Shown mid-capture.
Klinzing, Niethard & Born (2019) — Mechanisms of systems memory consolidation during sleep
⌘K
⌂
?
▤
◈
⌖
◻
⚙
OUTLINE
1 Introduction
2 Slow oscillations
2.1 SO–spindle coupling
2.2 Density measures
3 Replay
4 Open problems
QUESTIONS HERE4
Does slow-wave density predict recall gain…
p.4 · open
Which density metric is standard?
p.5 · resolved
Are nap designs comparable at all?
p.9 · open
37 highlights · 12 notes
2 Slow oscillations and the density hypothesis
Across three decades of work, slow-wave activity has consistently been associated with overnight retention benefits in declarative tasks. The magnitude of that association, however, varies by almost an order of magnitude between laboratories, and the reasons for this dispersion have not been systematically examined. Where authors report density rather than power, the definitions used are rarely stated explicitly.
A second difficulty concerns the direction of the inference. Participants who encode more material during learning also tend to exhibit greater subsequent slow-wave activity, which makes the overnight benefit difficult to attribute to consolidation as such rather than to prior encoding strength.
We return to this problem in Section 4, where the available dissociation designs are reviewed and their statistical power is assessed.
4
NEW QUESTIONfrom highlight · p.4
Is the between-lab dispersion an artefact of density definition, or a real moderator?
#methods
#swa
+ tag
Capture ⏎
Promote ⇧⏎
esc
ANNOTATIONS · P.4
#12 · 09:41 · linkable
"…varies by almost an order of magnitude between laboratories…"
→ 1 question captured here
#13 · 09:44
"prior encoding strength"
This is the confound Schreiner tries to design out. Cross-ref §4.
#14 · 09:52
"available dissociation designs are reviewed"
SCOUT · THIS SOURCE
Watching citations since Mar. 3 new proposals →
· Rail is 54px and iconic; the second pane changes meaning per surface.
· Capture popover is anchored to the highlight — provenance is automatic.
· Questions-in-this-source sit in the outline pane, not a separate trip.
· Cost: six icons is six things to learn; needs tooltips + ⌘K as the real nav.
1cQuestion column — the open questions are permanent furniture. Most opinionated.
consolidation-vault — 23 open questions
⌘K
Open questions
liveall
RESEARCH QUESTIONS · 9
Is the overnight retention benefit attributable to consolidation, or to encoding strength?
7 for3 against28d
Does TMR show a dose–response relationship?
4 for4 against6d
Is awake replay measurable non-invasively in humans?
2 for0 against61d
What would falsify the active systems account?
0 for0 againststale 94d
CAPTURES · 14 UNTRIAGED
◆Does slow-wave density predict recall gain…
◆Why so little on awake replay in humans?
◇Is TMR dose-dependent?
+ 11 more
⌂ Home◻ Vault⌖ 6
RESEARCH QUESTION
OPEN · 28d
Add sourceSet scoutResolve
Is the overnight retention benefit attributable to consolidation, or to encoding strength at learning?
WORKING ANSWER · edited 3d ago
Probably both, but the dissociation designs that could separate them are underpowered. The strongest case for consolidation-proper is TMR, because the cue is experimenter-controlled and orthogonal to encoding. Everything correlational is contaminated.
SUPPORTING · 7
Rasch & Born 2013
Review; TMR effects survive encoding controls. p.699 ann#4
Schreiner 2021
Cue-locked reactivation predicts gain within subject.
Diekelmann 2010
stub · no PDF
+ 4 more
OPPOSING · 3
Cordi & Rasch 2021
Meta-analysis; effect shrinks with preregistration. p.12 ann#9
Klinzing 2019
Concedes the confound is unresolved in §2.
+ 1 more
ORIGIN
Captured 22 Aug 2026, 16:07 from Rasch & Born 2013, p.699 — promoted 24 Aug.
FEEDER CAPTURES · 5
Does slow-wave density predict recall gain…
What counts as a reactivation event?
SCOUT
encoding-confound
Weekly · watches citations of Rasch 2013
3 proposals waiting →
LINKED NOTES · 4
Dissociation designs — power
Lab meeting 04 Sep
· The left column never leaves — open questions are the app's spine.
· For/against columns make a question's state readable in two seconds.
· Staleness is first-class: 94d unanswered is shown, not hidden.
· Cost: Reader must run in a second window or full-screen mode.

Try next: "take 1c's left column into 1b's rail" · "show me Home and the Scout Queue in 1a" · "darker, and one notch denser" · "now the iPad reading mode"

0Assumptions, surface map, visual language

You left both question forms empty, so these are my calls — argue with any of them. Everything above is built on them.

ASSUMPTIONS I MADE
Capture must cost nothing. A question you have to leave the page to record is a question you don't record. Every model above captures in place, with provenance attached automatically — never typed.
Promotion is the only ceremony. Capture is one keystroke; becoming a Research Question is a deliberate act with a form. That asymmetry is the whole product.
Time is a property of a question. Age and staleness are shown everywhere. A 94-day-old unanswered question is information, not clutter.
The Scout is a subordinate. It proposes into a queue and never writes to the vault. It gets one colour, one badge, and no notifications.
Keyboard-first, mouse-tolerant. ⌘K is the real navigation; the sidebar is the map you consult until you stop needing it.
HOW THE SIX SURFACES RELATE
3 · Reader
Source + annotations. Where wondering happens.
6 · Vault
Notes, links, graph. Where thinking is written down.
↓ capture↓ capture
2 · Question Inbox
Every capture lands here with its provenance. Triaged, never accumulated.
↓ promote
4 · Research Question
Working answer · supporting · opposing. The only place an answer is asserted.
↓ sets a watch
↑ rolls up to
5 · Scout Queue
Accept → Source stub → Reader. Reject → the scout learns.
1 · Home
Counts and what's stale. Read in five seconds, then left.
VISUAL LANGUAGE
Source Serif 4 — questions, quotations, working answers
Anything authored or quoted is set in serif. It reads like text, not like UI.
IBM Plex Sans — interface, labels, controls
450/500/600 only. No light weights; they die at 12px on a dense screen.
IBM Plex Mono — p.4 · ann#12 · 09:41 · 94d
All provenance and all counts. Machine facts look like machine facts.
paper
#FDFAF3
chrome
#F7F1E6
ink
#09111D
muted
#707885
open Q
#836202
scout
#105D41
opposing
#8F133E
highlight
#FEEBC4
Three hues, all at the same chroma and lightness, all load-bearing: amber means an open question, teal means the machine did this, rust means evidence against. Nothing else is coloured. Rows are 32–40px, the grid is 4px, radii are 5–6px, and there are no shadows inside the window — only hairlines at #B2BBC9. Warm paper rather than grey because you are looking at it for six hours.

Next steps I'd take: Home and Scout Queue as real screens · the promote-a-question form · the iPad reading mode · a dark variant for night sessions.