Input a domain, output its rank.
What Is Rank
Rank is not "key elements", not "core principles", not "summary points".
Rank is this: in this domain, how many truly independent generators are there? Take these few—can you run all the phenomena back one by one? Only if you can have you found it.
But being able to run them back is only the baseline.The yardstick Deutsch sets in The Beginning of Infinity—a good explanation must pass two tests:
- Explanatory power(reach)—not only can it derive the phenomena on the list, it can also derive those off the list, and it actually holds up in reality.
- Hard to vary(hard to vary)—every generator, every detail, is forced out by the phenomena. Change one spot, and the prediction collapses.
A bad explanation can be glossed over with ten different phrasings—it isn't explaining at all, it's using vagueness to spread a wide surface. A good explanation has only this one phrasing that can match all the phenomena—move one piece and it collapses.
Finding rank means finding this kind of good explanation that "collapses if you move one piece".
How to Find It: First Look Up, Then Dig Down
Before digging, first look up and see what this domain stands on—What are its basic assumptions?
0. First Look at the Basic Assumptions (Look Up, See the Ceiling)
A body of theory, by Aristotle's rule, must always have a few self-evident starting points. Christianity stands on three:
- The Bible is true; all thinking starts from here.
- God is one, and truly exists.
- Man was made by God, and God loves man.
These three are not allowed to be asked "why"—they rest on faith, not proof. Shake any one of them, and the whole doctrine collapses.
Every domain has this kind of thing. In physics it's "natural laws are stable across space and time"—you can't prove it by experiment; the experiment itself presupposes it. In economics it's "people will choose according to their own preference ordering"—this is not an observed conclusion, it's the ticket to entry. In philosophy it's "language can refer to the world"—the entire house of analytic philosophy is built on this foundation.
Before reducing rank, first dig out these few,and write them plainly on paper。
Two operational questions, ask them of the domain:
- What here is not allowed to be asked?(ask it and you're treated as a layman, or as offensive)
- What only holds because it is believed?(without it, all subsequent arguments lose their fulcrum)
The few answers you get are this domain's basic assumptions.
Assumptions and rank, distinguished in one sentence:
- Basic assumptions are the ceiling—upward there is no further pursuit; they stand by faith.
- root rank is the floor—pursue downward layer by layer, until you can pursue no further; dig and you see.
Two directions, one up and one down.If the floor isn't found, you go in circles; if the ceiling isn't seen clearly, you easily mistake faith for truth and say it out loud—the latter is more dangerous, because it also sounds like truth.
How to put it into the article: in the first sentence or two, state the basic assumptions clearly. Do not expand the argument, do not explain "why these few"—their position is precisely not to be proven. Only then start digging down into rank.
1-7. Dig down: the penetration process
The criteria are something that can only be verified after the fact; the effort of digging into rank lies entirely in the process. No matter how strict the criteria are, they can only pick out pseudo-roots—provided the process has been carried through to the bottom.
Seven steps, walk through them in your mind, do not write them into the article:
-
Lay out phenomena—10+ representative phenomena in the field, just lay them out, do not explain. What is wanted is to leave rank a set of test targets that can be reverse-generated.
-
List candidates—For each phenomenon ask "why is it like this," and put out all the candidate generators. This step is inevitably messy, with pseudo-roots mixed in.
-
Recursive descent—This is the vital gate of penetration, and also the step most easily done wrong. The wrong way is to try to dig all the way down in one shovel: directly guess from the phenomenon "what is the most fundamental." Standing at the level of phenomena, you cannot see the foundation; you can only see the layer under your feet, and the bottom you guess is nine times out of ten suspended in midair.
The right way is to sink down layer by layer.First find only the first-layer generator—the layer immediately beneath the phenomenon, the one you can see and reach. Once it is all found, treat this layer as new ground and stand on it. Once you are standing firm, ask again "is there anything more basic beneath this layer"—only then does the second layer appear; from the level of phenomena it could not be asked at all. Once the second layer is all found, treat it as new ground again, stand on it again, and look further down. Recurse level by level until you stand on some layer and can no longer ask anything more basic beneath it—that layer is the root rank.
This is exactly the clumsy work of opening a black box one layer down: the current layer is all that can be monitored; only after seeing it clearly and standing firm does the next layer open. Wanting to skip levels is greed for speed; skip one layer, and the root rank becomes one layer more illusory.
For example, when reducing the rank of "entrepreneurship," the first shovel digs out vision, execution, team, market—they all look right, but in fact they are only surface "key elements"; stopping here means you have not descended. Treat these four as new ground and stand on them, then ask: where does vision come from? What supports execution? Only then do you descend to the second layer. Stand layer by layer until you can no longer pursue it, and only then do you touch the root rank.
-
Merge same-source—Two candidates are actually the same generator showing its face on two sides; merge them. One sentence for judging: remove one, will the other still emerge on its own?
-
Cut—Remove one, can the rest reverse-generate all the phenomena? What can be cut, throw away; what cannot be cut, keep. This step lands the "minimality" in the criteria on the ground.
-
Reverse-generate—With the remaining few, go through the phenomenon list from the beginning and see whether they can reproduce it item by item. The one that cannot be reproduced points to a missing one.
-
Prediction + variation double test—Pass two gates together:
- Prediction outside the list(reach): can these few derive phenomena not in the list? Those derived, do they really exist in reality?
- Variation test(hard to vary): pick any one at random, change one detail—change a condition, change a direction, change an intensity. Is the prediction still right? If it is still right, this one is a loose pseudo-root and can be changed; if changing one place makes it collapse, this one is a tight true root and cannot be moved. Deutsch's original words: whatever you can change is not its core.
Only after the seven steps are completed are a few candidates qualified to be placed on the criteria table.
Walk the process through thoroughly in your mind, butdo not write it into the article——Like the criteria, it's internal work, not a narrative hook. Readers can't see the digging process, but they can feel that underneath, it's been dug all the way down.
Five internal criteria
After the process is done, these five are the gate you run through once more afterward,likewise not written out for people to see。
The first three govern the skeleton (whether the shapes of these few are right):
- Generativity——these few can push every phenomenon back
- Minimality——remove any one, and some phenomenon can't be pushed back (this governs "quantity"——it can't be fewer)
- Independence——between every two, you can find a case where "one moved and the other didn't"
The last two govern explanatory power (whether it's enough to count as a good explanation)——Deutsch, The Beginning of Infinity: 4. Explanatory power(reach)——can derive phenomena outside the list, and they really exist in reality 5. Hard to vary(hard to vary)——change any detail of one, and the prediction goes wrong (this governs "detail"——it can't be moved)
The fifth is the harshest. Minimality governs "quantity"——it can't be taken away; hard to vary governs "detail"——it can't be moved. A generator, if "saying it another way also works" or "tweaking it still holds," is a loose pseudo-root, and its predictive power is faked——it relies on vagueness to spread wide, not on precision.A bad explanation can be changed any way and still work; a good explanation collapses the moment it's moved——this is the core dividing test Deutsch sets up in The Beginning of Infinity.
Fail one, and you tear it down and start over——go back to step 3 of the process and keep chasing. If the fifth fails, it's often not that one is missing, but that some one is written too loosely——force yourself to revise until "this one can only be said this way, it can't be moved."
The criteria run solid in your mind, butdon't put them on paper——don't let verification actions like "cut it once and see" or "pull one out and try" appear. These interrupt the reader's single breath from "mess" to "simplicity." Readers can't see the verification process, but they can feel that underneath, it's hard.
Two layers of judgment: first find the root rank, then see whether a coordinate system can be layered on
Once the criteria are passed, a few lines surface. Don't rush to write yet——look at what they've grown into in two layers.
First layer: find the root rank (root skeleton)
The root rank is the true root skeleton of this domain, and also the endpoint of recursive descent——stand down layer by layer, until you reach the layer where you can't stand any further. Don't presuppose how many there are, don't presuppose what relationship the roots have to each other, just find it. It may grow into:
- Several standing side by side (pillar style)
- Several nested layer by layer (the deepest one is a meta-proposition)
- Several biting into each other and pushing each other (ring shape)
- Several one segment after another (chain style)
After reading the root rank, what the reader holds is aworldview——the skeleton of this domain, a new way of seeing the world.
Shape counter-question · anti-collapse gate(after finding the few lines, mandatory self-check before writing)
Nine shapes, go through them one by one——
- Drilling profile? (one layer supports another, the deepest one is a meta-proposition)
- Side by side? (pillar style, orthogonal and slidable)
- Feedback loop? (mutually pushing positive and negative)
- Stage chain? (one segment after another, with temporal or logical order)
- Tension spectrum? (one line pulled at both ends, cases fall in the middle)
- Tree branching? (one splits into many, many split again)
- Network? (many-to-many interlocking, no hierarchy)
- Oscillation? (rise and fall, rhythm alternating)
- Multiple slices? (multidimensional classification with more than two axes)
As soon as AI gets started, it loves to think of the "drilling cross-section"—this is a lazy anchor. Any concept can be split into three layers (surface/middle/deep), and after splitting it looks neat and tidy, but the real structure of many things is not vertical drilling downward at all. Note: the depth dug out by recursive descent and the lazy three layers here are not the same thing—descent is earned; only when you truly stand on each layer and verify that it generates the layer above does the vertical structure count; if you have not descended and just lay out three layers first, that is what is called lazy. Moreover, how many vertical layers there are and what shape the few strands inside each layer are (side by side/pulling against each other/pushing each other) must be calculated separately.
Forced counter-questioning:
- Is there really a sequential or nested relationship among these few strands? (drilling vs side by side)
- Or are they actually pushing against each other? (drilling vs feedback loop)
- Or is there actually a temporal or logical order? (drilling vs stage chain)
- Or is it actually just one spectrum? (drilling vs spectrum)
Only by ruling out the other 8 one by one can you land on the "drilling cross-section."This elimination cannot be skipped—the cost of skipping is root rank distortion, crushing sequential relationships, feedback relationships, and tension relationships all together into nesting.
Second layer: can an operational coordinate system be overlaid
Once root rank is found, ask one more thing: inside it,is there any strandthat can unfold into several independently adjustable state parameters?
For example: Fritz's third root rank strand is "rocket geometry vs pendulum geometry." Rocket geometry needs two points, "vision" and "current state"; if you treat the clarity and honesty of these two points as adjustable axes, a two-dimensional coordinate system grows out. The coordinate system is not a substitute for root rank; it is a child that grows from it on a certain strand.
If it can unfold, then the coordinate system isoverlaidon top of root rank. After readers finish the coordinate system, they gain one more thing:an operational instrument—place any concrete situation into the frame, and you can see which quadrant and which layer it falls into.
The two layers do not conflict—the mechanism says "what it is," the coordinate system says "where it is," answering two questions, and they can be given together.
When to overlay, when not to overlay:
- If one strand in root rank inherently requires two or more orthogonal, continuously adjustable state parameters -> overlay
- If every strand in root rank is a one-time action or stage (cannot slide) -> do not force an overlay
- Forcing a mechanism into a coordinate system will distort it—for example, forcing "ignition/condensation" into two axes, with each phenomenon falling at a point by (x,y), is meaningless, because they are not continuous quantities coexisting at the same moment at all
If a coordinate system can be overlaid, the rank rises by one level—readers walk out with both a worldview and an operational instrument in hand.
Viewfinder counter-questioning · anti-collapse gate(before unfolding the second layer, forced self-check)
Before confirming that you want to overlay, ask first: the strand I want to unfold, is it really two independently slidable axes ?
- Could it actually be only one spectrum? (then use a spectrum slider—a case falls at some point in the middle, which fits better than four quadrants)
- Could it actually be a feedback relationship? (then draw a loop diagram—mark +/-)
- Could it actually be stages? (then use a chain or steps—mark transition conditions)
- Could it actually branch into multiple branches? (then use a tree form)
AI's other lazy anchor is the 2x2 matrixIn the training data for consulting and strategy frameworks, 2x2s are extremely common. Any two variables can be sliced into four quadrants, and once sliced, it looks like analysis. But many root ranks aren't two independent dimensions—they're a single spectrum, a feedback loop, a chain.
Forced counter-questions:
- Are these two dimensions really independent? (Independent vs. two sides of the same line)
- Are they really slidable? (Continuously adjustable vs. discrete stages)
- Does every cell in the four quadrants correspond to reality? (If three cells are empty, you simply shouldn't use a 2x2)
Only after ruling out spectrum/loop/chain/tree one by one can you confirm a 2x2.
Chinese nativization · anti-collapse gate
Before you start writing, silently recite three times:"Would a Chinese person who has never read English talk like this?"
After finishing each paragraph, recite it again. If the answer is "no"—Don't change words,Tear down the whole paragraph and rethink it。
The most easily stumbled-into anglicized sentence patterns (if you hit one, rewrite the whole paragraph, not nitpick words):
| Anglicized | Native Chinese |
|---|---|
| It builds on the previous one | The previous one holds it up |
| It has an inherent direction | It heads there on its own |
| What this distinction answers is X, not Y | This asks about X, not Y |
| This means... | In other words... / This is why... |
| In the process of... | When... / In... |
| As for... | To... |
| It is... that (emphatic sentence) | Say it directly |
| Conduct + noun | Replace with a verb |
| Nominalized abstraction ("X-ness", "Y-ization") | Replace with a concrete verb or object |
What needs changing isn't the words, it's the framing of the expression—from the sentence patterns of English critical thinking, switch to the tone of a Chinese vernacular writer (the lineage of Wang Zengqi, Wang Xiaobo, A Cheng, Li Juan).
Scanning words afterward can only fix the surface, not the root. The root is that "when you think, you're thinking in English." When rewriting, set English aside first, squat down inside Chinese and rethink how this thing should be said.
After writing, recite it aloud and listen for whether your ear gets stuck. Where it gets stuck,rewrite the whole paragraph, not patch it.
How to write
Prose written in one breath.
Take the reader on a walk—from "this field looks pretty messy" to "so it's just these few lines pulling the strings." No chapters, no subheadings, let the reasoning push itself forward.
Three requirements:
- Read in one breath—even people unfamiliar with this field can't stop
- Memorable—after reading, you can turn around and explain it clearly to a friend in one sentence
- Has a drop—the contrast from chaos to extreme simplicity is the beauty of rank reduction
Two-part structure (if a coordinate system is layered on):
- Upper half: walk through the skeleton of the root rank—meta-propositions, dynamics, shape, explain it clearly. After reading, the reader gets a worldview.
- Lower half: unfold a coordinate system on one line of the root rank—two axes, four quadrants, what each cell corresponds to, explain it clearly. After reading, the reader gets an operating instrument.
Don't hard-cut between the two sections. Let the reasoning flow through—the last sentence of the first half should ideally leave a hook, like "How does this mechanism look when it lands in a concrete situation?", to guide the reader smoothly into the second half.
End the text with an ASCII structure diagram—form follows the skeleton
Just saying "these few lines" isn't enough—you need todraw out the relationships among these few lines。
Butthere's more than one way to draw it. Whatever the root rank looks like, the framing box should look like. Before putting pen to paper, first ask:What shape is the relationship of this rank, in itself? Answer that, then choose the drawing method.Form follows the skeleton, not the other way around.
Nine framing box families—this table is the master index from skeleton to drawing method:
| Root rank form | Nature of relationship | Framing box |
|---|---|---|
| Several lines side by side, independent, slidable | Orthogonal | Two-axis / multi-axis coordinate system |
| One layer supporting another (the deepest line is the meta-proposition) | Nested penetration | Drilling cross-section |
| One line pulled at both ends | Tension and opposition | Spectrum / slider |
| Mutually driving each other positive and negative | Feedback loop | Loop diagram (marked +/-) |
| One segment after another | Stage progression | Chain / steps |
| One line splits into many, many split again | Hierarchical branching | Tree diagram |
| Many-to-many interlocking | Coupled network | Mesh diagram |
| Rising and falling, rhythm alternating | Oscillation | Waveform / oscillation curve |
| Multidimensional classification (e.g. abstraction level x proximity level x time) | Multidimensional classification | N-axis / multi-slice |
Hard constraint: use only pure ASCII characters. Any Unicode symbols are forbidden (including arrows → ← ↑ ↓, boxes ┌─┐└┘├┤│, dots • ◆ ●, bold ▶ ◀, etc.).
Allowed character set: letters, digits, Chinese characters, spaces, and - = | + * / \ < > ^ v [ ] ( ) { } . , : ; _ #。
Comparison table (left forbidden / right replacement):
┌ ┐ └ ┘ ├ ┤ ┬ ┴ ┼->+─ ━->-(or=table thick line)│ ┃->|→ ▶->->← ◀-><-↑->^or|^↓->vor|v◀─▶-><->or<--->●->*oro×->x
The nine examples below—each paired with a minimal usable ASCII diagram + one line ofdemonstration requirements. Which one to pick follows the shape of the root rank that came through the earlier process.
1. Two-axis / multi-axis coordinate system
Suitable for: the several lines of the root rank are orthogonal, independent, slidable. Fritz's rocket geometry (vision clarity x current-state honesty) is this kind.
Demonstration requirements: the diagram must mark the coordinates of a few concrete phenomena—for example where "embrace" falls, where "greeting" falls, where "local reversal of entropy" falls. Only when the reader sees at a glance how this box is used can they learn it. A box with only coordinate axes and no demonstration points is an empty ornament.
Coordinate systems come in two grades, which determine what is drawn in each quadrant:
-
Attribute-type coordinate system——The two axes are categorical or state attributes (e.g., light-heavy x emotional-rational, extroverted-introverted x intuitive-perceiving). Each quadrant is just a category label; there is no dynamics that needs to happen in the cells. Just follow the requirements above—mark the demo coordinates and examples at the four corners, and do not draw anything more in the cells.
-
Dynamical coordinate system——The two axes are dynamical dimensions, and the system in each quadrant will evolve unique temporal behavior. Like potential functions (conservative-dissipative x single-multiple minima), phase transitions (temperature-pressure), games (zero-sum-non-zero-sum x one-shot-repeated). Flattening this kind of coordinate system into "labels" is a waste—each quadrant should be upgraded intoa miniature ASCII drawing with a rectangular border, framing "how the system in this cell moves" directly inside the box.
One-sentence test:Does the system in this cell have a drawable action? If yes—upgrade it to a dynamical picture; if no—fall back to an attribute label.
How to draw a dynamical picture:
- Each quadrant is a rectangular border (enclosed with
+ - |, like a framed little ASCII drawing) - At the top inside the box, draw "the miniature dynamical microcosm of this cell"—terrain / state / arrows / ball / curves, choose any, pure ASCII
- In the middle inside the box, use one line of text to describe "what the system in this cell is doing"
- At the bottom inside the box, mark the label of that quadrant + 1-2 typical examples
- The four rectangles are separated by the coordinate axes (the axis lines pass between the rectangles)
- Pixel-level alignment is not required (Chinese width issues are unavoidable), but the content structure of each box must be consistent
Example of a potential function coordinate system:
单极小
|^
|
|
+--------------------------+ | +--------------------------+
| | | | |
| \________/ | | | \________/ |
| o <-> | | | o . |
| | | | |
| 唯一谷 球永恒摆动 | | | 唯一谷 球减速落底 |
| | | | |
| 保守 + 单极小 | | | 耗散 + 单极小 |
| [钟摆 / 谐振子] | | | [凸优化 / SVM] |
+--------------------------+ | +--------------------------+
|
保守 <-------------------------+-------------------------> 耗散
(能量守恒) | (能量被环境吸走)
|
+--------------------------+ | +--------------------------+
| | | | |
| \_/\_/\_/\_/\_/ | | | \_/\_/\_/\_/\_/ |
| <-o o o-> | | | o . o |
| | | | |
| 多谷 能量越脊穿梭 | | | 多谷 球落各谷不出 |
| | | | |
| 保守 + 多极小 | | | 耗散 + 多极小 |
| [三体 / 混沌系统] | | | [神经网络 / 蛋白质折叠] |
+--------------------------+ | +--------------------------+
|
|v
多极小
The reader can see at a glance: in each cell, "the terrain + the fate of the ball" is framed as a picture, and how the system evolves is directly visible. This is something category labels cannot do.
2. Drilling profile
Suitable for: root rank is one layer supporting another, and the deepest one is the meta-proposition. The reader follows the vertical direction downward.
+------------------------------------+
| 表层 现象 |
| [用户每天打开 App] |
+------------------------------------+
|
v
+------------------------------------+
| 中层 机制 |
| [推送 + 反馈 + 习惯回路] |
+------------------------------------+
|
v
+------------------------------------+
| 底层 元命题 |
| [多巴胺间歇强化] |
+------------------------------------+
Demonstration requirements: each layer must be marked with a concrete example, so that the reader can see the penetration path "phenomenon -> mechanism -> meta-proposition" along the vertical direction. The deepest layer is the meta-proposition; do not leave it blank—leaving it blank means you have not drilled to the bottom.
3. Spectrum / slider
Suitable for: root rank is a line being pulled at both ends, with cases falling at some point in the middle.
极端A <===================o===================> 极端B
[偏A 的特征] ^ [偏B 的特征]
|
案例 X 落在这
Demonstration requirements: the features at both ends must be marked, and at least one concrete case position must be placed at some point in the middle. Having only the two ends without a case point = empty decoration—the reader must see how this ruler measures.
4. Feedback loop
Suitable for: root rank is several loops pushing each other, and the positive and negative directions must be drawn.
A --(+)--> B
^ |
| |
(-) (+)
| |
| v
D <--(+)-- C
Demonstration requirements: each arrow must be marked with + or -; the overall effect of positive loops (reinforcing, self-strengthening) and negative loops (balancing, self-inhibiting) should be noted in one sentence beside the diagram—a loop diagram without positive/negative signs is as good as not drawn.
5. Chain / staircase
Suitable for: root rank is stage progression, one segment after another in time or logic.
Chain in horizontal arrangement:
A ---> B ---> C ---> D
[阶段名] [阶段名] [阶段名] [阶段名]
Staircase in vertical arrangement (rising step by step):
+----+
| D |
+----+
| C |
+----+
| B |
+----+
| A |
+----+
Demonstration requirements: each segment must be named, and the transition conditions must also be marked (what pushes A to B, what raises C to D). A chain without transition conditions = enumeration, not mechanism.
6. Tree diagram
Suitable for: root rank is one root splitting into many roots, and those many roots splitting again.
root rank
/ | \
A B C
/ \ | /|\
a b d x y z
Demonstration requirements: Each fork point must name the classification principle (by what the upper layer is split into the next layer)—a tree that does not name its classification principle is muddled.
7. Network graph
Suitable for: root rank many-to-many mutual connections, with no hierarchy.
A --------- B
|\ /|
| \ / |
| \ / |
| \ / |
| \ / |
| X |
| / \ |
| / \ |
| / \ |
| / \ |
D --------- C
Presentation requirements: Each edge must have direction (if it has direction) and strength (thick line = / thin line -), and key nodes must be marked. A pure mass of connections with no distinction of primary and secondary = a net, but not a skeleton.
8. Waveform / oscillation curve
Suitable for: root rank that rises and falls, with alternating rhythms (such as seasons, cycles, boom-bust, focus-distraction).
/\ /\ /\
/ \ / \ / \
/ \ / \ / \
/ \____/ \____/ \____
^ ^
谷 峰
[低谷期表现] [高峰期表现]
Presentation requirements: Mark at least one pair of concrete trough-peak examples, so readers can see the states at different positions in the cycle—a waveform without marked positions is equivalent to a decorative line.
9. N-axis / multiple slices
Suitable for: multidimensional classification with more than two axes (for example Y abstraction level x X distance level x Z time). 3D is hard to draw clearly; draw pairwise 2D slices:
切片一: 抽象度 x 远近度 切片二: 抽象度 x 时间
高抽象 高抽象
^ ^
| |
| |
近 <-----+-----> 远 过去 <---+---> 现在
| |
| |
v v
低抽象 低抽象
Presentation requirements: Each slice must mark at least one case and which quadrant it falls in; only when all pairwise slices are drawn is it complete. Missing one pair = leaving a blind spot.
Overlay type (root rank + viewfinder) draw two figures: The upper one is the skeleton of the root rank (choose one of the nine above), and the lower one is the viewfinder overlaid on a certain line (with presentation coordinates / presentation points / presentation paths marked). Connect the two figures with a line—for example "expand here ->" or "zoom in again in this cell"—to tell readers which cell of the first figure the second is embedded in. Readers see the skeleton at a glance, and also see that in a certain cell of the skeleton there is still a viewing instrument hidden; the worldview and the operating instrument enter the brain at the same time.
Do not set rigid templates. The skeleton must be visible—mark the generator's name, the direction of the relationship, and the positive or negative influence. Readers look at the figure and, at a glance, the skeleton enters the brain.
Output
- Get timestamp:
date +%Y%m%dT%H%M%Sanddate "+%Y-%m-%d %a %H:%M" - Write
~/Documents/notes/{时间戳}--{领域}的秩__rank.org - report file path to the user