NoteRank — your notes, ranked by what matters now.
Most apps sort notes by date. NoteRank composes 14 personalized signals across 6 layers into one rank — the academic algorithm that ranked the web, ported to your second brain.
Move the dials. Watch the rank change.
Eight real-shaped notes. Four signal families. Drag any rail and the order recomposes — with a one-line reason for every position.
Top result: Q3 OKRs — draft v3. Balanced across all six layers — no extreme weight applied.
Free uses fixed weights across 4 basic signals. Pro unlocks all 14 signals with adjustable weights → See Pro
Why this rank?
Fourteen signals, six layers.
Each signal answers a different question about a note. Together they compose the rank — and every contribution stays auditable.
Family I Graph & Links
1 signal - Typed link centrality Personalized PageRank where link types carry different weight —
depends_on1.5×,references1.0×,related0.8×,mention0.5×,supersedes0.3×. A note pointed at by a high-ranking note inherits some of that standing — recursive over up to 5 power-iteration passes, by design.
Family II Knowledge Maturity
3 signals - Status × type maturity A completed plan with a positive outcome scores higher than a draft of the same age. Status, type, and outcome compose into a single maturity score (completed plan +0.12, draft −0.05).
- Confidence boost Notes flagged
metadata.confidence: highearn a lift — your own attestation that this knowledge has been verified or shipped. - Precedent value How often this note has surfaced as a precedent for a new decision. Notes that the Experience Envelope keeps citing earn a quiet promotion.
Family III Time
3 signals - Type-aware freshness Different note types fade at different rates — drafts in 7 days, active plans in 30, precedents in 180. Recency, without overweighting it.
- Staleness urgency A spaced-repetition signal — important notes that have not been touched in a long time generate an attention pull, surfaced for review.
- Completion momentum Past 70% checkbox progress, a note picks up a "nearly there" boost — the system nudges you toward what is closest to done.
Family IV Your Engagement
2 signals - Engagement velocity Views and edits over the past week, smoothed by an EWMA so a spike on Monday does not dominate Friday's rank. Notes you keep returning to compound their position.
- Interaction recency When did you last touch it? Distinct from creation date — a note edited yesterday is alive even if it was authored last year.
Family V Contextual Real-Time
3 signals - Contextual relevance Tag and link overlap with the past three days of activity. The note about auth ranks higher when you have been writing about auth all week.
- Priority weight Notes flagged
p0orp1earn a multiplier — important things stay surfaced even when the graph would otherwise quiet them down. - Dependency urgency Where this note sits in a
depends_onchain. A blocker on the critical path of an active plan rises before its date or activity would justify on their own.
Family VI Discovery
2 signals - Weak-tie bridge score Granovetter's signal — notes that connect otherwise distant clusters of your knowledge graph earn a serendipity bonus, so cross-domain ideas get a chance to be seen again.
- Exploration bonus Thompson Sampling on a Beta(α, β) posterior, plus a type diversity constraint that keeps at least three distinct note kinds in your top 20. Filter bubbles cost more than they pay.
Date-sorted notes are 1990s file managers —
and your second brain deserves better.
Google ranks the web. Gmail ranks your inbox. Spotify ranks tracks you have not heard yet. Every domain that mattered got organized.
Personal knowledge bases never did. Notion sorts by edit date. Obsidian by alpha. Mem.ai by chronology. You scroll, you pin, you remember where things are.
- The cost
- Information foraging research: ~70% of personal-knowledge users still browse — and the first 20 visible items decide what they think they know.
- The fix
- Stop showing edit order. Show what matters now.
From PageRank to your second brain.
NoteRank is not invented from nothing. Five algorithmic ancestors taught it how to rank.
- 1973Granovetter — Strength of Weak Ties Information flows between distant clusters through bridge connections — sociology that became serendipity engineering. Sociology · network theory
- 1998Brin & Page — PageRank A node's importance is recursively a function of who links to it. The ranking algorithm that made the web navigable. Stanford · Brin–Page paper
- 2010Gmail — Priority Inbox Per-user signal blending — engagement, recency, sender importance. Proved ranking can be personal, not editorial. Google · production system
- 2018Spotify — BaRT Thompson Sampling for music discovery — the explore–exploit balance that lets serendipity beat pure relevance. Spotify · ML production
- 2026Knovya NoteRank All four ancestors composed into 14 signals across 6 layers. Personalized, recursive, transparent — for the one place ranking was missing. ★ Knovya · production
First of its kind
Nobody else ranks notes.
Notion sorts by date. Obsidian by name. Mem by chronology. Roam tried link weight, alone. There is no second product that composes 14 personalized signals across 6 layers into one rank — that shows you why, and lets you tune the weights.
- Notion date · alpha · property
- Obsidian date · alpha · plugin
- Mem.ai chronological
- Roam Research link weight only
- Tana property sort
- ★ Knovya 14 signals · 6 layers · personalized
NoteRank touches everything you reach for.
Once it knows your work, it shapes every list, search result, and AI response.
Folder lists, recent panel, home feed — all ordered by NoteRank. The first 20 items become a curated entry into your knowledge.
Hybrid search fuses semantic + keyword via Reciprocal Rank Fusion — then NoteRank reorders for personal relevance, surfaced atop the result list.
When Claude or Cursor calls knovya_search, NoteRank
shapes which notes the agent sees first — without needing to learn
your structure.
When you draft a new plan, Knovya surfaces precedents grouped by outcome — and NoteRank decides which precedents actually warrant your attention.
NoteRank composes with the rest of Group II.
14 ★
Hybrid search BM25 + pgvector + Reciprocal Rank Fusion + fuzzy fallback. Four engines, native to one Postgres, in one search bar.
12 ★
Experience Envelope Personal Decision Intelligence — every note wrapped in 3 layers (precedents, evolution, evidence).
13 ★
Knowledge graph Bidirectional links, AI-suggested edges, bridge detection, cluster intelligence. The graph reads what you have written.
16 ★
Agentic memory MCP-native persistent memory for six AI agents — four recall modes (semantic, neighborhood, time-aware, memory health) over one shared workspace.
A few honest answers.
What is NoteRank?
A personalized ranking system for your notes — a 14-signal extension of Personalized PageRank, blended with engagement, maturity, recency, and serendipity. It surfaces what matters now, not what was edited last.
How is NoteRank different from PageRank?
PageRank ranks the public web by counting hyperlinks. NoteRank ranks your private knowledge base by composing 14 personalized signals — typed link centrality, type-aware freshness, engagement velocity, weak-tie bridge score, and ten more — into a single score, computed per-user, per-workspace.
Can I see why a note ranks high?
Yes. Hover any note in the browse view and Knovya shows the per-signal breakdown — which factor pushed it up, which pulled it down. Transparency is a default, not a setting.
How are the 14 signals weighted?
By default, six layers — Graph & Links, Knowledge Maturity, Time, Your Engagement, Contextual Real-Time, and Discovery — contribute equally. Pro users can adjust weights from Settings → Ranking. Free users use fixed defaults across four basic signals.
Is my ranking data private?
NoteRank is computed per workspace. No ranking data leaves your account. End-to-end encrypted notes are excluded from the global rank computation since they cannot be read on the server.
Does NoteRank work for new users?
Yes. With fewer than 20 notes, NoteRank gracefully falls back to a freshness-weighted default. As your graph grows, the link, engagement, and contextual signals begin to dominate. Cold start is handled — you do not need to teach it anything.
Can I disable specific signals?
On Pro, yes. Each signal weight has a slider that goes to zero. The most common adjustment is dialing serendipity (weak-tie bridge) up or down based on whether you are in focus mode or exploration mode.
Try the Lab in your own notes.
NoteRank Basic on day one — four signals, free forever. 14 signals and full weight control on Pro.
★ element 11 · Group II — Intelligence