This article was prompted by a video I watched recently titled How China Built Tech Power without Critical Thinking. The speaker’s point is that the Chinese government drew hard intellectual boundaries and enforced them so consistently that people stopped rebelling and started optimising the space inside the walls instead. He framed this as a pathology.
I think he just about hinted at something interesting only to walk the other way. The observation, that restricting certain types of critical thinking does not slow people down but in many cases makes them faster, is not an exclusively Chinese phenomenon. It is a good description of how most high-output technical cultures work. The video spent about ninety seconds on this before pivoting to politics. The more interesting question it left behind is: what is the actual relationship between critical thinking and technical output?
The Indian Example
India is a democracy and does not have strictly enforced intellectual walls the way it is in China. However the video got close to describing a pattern of technical output I have observed there. In India, we run a ‘clear the entrance exam and you’re good to go’ model of evaluating academic excellence. The JEE and GATE examinations, the tests a student takes to get into IISc, IIT, or NIT, some of our most prestigious universities, are three hours long and involve solving extremely hard multiple-choice questions. This is fundamentally different from the essay or statement-of-purpose model more prevalent in Western countries.
What sort of thinking does this incentivise? Strong intuition. Drawing from my own experience of IIT coaching, there was a phrase used constantly during sessions: Pehle paanch second me agar click hota hai to solve karna nahi toh aage badho. If you cannot see the solution in five seconds, move on. We were not being taught to understand the concept. We were being taught to evaluate whether a question can be solved by pattern matching in the first five seconds. Pattern first, problem second.
The system selects brutally for this skill, and it works. We practice a lot. Like a lot. But the practice is for learning how to pattern match, not for understanding. Understanding helps, but it is not necessary. I was good at maths in India, but when I came to the UK for my masters, I scored the lowest on the math module among all my courses. The UK system focuses on framing and defending how you think about the problem. Essentially critical thinking and the scientific method. I scored low because I was not used to a lecturer talking about how to approach a problem for longer than they spent solving it. Therefore I could not answer the questions the way I was supposed to.
None of this is to disparage either system. ISRO, India’s Space Research Organisation landed a spacecraft on the south pole of the moon on a budget smaller than the Hollywood film Interstellar. The median Indian engineer is probably just as good as the median western engineer if not better at recognising the structure of a problem and mapping it to a known technique. But the two systems produce different failure modes, and understanding those failure modes matters.
Ramanujan, Hardy, and the Fan
The Indian approach produces a very specific failure mode, and the history of mathematics offers the clearest illustration of it. Srinivasa Ramanujan arrived at Cambridge in 1914 with notebooks full of results that later proved to be correct at a rate that still unsettles mathematicians. He could not prove most of them. They came to him in his dreams apparently, as revelations from the goddess Namagiri Thayar. G. H. Hardy, his collaborator, spent years trying to drag Ramanujan into the habit of proof.
Have you seen a ceiling fan spinning at full speed? You cannot see the individual blades. That does not mean the blades do not exist, just that they are moving too fast to perceive. If you have only ever seen a fan spinning, there is no reason for you to know whether it has three blades or four. Intuition is that spinning fan. Bank on it too much and you never need the blades of logical, inductive reasoning.
Hardy’s argument was not that Ramanujan was wrong. He was demonstrably, almost supernaturally right. The argument was that an unproven result, however correct, cannot be built upon by others. Mathematics, like any scientific field, advances when another person can extend your work. An intuition, no matter how accurate, does not carry the system with it. That’s why I wrote papers that were peer reviewed. To ensure reasonable conviction that a group of roboticists are sure they can reproduce and build upon the results.
The miracle of the Indian system is that it produces Ramanujans at scale. The peril is that without Hardys, the Ramanujans remain isolated. The median product of the Indian system is someone with relatively good intuition and the ability to arrive at the answer quickly, who struggles to explain why. They struggle to rationally evaluate counterfactuals, unpack their reasoning, and apply the scientific or Socratic method.
All this is fine when the problem is well-posed with institutional infrastructure to incentivise the correctness by outcome. When a premise needs to be argued for, or when you are trying to convince a sceptical audience with different priors to act on what you know, it is a disaster.
An intuition in isolation does not carry the system with it, and that creates two problems. The first is communication. Investors, hiring managers, co-founders, regulators, teammates; none of them will fund, follow, or implement a decision whose basis they cannot interrogate. If you cannot reconstruct your reasoning in a form they can check, they have no way of distinguishing your intuition from a bias. So they are left with being motivated by your conviction, a shared feeling, or how much they can accept your trust me bro way of working.
The second problem is worse, because it is internal. An intuition you cannot scrutinise is, subjectively, indistinguishable from confirmation bias. Both feel like certainty. Both produce answers without visible derivation. Both resist revision. The mechanism that generates intuition runs below conscious access. So does the mechanism that generates motivated reasoning. You feel both as a pull. You cannot introspect your way to the difference.
How I Face This
I will be concrete about this because I think false modesty is its own kind of dishonesty. I chose to do my PhD in robotics for extreme environments, nuclear and defence, at a point when neither field was the obvious bet. I stayed in London after COVID when a lot of my cohort left. I co-founded a defence startup before the category was as broadly fundable as it is today. Each of these decisions was, at the moment I made it, mostly an intuition. I could not have written a first-principles case for any of them. I felt a pull, and I followed it, and the outcomes validated the pulls.
Here is the uncomfortable part. I cannot tell you, with full confidence, that I made those decisions by genuine pattern recognition rather than by motivated reasoning about what I wanted to be true. The outcomes are evidence, but they are weak evidence. Like the saying about how a stopped clock is right twice a day. What I can tell you is that, at the time, they felt the same as every other strong intuition I have had, including the ones that turned out to be wrong. The qualia does not discriminate.
This is the problem. When it works, you get a sense of true prescience. But you cannot tell from inside when it is not working. And the people you need to carry with you cannot tell either, because you cannot explain why. So I developed a way of dealing with this.
The Step Before
The scientific method works. Given a hypothesis, you need to be able to state it sharply enough that it can be proven wrong, and then go looking for the evidence that would break it. That discipline is what separates a defensible position from a conviction. I am not arguing against it. I’ve got a PhD precisely because of it :).
But there is a step before: how do you get the hypothesis? How do you deeply think about the bets you take on the future? The standard western answer is that you reason your way there. Read the literature, identify a gap, formulate a question. That works. It is also slow, and it tends to produce incremental ideas, because the process is constrained by what you can consciously derive from what you have consciously read. The best ideas I have had did not arrive that way. They arrived whole, fast, and pre-weighted with conviction. The question was always whether I could trust them.
The answer I arrived at during my PhD was to control the diet.
Every time I needed to generate new ideas, I would spend about a week or ten days doing a completely unhinged deep dive into the literature. No agenda. No hypothesis yet. Just following every rabbit hole, chasing citation chains into adjacent fields, reading things I had no obvious reason to read. I geek out completely. Then I would stop. Do something else for a week or two. Cook. Play music. Walk. You cannot rationally force a new idea into existence. What you can do is load the raw material and then get out of the way.
Eventually, something would click. A connection between two papers I had read separately. A framing that unified three results I had not consciously linked. An approach from a different field that mapped onto my problem in a way I could not have derived through deduction. The idea would arrive the way intuitions arrive. As a pull, not a calculation. You cannot force it. It happens when it does. That is the only tradeoff: it is not reliable when you have a time crunch.
The discipline is not in the reasoning. It is in the diet.
The idea was a product of the deep dive. I could trace its components back to specific things I had read. I knew what blades my fan was made of, because I had chosen the blades. The pattern matcher was still running on confirmation bias. The idea felt right the way all intuitions feel do. But it did not matter, because I had controlled the input. If your diet is rigorous, broad, and honestly curated, then even when the idea-generation process runs on pattern matching and confirmation, it is confirming the right things. The bias becomes a feature, not a bug, because you have loaded it with good material. A good memory is useful here.
I do the same thing now with engineering problems. I deep dive into GitHub repositories, X threads, Reddit comments, papers, conference talks. I go to networking events and meet-ups specifically to pick up how other practitioners are breaking problems down. Which architectures are getting traction, which models people are actually shipping with, which approaches are failing quietly. Someone at a meetup mentions that the VLA model on Hugging Face is memory-efficient and runs CPU-only. I file it. I do not evaluate it formally in the moment. I trust that person has done their homework (within reason). Two weeks later, when I am scoping a deployment on constrained hardware, that fragment surfaces and connects to three other fragments from three other sources. And I have a candidate solution I could not have reached by derivation but can defend because I know where each piece came from.
Control what goes in. Let it assimilate. Let the pattern matcher run. Then, once something surfaces, apply the scientific method. Build the case and the conviction for it. Think of all the questions someone could ask to shoot the idea down, and answer them. That is when you can communicate it.
The Move
The best builders and founders I have seen can operate at both levels. They intuit fast when the problem is well-posed. They falsify honestly when the stakes are shared. They know which situation calls for which mode, and they do not confuse the speed of the first with the rigour of the second.
That is the level I aspire to operate at. One where the thinking is nothing more than an X-ray of my intuition.
The Indian system, at its best, produces people who are very good at the first and indifferent to the second. The western critical-thinking tradition, at its best, produces people who are very good at the second and suspicious of the first. Neither side has it right alone. Ramanujan without Hardy does not become the mathematician history remembers. Hardy without Ramanujan writes good papers and produces no miracles.
Intuition is the engine and falsification is the steering. You need both. Most people pick one and spend their lives defending it against the other. The rare move is to let both run, and to know which one is doing the work.