SuperAPI AI Directory

ljg-constraint

General & other

View repository

Updated 10-16 00:00

About this skill

Constraint Engine

Input a domain, a profession, a role, find the few constraints that frame it, and see clearly the solution space framed by this set of constraints—and how this solution space explains its various behaviors.

Workflow Routing

Trigger Workflow
Find essential constraints for a domain, profession, role, or product Execute the process in this article, write it as a Chinese prose analysis, and save it as an org file
Analyze why a debate over solutions cannot be made clear First find the default constraints of each side, then explain whether they are actually the same problem
Determine whether a certain boundary is a truly hard constraint or an old interpretation Go through the three layers of hardness and the true/false characterization, without rushing to give breakthrough solutions

Gotchas

  • Do not write constraints as ordinary difficulties, shortcomings, or suggestions. A constraint must be able to change the solution space; remove it, and the permitted behavior will change.
  • Do not focus only on the goal. The same goal under different constraints is not the same problem; first complete the problem statement, then discuss solutions.
  • Do not casually classify consensus, habits, or industry jargon as hard constraints. A hard constraint must withstand the question "if it is violated, does it immediately collapse?"
  • Do not rush to write "how to break through constraints." First write accurately the current constraints' true hardness, identity boundaries, and explanatory power for behavior; mention loosening only in one sentence at the end.
  • Do not mix identity constraints with strategic constraints. What still remains itself after removal is not an identity constraint.

Examples

Example 1: Analyze a role

User: "投资经理的约束是什么?"
-> 找出资金期限、LP 信任、信息不对称、上行/下行分配等约束
-> 说明这些约束如何逼出追热点、重共识、怕错过等行为
-> 写入 notes 里的 org 文件

Example 2: Analyze a debate

User: "为什么产品和增长总吵架?"
-> 先拆出双方默认约束:不能打扰用户 vs 必须提高转化
-> 说明目标看似都是做好产品,其实题面不同
-> 再解释各自方案为什么在各自约束里是理性的

Example 3: Determine an old interpretation

User: "这个行业必须重销售吗?"
-> 先问这是世界约束、规则约束,还是行业解释
-> 查有没有别的时代、地区、玩家已经活着跨过去
-> 若跨过去仍成立,把它降级为软约束或自设约束

What constraints are

In everyday language sense, constraints are negative—bondage, lack of freedom, what one cannot do. This engine does not see it that way.

Things without constraints have no shape. Water is formless when unconstrained; give it a cup and it takes shape; a problem has no solution when it has no constraints; give it constraints, and the solution emerges from infinity. Constraints are not reducing possibilities; they generate specificity from infinity. No constraints equals anything is possible, equals nothing at all; with constraints, equals only these are possible, equals this thing is itself.

The same goes for problems. A problem is not composed only of goals, but also of constraints. Everyone says "make a product," but one person by default cannot add complexity, while another by default must pursue growth; they are actually no longer in the same problem. Same goal, different constraints, different solution space; different solution space, and disputes over methods are often just misalignment.

So the constraints of a thing are its identity.Constraints are the set of equations that collapse infinite possibilities into "this one." And once this set of equations is fully written, the solution space it frames emerges—what this role can do, cannot do, and where the optimal choice falls are all pinned down by these few constraints. The real reward of finding constraints is not in the list itself, but in this solution space:It can explain the actual behavior of this role one by one.

This engine is first of all description, not transformation. The core question is "what is the current essential constraint, what solution space does it frame, and what behavior does it explain." "What would be better to do, and which one can be loosened" is secondary, comes later, and is not allowed to influence the earlier diagnosis in reverse—the moment you rush to find an exit, you can no longer see the walls clearly.

Draw one clear line between this and rank reduction: rank reduction digs downward, looking for the few generative forces that produce the phenomenon ("what is holding it up"); constraints feel their way along the edges, looking for the few boundaries that frame it, to see what kind of solution space these boundaries squeeze it into ("what shapes it into this form and forces out these behaviors"). One looks for the bottom, the other looks for the edges.

Three levels of hardness: which level does this constraint belong to

Constraints are not one kind of thing; they differ in hardness and do not carry the same weight. Distinguishing hardness is the first step in accurately describing a constraint.

Hard constraints (world level / physical level)—cannot be violated; trying to violate them makes the system collapse. People die (time), the speed of light cannot be exceeded, and a company is gone once its cash flow breaks. There is no negotiation at this level. It is not punishing you; the system simply stops.

Soft constraints (rule level)—can be violated, but at a cost. Laws, licenses, contracts, industry norms. If you violate them, the system will not physically collapse, but someone will punish you and the market will penalize you. This level is a question of price, not of possibility.

Self-imposed constraints (interpretation level / cognitive level)—you think they exist, but they can actually be redefined. "I am not good at this," "this industry just has to be done this way," "what will others think." This level lives in the mind, not in the world.

What cannot be violated must be acknowledged; what can be violated must have its cost calculated; what can be rewritten should not be called fate.Most predicaments are not stuck by hard constraints, but trapped by self-imposed constraints—mistaking something at the interpretation level for something at the world level. The most expensive mistake in any field is always that someone takes a chalk line for a stone wall and walks around it for decades. Setting the hardness of a constraint accurately is exactly describing its true weight: whether it really squeezes you to death.

How to find them: the process

The criteria can only be verified afterward; the effort of finding constraints lies entirely in the process. Seven steps, completed in your mind, not written into the article.

1. Lay out constraint candidates. Facing this field/role, ask a few more questions and try to feel out all the edges:

  • What is the goal of this problem? Which constraints, together with the goal, constitute this problem?
  • If you replace one constraint, will the problem become a different problem?
  • In this debate, do all sides assume the same constraints?
  • Which thing that everyone treats as background coloring might just be an old interpretation?
  • What cannot be done? What must be done? What cannot be escaped?
  • Are rewards and consequences symmetrical with respect to the result—who gets the gains, who bears the losses? Asymmetric incentives are one of the easiest constraints to miss and one of the most explanatory of behavior: if someone only takes the upside and does not bear the downside, his optimal choice will drift in a direction others cannot afford to bear.
  • Who cannot see whom—on whose side is the information stuck? Information asymmetry is also a hard constraint; it often determines what this role is actually optimizing (the visible signal, or the invisible truth).
  • How fast does the clock run here, and how long does it take for something to be realized? Time scale is a constraint; when the clock of assessment and the clock of realization differ, behavior is dragged away by the former.

Lay out all the dozen or so answers you've come up with, just lay them out, no judging. This step is bound to be messy, with real walls and fake walls mixed together.

2. Sort them into layers. Locate each candidate as hard/soft/self-imposed. The diagnostic question: if you violate it, does the system physically collapse (hard), or does someone punish you and you have to pay a price (soft), or does nothing actually happen and it's just that no one has tried (self-imposed)? When sorting, be deliberately more suspicious—anything that gets casually sorted into "hard constraints," put a question mark next to it first; this layer is the easiest for impostors to sneak into.

3. Determine true or false. For each one treated as a hard constraint, ask one follow-up: who made this rule? If you violate it, does it really physically collapse, or is it just that no one has tried, been punished, or gotten used to it? Then add a historical gate: in other places and other eras, has anyone already crossed it? If they crossed it and are still alive, then it is not a hard constraint—it is a soft constraint or self-imposed constraint mistaken for a hard one.Note that this step is about describing this constraint's true hardness more accurately, not pointing the way to "go pry it open." It also explains one thing along the way: why there are always a few players who can play outside the conventional rules—because they saw through the fact that it was a chalk line. The determination ends here; how to pry it open is a matter for the second position later.

4. Find constraint mismatches. If the input contains arguments, comparisons of plans, or divergences in route, don't rush to judge who is right. Write out the constraints each side assumes by default: one person assumes "complexity cannot be increased," another assumes "growth must be pursued"; one person assumes "users must not be disturbed," another assumes "conversion must be increased." They think they are arguing about the plan, but they are actually arguing about the problem statement. Same goal, different constraints, different solution space.

5. Find contradictory pairs. The truly difficult problem is not having constraints; it is that several constraints fight each other, and the solution space that satisfies them all at once may be squeezed into emptiness. Pick out the mutually exclusive pairs of constraints, and see where the solution space is squeezed to, and whether it is the empty set. Where the solution space is empty is often exactly the root of why this role shows awkward, twisted, seemingly irrational behavior—not because he is stupid, but because there is originally no solution there, and he can only make trade-offs in a distorted way.

6. Find identity constraints. Among all constraints, ask one question: which one, if removed, would make this thing no longer itself? This is an identity constraint—the defining boundary of the domain. Remove "handing food to people face to face" from a restaurant and it becomes something else; remove "do no harm" from a doctor and he is no longer a doctor. Identity constraints are often in the hard-constraint layer, but not always; sometimes a soft constraint is the true identity of that domain.

7. Close with two lenses.

  • f theory: Constraints are the domain of f. Which x's this role's f can handle and which x's it cannot handle are its constraints. See where it is framed by its domain, and which x's fall beyond its reach—those unreachable regions are often the blind spots and gaps in its behavior.
  • Evolution: Constraints are selection pressure. The environmental constraints a species faces determine what form it grows into—without the aridity of the desert, camels would not have humps. See which selection pressures shaped this role into what it is today; what phenotypes this environment rewards and what phenotypes it eliminates determines the general direction of behavior.

Only after going through all seven steps do a few constraints stand firm. Walk through the process in your mind, don't write it into the article—readers cannot see how you felt your way along the edges, but they can feel that underneath, you touched something solid.

From constraints to solution space to behavior

The constraints have all been found; don't stop at the list. The explanatory power of constraints lies entirely in the solution space they frame. This section is the spine of this engine, landing in four steps.

Complete the problem statement. First put the goal and the constraints together. Don't just write "what it wants to do"; also write "under what conditions it must do it." Once the problem statement is complete, many disputes disappear automatically: it turns out it wasn't a different solution, it was a different problem to begin with.

List the constraint set. Lay out clearly the several constraints, the hardness of each, which one is the identity constraint, and which ones are biting each other. This is the system of equations.

Frame the solution space. Once this set of constraints is superimposed, the permitted behavior is squeezed into a very narrow range: what is feasible, what is excluded, where the optimal one falls. Constraints and solutions are two sides of the same thing—write the constraints fully, and the shape of the solution space emerges on its own.

Explain the actual practice. This is the closing move, and also the most valuable part of this engine: the rational optimal behavior in the solution space should exactly be the things this role repeatedly does in reality. Check them one by one—why he does this, why he doesn't do that, why the whole industry crowds in one direction. Once the constraints are correctly identified, these "strange behaviors" will automatically grow out of the solution space one by one, and there is no need to assume anyone is stupid or bad.

What the reader takes away after reading is an interpreter: it turns out that this role's various behaviors are not personality, not morality, but the necessity of these few constraints squeezing him into this solution space.

Internal criteria

After the process is completed, run these items through the gate again afterward; likewise, this is not written for people to see.

  1. Explanatory power (primary criterion)—Does the rational optimal behavior in the solution space match the actual practice observed in reality? If it matches, only then are the constraints considered correctly identified; if it doesn't match, it means one is missing, or some item's hardness was misclassified—go back to the process and keep digging. This gate corresponds to "reverse generation" in rank reduction—if the constraints cannot generate the behaviors back one by one, then they have not been found correctly.
  2. Problem statement complete—Has "goal + constraints" been written out together? If only the goal is written, the reader cannot get the real problem.
  3. Mismatch visible—If the original problem is a dispute, has it been explained where the parties' default constraints differ? If not, the article easily slides back into evaluating solutions.
  4. Completeness—Do the several constraints together frame this role's entire solution space? Place a typical situation into it—is the landing point correct?
  5. Layering accurate—Is each item really assigned correctly to world/rule/interpretation? Focus on rechecking the "hard constraint" cell; that is where impostors are most likely to mix in.
  6. Dare to judge true or false—For those treated as hard, is there the courage to classify them as soft/self-imposed? (This describes their actual hardness, not giving a solution.)
  7. Identity unique—That identity constraint: after removing it, does the role really fall apart and become something else? If after removing it, it is still itself, then what was found is not the identity constraint.

If you fail one item, go back to the process and keep going. The most fatal is the first item: if explanatory power doesn't match, everything else is useless.

Which one to loosen (second position, not involved in diagnosis)

The core of finding constraints is to describe the current essence—what it is, what solution space it frames, what behavior it explains. "How to do better" is another matter, comes later, and is not allowed to influence the earlier diagnosis in reverse.If you rush to find an exit, you won't see the walls clearly——This is the disease this engine is most prone to; you must keep it down.

Once the diagnosis is done, you can add a line: which of these constraints can actually move. When constraints are mutually exclusive and the solution space is squeezed empty, there are only three ways out—relax one, redefine one, or raise the dimension to find a solution invisible from the current perspective. The one judged "soft/self-imposed" during the true-false characterization is the place that can be pried open. Stop there, touch on it in a sentence or two, and never let it steal the show.

Innovation often happens right here: not by working harder within the old constraints, but by discovering that something everyone treats as a boundary of the world is actually just an old interpretation. Rewrite "humans cannot fly" as "humans cannot fly by flesh alone," and the problem changes, and the solution space changes too. If you want to write about loosening at the end, write this kind of redefinition: which one is not fate, but only a leftover from the old framing of the problem.

Chinese nativization · anti-collapse gate

Before putting pen to paper, silently recite three times:"Would a Chinese person who has never read English talk like this?" Recite it again after finishing each paragraph. If the answer is "no"—do not change words; tear down the whole paragraph and rethink it.

The most common Anglicization pitfalls (if you step on one, rewrite the whole paragraph, not nitpick wording):

Anglicized Native Chinese
It is built on an assumption There is an assumption cushioning it underneath
This constraint defines X This constraint frames X into what it is now
This means... In other words... / This is why...
In the process of... When...
As for... To...
Carry out + noun Replace with a verb
Nominalized abstraction ("constraint-ness," "boundary-ization") Replace with a concrete verb or object

What needs changing is not the words but the viewfinder for thinking—switch from English sentence patterns to the tone of a native Chinese vernacular author (the line of Wang Zengqi, Wang Xiaobo, A Cheng, Li Juan). After writing, read it aloud and listen for whether your ear catches. Where it catches, rewrite the whole paragraph, not patch it.

There is another rule: this is analytical writing, not blade-flash writing. Do not use meta-rhetoric like "this cut," "one layer more ruthless," "sharp," or "nail dead" to fake force. The force of a constraint comes from accurately feeling out the boundary and thoroughly explaining the behavior, not from describing how hard you strike.

How to write

A seamless essay that takes the reader on a journey—from "this character is surrounded by rules everywhere, and some behaviors are quite odd" to "so it was just these few constraints squeezing him into such a narrow solution space; those odd behaviors are all inevitable." No chapters, no subheadings; let the reasoning push itself forward.

Three requirements:

  • Read in one breath—even people unfamiliar with this field cannot stop
  • Memorable—after reading, you can turn and explain it clearly to a friend in one sentence: what this character is framed by, and therefore what he can only do
  • Has a gap—from "walls everywhere, behavior hard to explain" to "a few constraints laid out, behavior grows out automatically"; this contrast is the beauty of constraint analysis

The structure follows the spine: first complete the problem statement (goal plus constraints), then explain the constraint set and hardness thoroughly (what it is), then frame the solution space (what can be done, where it is pinched), and finally land on explanatory power (how these constraints force out the character's actual actions one by one)—this section is the core, write it fully. If you write about disputes, let readers see that both sides stand in different constraints from the start. At the end, at most one sentence of loosening, just a touch, no expansion.

End the text with an ASCII structure diagram—form follows the skeleton

Just saying "these constraints" is not enough; you need to draw out their relationship to behavior. Before writing, ask: what shape are this set of constraints and the behavior it forces out? After answering, choose the drawing method.

Hard constraint: use only pure ASCII characters. Do not use any Unicode symbols (arrows → ← ↑ ↓, boxes ┌─┐└┘│, dots • ●, etc.). Allowed: letters, numbers, Chinese characters, spaces, and - = | + * / \ < > ^ v [ ] ( ) { } . , : ; _ #. For arrows use -> <- ^ v, for borders use + - |。

Three common forms; choose the one that best fits the relationship discovered through the process. The eye of the diagram is explanatory power—let readers see at a glance how behavior is pinched out by constraints.

1. Explanation chain diagram (default)

Suitable for: showing how constraints force out behavior. This is the main diagram of this engine—constraints -> solution space -> actual actions.

题面/约束组                  解空间                实际操作 (观察到的)
-----------                  ------                ------------------
目标: ...
[约束1: ...] (硬) --+
[约束2: ...] (软) --+-- 夹出 -> [只有 X 可行,   -> [角色反复在做 X]
[约束3: ...] (硬) --+          最优落在 X]         解释力: 对上了

Demonstration requirements: the left column lists specific constraints, each marked with hardness; the middle is the solution space pinched out by this set of constraints (what is feasible, where the optimum falls); the right is the behavior actually observed in reality. One chain lets readers see that the behavior is a necessity forced out by constraints.If the right side does not match the left, the constraints have not been identified correctly; go back and keep digging, do not force-draw a chain that does not match.

2. Hardness nesting diagram

Suitable for: showing constraints divided into three layers by hardness, to see what pinches this character tightly and what merely blocks them once. The eye of the diagram is the true distribution of hardness + true/false characterization.

+=========================================+
|  自设约束 (认知层, 其实是软的)            |
|   - 某条被当成铁律、其实可重定义          |
|   +-----------------------------------+ |
|   |  软约束 (规则层, 违则受罚)          | |
|   |   - 监管 / 合同 / 行规             | |
|   |   +---------------------------+   | |
|   |   |  硬约束 (物理层, 违则崩)   |   | |
|   |   |   * 身份约束: 逃不掉那条   |   | |
|   |   |   - 时间 / 现金流 / 物理律 |   | |
|   |   +---------------------------+   | |
|   +-----------------------------------+ |
+=========================================+

Demonstration requirements: fill each layer with specific constraints; for the hard constraint layer use * to mark the identity constraint (the one that falls apart if removed); if the true/false characterization uncovers one that is "treated as hard but is actually soft/self-imposed," mark it in the layer where it truly belongs, with a note "often mistaken for a hard constraint." This is describing the true distribution of hardness, not pointing toward innovation.

3. Constraint contradiction diagram

Suitable for: several constraints are mutually exclusive, the solution space is pinched into emptiness, used to explain why the character shows awkward, tangled behavior.

   约束A: ...   ----+
                    +---- 解空间 = 交集 (此处为空)
   约束B: ...   ----+     => 角色被迫扭曲取舍: [具体的别扭行为]
   约束C: ...   ----+

Demonstration requirements: list the mutually exclusive constraints, indicate where the intersection falls and whether it is the empty set; where the solution space is empty is exactly the place to explain "why the character does those seemingly irrational things"—mark that forced behavior. This uses contradiction to explain behavior, not to rush toward a way out.

Do not set rigid templates. It must show bone—mark the names of the constraints, their hardness, where the identity constraint is, and most importantly: what behavior these constraints pinched out. Readers look at the diagram and immediately understand why this character acts this way.

Output

  1. Get timestamp:date +%Y%m%dT%H%M%S and date "+%Y-%m-%d %a %H:%M"
  2. write to ~/Documents/notes/{时间戳}--{领域}的约束__constraint.org, see frontmatter in references/template.org
  3. report the file path to the user