The ZettelFlow manifesto¶

Most plugins help you write. Almost none help you think.

ZettelFlow is not a note plugin. It is an engine that makes knowledge evolve.

Stop managing notes. Start managing knowledge.¶

Obsidian is a wonderful place to store notes. But a store is not a mind. Luhmann's insight was that the value of a slip-box is not in keeping information — it is in making ideas evolve and connect until they produce new knowledge.

Obsidian is the store. ZettelFlow is the engine that makes the knowledge inside it evolve.

Everything ZettelFlow does follows from one architectural idea:

ZettelFlow does not manage notes. It manages the processes that transform knowledge.

A Step used to mean "do this operation on a note." It now means "advance this piece of knowledge." Steps compose into workflows — programmable thinking — that walk an idea through its life:

CAPTURE → CLASSIFY → PROCESS → CONNECT → DEVELOP → REVIEW → CONSOLIDATE

Begin with what matters to you¶

A well-organized vault is a means, not the destination. The destination might be a question you want to understand, a decision you need to make, or an argument you want to develop. Exploration without a fixed goal matters too. The system serves your purpose; it does not assign you one.

You choose what deserves attention — not just which suggestions to accept.

This is a direction for the whole product, not a fifth pillar or another project manager. Start with your own idea, use what you already know, inspect what challenges it, and decide what to do next. Recommendations should explain their basis and scope. A highly connected note is not automatically the most important thing for you to work on.

Progress can be a sharper question, a useful connection, a changed position, or an honest "I do not have enough evidence yet." A supporting link does not settle a question; a dense graph does not make a claim true. Evidence found, a proposed interpretation, and a conclusion you endorse are different things. You decide when a response is sufficient for now, and you can reopen it.

Unfinished thinking deserves continuity, not a demand to finish. Preserve the purpose, the reasoning and the remaining uncertainty so returning is easier than starting over. Pausing, changing direction or deliberately stopping can be the right next move — none requires improving a score first.

Cognitive agency (the meta-principle)¶

The four pillars below say what ZettelFlow builds. This says why it exists.

Knowledge tools increasingly offer a tempting bargain: hand over the thinking, keep the output. The risk is not receiving help; it is letting the output stand in for your judgement. Delegating a mechanical task is strategic and good. Considering a suggested interpretation can be useful too. Cognitive surrender begins when a system's evaluation is treated as your conclusion without your own decision. That is a design failure ZettelFlow works against, not a label to attach to its users.

ZettelFlow removes mechanical work. It protects cognitive work.

This is not an anti-AI stance, and not a demand that you do everything by hand. Automating "gather the recorded objections in this scope" removes mechanical work. Showing a possible reading can help you examine them. Silently adopting that reading as yours crosses the line:

The boundary is explicit: mechanical gathering and structural projections can be automated; interpretive output reaches your notes only through your accept / modify / reject decision, whether the proposal came from AI or a heuristic. The verdict is recorded, and a proposal must remain distinguishable from your own authored or approved position. Automation cannot bypass that boundary.

Deliberate friction is one design tool: where judgement is genuinely at stake, invite your reading before revealing the system's. It is not a reflection tax on every click. Skipping a prompt is not a verdict; accepting a good suggestion is a legitimate decision. Remove operational effort, and test whether the remaining pause helps people reason instead of merely slowing them down.

                         COGNITIVE AGENCY
             keep the human intellectually in the loop
                                │
            ┌───────────────────┼───────────────────┐
            ↓                   ↓                   ↓
        Lifecycle             Graph             Discovery
            └───────────────────┼───────────────────┘
                                ↓
                             Health

Obsidian stores your thoughts. ZettelFlow connects them, challenges them and makes them evolve — it can propose interpretations, but it never adopts them as your conclusions without your decision.

The four pillars (the soul)¶

• 🌱 Knowledge Lifecycle — every idea has a state (fleeting → literature → permanent → developing → evergreen → archived). ZettelFlow knows the phase and proposes the next move.
• 🕸️ Semantic Knowledge Graph — relations that carry meaning, so you can ask "show me every idea that contradicts this" and follow reasoning chains — not just backlinks.
• 🔭 Knowledge Discovery — surface the unexpected: connections you didn't know you'd made, challenges to your ideas, the hubs your thinking clusters around, a book outline hiding in a folder.
• 🩺 Knowledge Health — measure balance and richness, not volume: maturity, knowledge debt, a weekly review, composition, a heatmap of the ideas you actually developed.

Honest signals, not a measure of your mind¶

Metrics are consequences, not inventions: they must have an inspectable definition grounded in recorded data. Their scope and limitations matter. No result from find related means no candidate was found by that method in that scope. No recorded contradiction does not mean no opposing view exists. Connectivity, maturity and debt describe aspects of the recorded system — not truth, source quality or the worth of an idea.

Signals describe what was recorded; they do not certify understanding. Absence of a record is not absence of thought.

A decision log can show which proposals you accepted, changed or rejected. It cannot tell whether you understood them, thought elsewhere, or should have disagreed more. Confidence is yours to express, not something inferred from links or clicks. "Has my understanding changed, or has my vault just grown?" is a question the tool helps you examine, not a score it can answer for you.

Health should invite a useful next move, not turn every gap into an obligation. More notes, longer streaks, more rejections or more time in the plugin are not ends in themselves. No daily quota, no guilt for leaving an idea alone: the measure of usefulness is progress you recognize toward what matters to you.

What we will not become¶

• "Obsidian + AI" — competing with a thousand tools. AI is one Action, never the product. The system works fully — every pillar, every dashboard — for a user who never enables AI.
• An answer machine. Producing a plausible conclusion is not the same as helping you reach one. ZettelFlow gathers, proposes and challenges — it does not treat its output as your answer without your verdict.
• A judge of your intelligence or a productivity scoreboard. We describe records and expose gaps; we do not rank people by their notes, clicks, streaks or agreement with a machine.
• "Another plugin to create Zettelkasten notes." That ceiling is too low.
• A generic manager that pleases everyone and marks no one.

Preserve people's work — reposition capabilities¶

The original repositioning promise made this evolution possible without a rewrite. Its durable meaning is preserve people's notes, systems and established workflows — not every redundant interface forever. Consolidate overlapping paths, retain compatible access when an entry point moves, and make any migration explicit. Subtraction must not erase someone's work or strand their workflow. The canvas wizard remains the engine's front end, not a discarded first product.

The engineering shape follows the philosophy in five layers — Knowledge Model → Workflow Engine → Knowledge State → Experience → Community Gallery — mapped feature-by-feature in the reposition map.

Design by subtraction¶

Repositioning is only half of the discipline. The other half is subtraction: the plugin is designed by removing, not accumulating. Power comes from fewer, deeper capabilities, not more shallow ones.

• When two features resemble each other, we don't ship both — we centralize and empower one. One strong home for each idea beats two half-overlapping ones the user has to tell apart. (This is why the ~12 views became four surfaces, the community became one static gallery, and the whole system consolidated around one engine rather than a pile of features.)
• Every addition must earn its place against the cost of complexity. A feature that makes the product harder to hold in your head is a net loss, even if it "works".
• Keep the thinking system uncluttered. Config, templates and other vault tooling aren't knowledge — you can exclude those paths so they leave the graph, the health checks, discovery and everything else. Less noise, sharper signal.

If it can be centralized, centralize it. If it can be removed, remove it. Stay minimal and comprehensible — that is the feature.

Subtraction makes depth available without making complexity compulsory. Minimalism here is not a lack of ambition; it is the ambition.

Useful from the first idea. Deeper as your practice grows.¶

The best tools meet you where you are. Our design target is a first useful advance on your own material in five minutes — a target to test with people, not a result to claim without evidence. One existing note, or one real idea captured in an empty vault, should be enough to begin. Do not require reorganizing a vault, adopting a methodology, learning Canvas or enabling AI before giving value. A sparse vault deserves an honest next question, not invented connections.

That is the on-ramp: remove the setup and gathering work, not the user's choice of purpose or conclusion. Systems and the existing gallery offer scaffolding when wanted, not an admission fee. A beginner is not someone who thinks less — only someone who needs less configuration and clearer support. Keyboard and mobile access belong to that first experience, not a later reward.

And there should always be another level. Compose your own workflows, program conditions and events, author and share systems, ask your own knowledge graph questions — the ceiling rises as far as your practice does. That is the summit: depth that serves your practice, not difficulty for its own sake.

Useful from the first idea. Deeper as your practice grows.

The test for every addition¶

Before building, ask:

1. Whose purpose does this serve? Name a useful knowledge transformation, not an engagement metric.
2. What stays the person's decision? Make proposals, evidence, uncertainty and verdicts distinct.
3. What does the system actually know? State the basis, scope and limits; do not infer a mind from a log.
4. What do we deepen or simplify? Reuse a capability and preserve people's work rather than add a rival surface.
5. How will we know it helped? Test an observable outcome and the burden of using it, with consent and no hidden telemetry.

The constitution makes the engineering boundaries enforceable. This manifesto sets the direction, not a claim that every journey described here already ships. The Knowledge with purpose epic turns the next step into a bounded proposal: purpose-led Cultivate and first value on your own material. Its requirements, validation and deferred opportunities live there; current capabilities live in the product documentation.

For the few who will love it¶

This is not a feature list; it is an identity. ZettelFlow is not trying to be something everyone likes a little. It is trying to be something a few people love — because it changes how they think. If "stop managing notes, start managing knowledge" made something click for you, you're one of them. Welcome.