Welcome to Tinsight, where Tabbit shares long-form ideas worth your time. Our new series, “Rebuilding Individual Learning,” offers practical ways to rethink learning in the AI era. The complete essay is about 19,000 Chinese characters and will be published in six installments; this is Chapter 4. It argues that cross-disciplinary learning should begin with a mental model you have already earned through real experience—a “mirror” you can use to reveal the structure of a new domain.
Author: Dawei Geng
Source: https://offbook.press/essays/rebuilding-learning/
Chapter 4 · Modeling with AI (2): Cross-Disciplinary Learning Starts with the Mirrors You Already Have
The concept of "generalist" has been discussed repeatedly in recent years. Most discussions imply a path: you have to learn a lot of new fields (read introductory books in multiple fields, subscribe to various newsletters, and be exposed to various new concepts). The hidden premise of this path is that cross-disciplinary competence = knowing many domains.
This path fails most of the time.
The evidence for failure is concrete: People who "read a lot and cover a wide range of topics" are no significantly better at judging truth than people who "only go deep in their own field." They have more concepts in their head, but these concepts are not combined into real insights. Their "breadth" is an exhibition of knowledge, not judgment.
The failure of this path exposed a deep misunderstanding: the essence of cross-disciplinary capabilities is not "knowing many fields", but "being able to use existing depth to illuminate the structure of new fields."
This fix changes everything about being a generalist and learning new areas.
Why You Must Start with a Mirror You Already Have
Let me first illustrate this judgment with an extreme example.
For a person who has never understood any field in depth, if he reads a book about biological evolution and another about market competition, the most insight he can get is "Oh, they both talk about survival of the fittest." This insight is a textbook cliché and will not change any of his judgments.
A person who has been in market competition for fifteen years and has an in-depth model of this field will suddenly see this when reading about biological evolution:
Therefore, the concept of "moat" in the market actually corresponds to "niche specialization" in biology. This means that companies with deep moats and highly specialized species share the same vulnerability: they are particularly vulnerable to extinction when the environment changes drastically. Generalist companies, like generalist species (rats, cockroaches), are more resilient when the environment changes drastically.
This insight is his own. It was not in any book, not generated by any AI, it came from a field he was familiar with and was re-illuminated from the perspective of another field.
▍What’s the difference?
The first person has no depth model, so both sides only see the surface, and the similarity is only superficial. The second person has a deep model of one side, so he can use this model to query the other side, and what he queries is the real insight.
So cross-disciplinary isomorphism can only be recognized by a mind that already has a deep model. Without a deep model, you can only see superficial similarities; with a deep model in at least one field, you have a ruler for structural comparison.
This judgment has several important corollaries:
1. A generalist is not "knowing multiple fields", but "having the ability to quickly build a structure in any new field." The starting point of this ability is a mirror of reality, an area that you yourself have deeply understood.
2. The correct starting point for building cross-disciplinary capabilities is not "learning a new field" but "polishing your existing depth into a usable mirror." Most people have never done this. They have been doing it in the field for many years, but they have never systematically abstracted their experience into a usable structure.
3. “I don’t have any expertise, so I can’t build cross-disciplinary capabilities” is an illusion. Everyone who has done something seriously has the raw materials for a deep model: work, long-term hobbies, life practices, professional skills. The problem is not "nothing", it is "not organized into a usable mirror".
You already have a mirror
This paragraph is particularly important, because when most readers read this, they will default to "the field of depth must be something very professional", and then consciously exclude themselves. This default is wrong.
The only criterion in the depth field is: you have received real feedback on this matter, made real judgments, and suffered real consequences.
Under this criterion, a depth field can be anything:
A housewife who has been cooking seriously for ten years may have a deeper model in the field of cooking than most food bloggers who have read cookbooks, because she has ten years of real feedback data and knows when a recipe cannot be cooked according to the recipe, when violating the recipe is wrong, and the real interaction rules between different ingredients.
A father of three who has a wealth of in-depth observations on child psychology and relationships. He knows what kind of expressions can make children really listen, what kind of punishments will backfire in the long run, and how to deal with the dynamic balance between brothers and sisters.
A master who has been repairing cars for twenty years has a model of mechanical systems and fault diagnosis that far exceeds any book knowledge. He knows what might be happening to a car at a specific sound, the causal chain between different faults, when it's a minor problem and when it's a major problem
A person who has been in sales for ten years has a profound practical model in human nature, motivation, persuasion, and negotiation. He knows what the other party is telling the truth and what is an excuse, when to push and when to withdraw, and the true relationship between price and value.
The key is not to "appear professional" in this field, the key is that you have actually received realistic feedback in this field.
A person who has a low level of education and has never been to business school, but has run a small store seriously for twenty years, may have a far more in-depth model of "business" than a consultant who has an MBA but has never actually done business. Because the former has twenty years of real feedback data, the latter only has cases in textbooks.
The tragedy of the pre-AI era is that such people never knew how valuable the depth they possessed was. They call their experiences "common sense," "local methods," and "feelings," and never regard them as resources that can be systematized, abstracted, and used to understand the world.
The AI era gives such people a completely different starting point: you already have a mirror, you just don’t realize it.
Turn Experience into an Explicit Mirror
The first key thing that AI can do is to help you dig out the implicit experience in your mind and turn it into an explicit structure that can be used to interrogate other fields.
This process is not difficult, but most people have never done it.
Specific methods:
Step 1: Choose something that you have done for a long time, had real feedback, and suffered real consequences.
Work, long-term hobbies, life experience, and professional skills are all acceptable as long as they meet the two conditions of "real feedback + real consequences".
Step 2: Give the AI a specific description
Not an abstract description, but a description with concrete examples.
for example:
I have been in sales for fifteen years, and my main customers are small and medium-sized enterprise owners. I have observed several phenomena:
Phenomenon 1: When customers ask about the price during the first meeting, they usually don’t actually buy it. Customers who really know how to buy will ask other questions first: What problems can I solve for them? Have I done similar cases before? What is the implementation process like?
Phenomenon 2: Customers who say "We need to have a meeting to discuss" are almost always politely declined. Clients who really need a meeting to discuss will say specifically "I need to discuss it with X and Y, we have a regular meeting on Tuesday, I will bring it up at the meeting, and I will give you feedback on Wednesday."
Phenomenon 3: Price is rarely the real deciding factor. In all the failure cases I have done, although they said "it was too expensive" at the time, when I reviewed it, I found that the real reason was something else: they didn't trust me, they didn't trust the product, or they weren't actually the decision-makers.
Please help me organize these observations into a causal model of the B2B sales process. What is the underlying mechanism? What general patterns does my experience point to? Which ones might just be accidental?
Step 3: Let AI help you dig deep structures
AI will abstract mechanisms from your specific observations:
Your observations point to several underlying mechanisms:
1. Real demand vs superficial demand. The very act of a customer asking about the price does not mean that he is evaluating the value. It could be a social move (to show that I'm thinking seriously), or it could be a defensive move (an excuse for you to leave). Customers who really want to buy focus on "can this thing solve my problem?" This is the real need.
2. Concreteness vs abstraction as a signal of commitment. Customers who do buy use specific words (names of people, times, places) when describing their decision-making process. Customers who don’t know how to buy use abstract words (“we”, “we need to discuss”, “we’ll talk about it when we have time”). The degree of concreteness is a reliable proxy for commitment strength.
3. Price as reason vs. price as reason. Price is almost never the real reason for a decision, it's a socially acceptable reason for rejection. The real reasons for rejection (distrust, lack of decision-making power, low priority) are difficult to say. So 99% of the time when a customer says "it's too expensive" you should translate it into "I have other reasons not to buy it, but I don't want to discuss that reason."
4. Abstract these mechanisms: Your observation is actually about a more general phenomenon: when humans face social pressure, they will use acceptable superficial reasons to cover up the real reasons. This mechanism doesn’t just apply to B2B sales, it applies to any situation involving decision-making and social pressure.
Step 4: Check and correct
There may be something wrong with the abstraction of AI. You have to use your specific experience to verify: "Does this mechanism conform to my observations? Are there any counterexamples?"
After several rounds of back-and-forth, you will have an explicit domain model with the underlying mechanisms expressed in your own language.
This model is your first mirror.
Once established, it can be used to illuminate other fields: politics, education, medicine, investment, interpersonal relationships, any field that involves human beings making decisions under social pressure.
Use Your Mirror to Examine a New Domain
After having the first mirror, the way to enter a new field is completely different.
Traditional way: Start from scratch by reading introductory books, textbooks, and cases in this new field, and spend a few months to a few years establishing basic knowledge.
Use the mirror method: use your existing structures to interrogate new areas and quickly locate the deep mechanisms.
Specific methods:
Step 1: Identify the key mechanisms in your mirror
Pick out a few key mechanics from the depth areas you just sorted out. For example, in the sales example above: "real demand vs apparent demand", "concreteness as a signal of commitment", "price as a reason for rejection", "surface reasons mask the real reasons".
Step 2: Use these mechanisms to interrogate new areas
Let's say you want to understand the pharmaceutical industry. Instead of asking "who are the players in the pharmaceutical industry?" start with the mirror you already have.
An observation I’ve made in the sales world is that people use superficial reasons to mask the real reasons. Where does this mechanism appear in the pharmaceutical industry? For example, when a patient chooses a certain medicine, a doctor prescribes a certain prescription, or a hospital purchases a certain piece of equipment, how big is the gap between the superficial reasons and the real reasons for these decisions?
Step 3: AI provides corresponding mechanisms in new fields
In the pharmaceutical industry, this mechanism is very obvious, here are a few examples:
1. The doctor prescribes. The reason the doctor usually gives is "this medicine is the most effective for this disease." The real reasons may include: he is familiar with the drug (familiarity preference), he has recently interacted with the drug manufacturer more (relationship factors), the drug has evidence-based medical evidence (academic umbrella), and the hospital will receive rebates after prescribing the drug (profit factors). Moreover, these factors often exist at the same time, and he himself is not entirely sure which one is playing a dominant role.
2. The patient chooses the hospital/doctor. The reason patients usually give is "this hospital is good." The real reasons may be: recommendations from friends (social signals), ease of registration (convenience), having been here before (habit), and proximity to home (cost). The criterion of "good medical skills" cannot be truly evaluated by patients, so it is a superficial reason that covers up the real reasons that he is embarrassed to say.
3. Hospital procurement of equipment. The reason given by the purchaser is "the need for better diagnostic and treatment capabilities." Real reasons may include: the personal preferences of the department director, the public relations efforts of a certain equipment company, or the factional balance of the hospital leadership team. None of these real reasons can appear on the procurement report, so prima facie reasons must exist.
Step 4: Further questioning – similarities and differences
You will find that your sales mirror is partially true in the pharmaceutical industry (the underlying mechanisms of human behavior are the same), and partially not true (the pharmaceutical industry has its own unique structure: medical insurance, professional thresholds, and life-threatening situations that change decision-making dynamics).
These similarities and differences are insights:
The same parts allow you to see the underlying mechanisms across industries (how humans make decisions under social pressure)
The different parts allow you to understand what is truly domain-specific (uniqueness of medicine: professional asymmetry, separation of payers and consumers, life-threatening situations)
Both insights will increase your judgment.
non-linear accumulation
This method produces non-linear gains as the number of mirrors increases.
It's not that n mirrors allow you to do n-dimensional comparisons - that's just a linear increase. The real nonlinearity comes from pairwise combinations between mirrors: 1 pair for two mirrors, 3 for three mirrors, and 10 for five mirrors. Each pairing is an independent structural perspective, allowing you to see different mechanics in new areas. The number of combinations is n(n-1)/2, growing faster than the number of mirrors themselves.
What's more important - a certain mechanism in the new field may not be visible in one mirror, but it is clear in another mirror. When several mirrors illuminate a new area at the same time, the mechanism that can hit is far greater than the sum of the independent comparisons of each mirror.
Therefore, a true generalist does not "know ten areas", but has three to five clearly polished mirrors. The combination of these three or five mirrors allows him to quickly understand the structure of almost any new field.
And this accumulation process has a counter-intuitive quality: with each new mirror built, the existing mirrors become clearer. Because their application in new areas will expose their previously undiscovered textures.
Therefore, "generalist" is not a state reached after a certain level of accumulation. It is an accumulation process that adds value at every step. Even if you only have two mirrors, your cross-disciplinary judgment is far better than a person with ten shallow mirrors.
Avoid the Metaphor Trap
When looking in the mirror at a new field, there is a trap that must be avoided: degenerating "structural isomorphism" into "surface metaphor."
The metaphors are "Entrepreneurship is like mountain climbing", "Marriage is like a partnership", "Life is like a journey". These metaphors are useful in inspiring emotions but are almost worthless in understanding reality because they only capture superficial similarities.
The structural isomorphism is this: "The early explosive growth of a startup and the altitude sickness of mountain climbing share the same feedback structure: both are systems pushed to the limit without fully adapting to the environment, and both will produce irreversible damage at a certain critical point, and the only way to judge this critical point is to monitor a series of indirect signals rather than the final result, because the final result appears too late."
The accuracy of these two statements differs by an order of magnitude. Metaphors make you feel that you understand; structural isomorphism allows you to really make new judgments.
The key role of AI here is to help you descend from metaphor to structure.
You propose a metaphor and ask the AI to ask:
"Where does this metaphor hold true and where does it not hold true? Is the similarity superficial or structural? If it is structural, which causal mechanism is shared?"
AI will force you to refine from "A is like B" to "A specific mechanism of A is isomorphic to a specific mechanism of B." This compulsion itself is the improvement of thinking.
If You Don't Yet Have a Domain of Deep Expertise
Finally, let me talk about a situation - you may think that everything mentioned above is correct, but you really don't have an obvious depth area. You're too young, you haven't started working yet, or you're not deep enough in what you're doing.
At this time, the answer is not to "get a mirror", but to directly start doing something that can generate real feedback. The AI programming discussed in the next chapter is a starting point that almost anyone can start immediately, and the feedback is extremely clear. As you practice, your first mirror will grow naturally.
Mirrors are not found, they are made.
The previous two chapters talked about using AI to do two things: learn thinking tools and build cross-disciplinary depth. But both of these things are based on a key premise: there are parts that AI can do, and there are parts that only you can do. If this boundary is not clear, the entire method will slide towards failure unconsciously. The next chapter talks about this boundary.
Chapter 1 · Most people use AI to impair cognitive decoupling
Chapter 1 · Most people use AI to impair cognitive decoupling
Chapter 2 · What questions are worth asking?
Chapter 2 · What questions are worth asking?
Chapter 3 · Modeling with AI (1): Thinking Tools as Reviewers
Chapter 3 · Modeling with AI (1): Thinking Tools as Reviewers
Chapter 4 · Modeling with AI (2): Cross-Disciplinary Learning Starts with the Mirrors You Already Have
Chapter 5 · What AI can and cannot do
Chapter 6 · A starting point: AI programming
Chapter 7 · Closing Thoughts
