Computing Life

I often see discussions in coffee enthusiast groups like this: someone makes an amazing cup of pour-over coffee and immediately shares the complete brewing parameters – water temperature at 92.3°C, 15.2 grams of coffee grounds, a specific grind setting on a particular grinder, and precise timing and water volume for each pouring stage, down to the second and gram. Seeing such shares, many people's first reaction is: "Great, save it! Next time, I'll follow this recipe, and maybe I can replicate that divine taste."

But deep down, we all know it rarely turns out that way. Even with the same beans, and even if we meticulously replicate every parameter, the taste often varies wildly from person to person. This discrepancy is sometimes blamed on equipment differences, water quality, or an unstable brewing technique. However, I believe there's a deeper issue at play, one that might not just be about operational precision but a subtle misunderstanding of control and learning itself.

This phenomenon isn't unique to the coffee world. In many areas, we have a strong tendency to replicate successful formulas or templates.

For instance, high-achieving students share their study methods: wake up at 5 AM, eat at 7 AM, and work through specific mock exams. Many will immediately copy this schedule and behavior, hoping to get into their dream school, yet they often overlook that this top student likely has an internal adjustment mechanism. They dynamically adjust their study pace and focus based on their energy levels, mastery of topics, and even interest in different subjects.

The same applies to investing. When a prominent influencer recommends a "hot stock" or a quantitative strategy shows explosive short-term gains, many will unthinkingly "copy the homework," investing their hard-earned money. But when market trends shift or the strategy fails, they often don't know how to adjust their positions or manage risk.

Fitness is another example. People see fitness influencers share their workout plans and diet regimens and immediately try to follow them. However, because everyone's physique, basal metabolism, and lifestyle habits differ significantly, the results are often underwhelming.

Even in relationships, we might try to replicate advice from dating guides: send a morning greeting every day, apologize first after an argument. But if these actions aren't adjusted based on the other person's reactions, they can backfire and make the relationship more strained.

These examples make me increasingly aware that in our daily thinking and actions, we often unconsciously adopt what I call an open-loop control mindset.

Open-Loop vs. Closed-Loop Control

Open-loop control is a concept from control theory, contrasting with closed-loop control. It might sound academic, but it's quite easy to understand.

Let's use coffee again as an example. When you strictly follow preset parameters—water temperature, coffee-to-water ratio, grind size, time—it's like running a pre-programmed sequence. Once you hit "start," the entire process unfolds according to the set steps, without any adjustments based on real-time conditions (like the coffee bed collapsing, flow rate changes, or aroma shifts).

This mode is like a speaker reading a script on stage, delivering the prepared speech from beginning to end without observing the audience's reactions—whether they're engaged or already dozing off. Their task is simply to deliver the words. Or think of an old-fashioned music box: you wind it up, and it plays a fixed tune, entirely based on its internal mechanics. It won't play your favorite song more times or change the melody if you're tired of it. Its core characteristic is a one-way command, lacking feedback. Once initiated, it's locked into the preset process, stubbornly following a single path.

In contrast, there's another mode I prefer: closed-loop control.

Making coffee with a closed-loop mindset, you might start with a general range of parameters, but you rely more on real-time perception and adjustment during brewing. You'd carefully observe the water flowing through the coffee bed, note the changes in dry and wet aroma, and once the coffee is brewed, you'd taste its body and aftertaste. If you think, "This is a bit astringent, the body is thin, perhaps it's over-extracted," then for the next brew, you might coarsen the grind or slightly lower the water temperature. Then you taste again, evaluate again, adjust again... This process is you, continuously measuring and adjusting the brewing system based on the outcome.

The core of closed-loop control is a simple yet powerful cycle: Perceive → Compare → Correct → Execute:

1. Perceive: Your taste buds (or other senses, measuring tools) capture the actual result or state.
2. Compare: You compare the perceived actual result with your desired target, identifying the deviation.
3. Correct: Based on this deviation, you analyze the cause and decide what action to take to reduce the gap.
4. Execute: You implement this corrective action, then return to the first step, perceiving the new result.

"Closed-loop control" sounds technical, but this idea of perceive-correct, constantly adjusting, is hidden in many mundane corners of our daily lives; we're just so used to it that we barely notice. For example:

• Cruise control in a car: You set your speed to 100 km/h on the highway. When going uphill, the speed sensor immediately detects a tendency for the speed to drop, and the car's computer automatically increases the throttle to bring the speed back to your set 100. Going downhill, if the speed is about to exceed 100, it will appropriately reduce the throttle. It's not blindly maintaining a fixed throttle position but continuously perceiving the actual speed, comparing it to the set speed, and dynamically adjusting the throttle.
• Home air conditioning: You set the temperature to 26°C. When the room's temperature sensor detects the temperature is above 26°C, the AC starts cooling. When the room temperature drops near 26°C, it automatically reduces its cooling power or temporarily stops to avoid overcooling. It's also a typical perceive room temperature - compare with setting - adjust operation loop.
• Even our own bodies are closed-loop control systems: When it's cold, you involuntarily shiver, and goosebumps appear on your skin. This is the body generating more heat through rapid muscle contractions while reducing heat loss from the skin, trying to maintain core body temperature. When it's hot, you sweat profusely; sweat evaporation takes away excess body heat, helping to cool down. Our body's temperature regulation, blood sugar regulation, blood pressure regulation, and so on, are all extremely complex and efficient biological closed-loop systems.

These closed-loop control systems we use and rely on daily don't just execute a rigid command; they continuously adjust based on the actual situation, striving to help us achieve our desired state.

The Gifts of Closed-Loop Thinking: Enhanced Control and Insight

So, what tangible benefits can this adaptive, "sense-and-respond" closed-loop thinking bring us? In my view, it offers at least two core upgrades:

First, and most directly, it gives us true control—an ability to embrace change and respond flexibly.

We often seek "control," but many times, what we're pursuing is merely control over input parameters or execution processes. As mentioned earlier, strictly following a recipe's measurements and timing for cooking seems like controlling every step. But if the ingredients' water content changes or the stove's heat output varies, this control over the process doesn't guarantee control over the outcome (the dish's taste). Closed-loop thinking is the opposite. It acknowledges that the world is dynamic and that surprises and disturbances are normal. It doesn't demand perfect pre-setting of every detail. Instead, through continuous feedback and adjustment, it ensures the final outcome is as close as possible to our expectations, or helps us adapt to new goals when the environment changes. It's like equipping a ship with sensitive radar and an experienced helmsman; even if it encounters unexpected storms or reefs, it can detect them and adjust course in time, rather than rigidly sticking to the straight line drawn before departure. This ability to maintain stability amidst uncertainty is a deeper and more valuable form of control.

Second, it significantly promotes our deep learning and helps us see the essence of things.

Each cycle of perceive-compare-correct in a closed loop isn't just about correcting the current deviation; it's a valuable learning and exploration process in itself. When you try adjusting a variable and carefully observe its impact on the entire system—and how different that impact is from your expectations—you unknowingly begin to understand the system's internal workings and how different factors interact. Let's go back to coffee. By constantly adjusting grind size, water temperature, and pouring techniques, and by carefully tasting the flavor differences in each cup, you gradually build your own intuitive understanding of extraction. You'll learn that dark-roasted beans might require slightly lower water temperature or a faster flow rate to avoid bitterness; that beans with insufficient degassing time (high CO2 content) might need more thorough blooming or gentler pouring to prevent channeling. This process of "adjusting and understanding" transforms you from a novice who merely copies others' parameters into an expert who truly understands the principles of coffee extraction. This knowledge, gained through hands-on practice, repeated trial-and-error, and distilled from feedback, is far more vivid, profound, and practical than dry theories from books.

The Closed-Loop Approach to Learning

When we think about learning, the images that often come to mind are diligently listening to lectures, memorizing facts, accurately solving every problem, and ultimately achieving high scores on exams. This pursuit of mastering correct knowledge and perfect execution is undoubtedly a very important part of learning. But if we broaden our perspective on learning, especially when facing real-world challenges that lack standard answers and are full of uncertainty, I increasingly feel that the core, and more valuable, part of learning may not lie in how accurately we initially aim or how perfectly we execute. Instead, it lies in whether we possess a mechanism that can continuously calibrate our direction and ultimately lead us to our goal—this mechanism is the debugging and correction capability endowed by closed-loop thinking.

Let's first examine the learning model that pursues a "one-shot" success: the examples of coffee, cooking, stock trading, and fitness we discussed. It often implies an assumption of succeeding on the first try: if I make my plan perfect enough, if I memorize the knowledge points thoroughly enough, I should be able to solve the problem in one go and achieve my goal smoothly. Under this model, we might spend a lot of energy searching for optimal learning paths and standard templates, hoping that by precisely replicating these inputs, we can directly obtain the desired output.

This learning style is indeed efficient for structured problems with fixed answers. But its fragility is also obvious. Once reality deviates from the perfectly laid out plan, or when we encounter unprecedented situations, those accustomed to "one-shot" aiming easily feel lost, frustrated, and may even give up. Their toolkit might only contain tools for precise execution, lacking the weapons for dynamic adjustment.

Closed-loop learning, however, is entirely different. It doesn't expect to succeed overnight; it doesn't even care much if your initial attempt is way off target. In the philosophy of closed-loop thinking, where you start and how large the initial deviation is are not that important. What truly matters is whether you have the capability and methodology to continuously calibrate your direction and incrementally approach your goal. In this process, so-called errors and deviations are no longer frustrating signs of failure but become extremely valuable navigation signals. Every time you find yourself straying from the intended path, every time your attempt doesn't yield the expected result, it's as if the "fog of learning" is communicating with you: "This path might not be right; you might need to try adjusting over there." This itself is highly informative.

The true outcome of learning isn't just how many facts you've mastered or how many questions you've answered correctly. More importantly, it's whether, in the process, you've honed this ability to keenly capture navigational signals and effectively adjust your behavior accordingly. This ability to learn how to learn, to learn how to solve unknown problems, is the most core and lifelong asset.

Therefore, a crucial aspect of the learning process is a shift in mindset: from fearing mistakes to actively embracing controlled trial-and-error. View the learning process as a journey full of exploration and discovery, not an exam that only permits success and forbids failure. Every attempt, regardless of the outcome, is collecting valuable data and feedback for us. At the same time, we can use specific tools:

• Error Log (or "Mistake Journal"): Simply recording your mistakes and reviewing them regularly is immensely valuable. After all, every mistake is the system speaking to you. An error log based on closed-loop thinking can take this further. It should be more like a detailed debugging log, not only recording what went wrong but also digging deeper: How was I thinking about this problem at the time? Why did this error occur? How can I recognize such errors in the future? What corrective methods did I try? Which ones worked, and which didn't? Through such detailed reflection and recording of each debugging process, we can truly learn from our mistakes and avoid stumbling over the same stone repeatedly.
• Embrace Small Steps and Rapid Iteration: When learning a complex new skill or undertaking an exploratory project, don't expect to create a flawless, all-encompassing grand plan from the outset (what software engineering calls "waterfall development"). A better approach is to break down large goals into small, quickly verifiable sub-goals. Attempt each sub-goal, rapidly obtain feedback, and then adjust the next steps based on that feedback (what software engineering calls "agile development"). This approach of small steps and rapid iteration allows us to more efficiently approach the final goal through continuous trial-and-error and correction.

Ultimately, closed-loop learning emphasizes the process over a momentary result. The value of learning lies more in the dynamic process of continuously perceiving deviations, analyzing causes, attempting corrections, and gaining cognitive improvement from it, rather than merely stopping at an isolated correct answer or moment of success.

From Replication to Creation

Of course, having said so much about the benefits of closed-loop thinking, it's not a panacea for all problems. We must admit that practicing closed-loop control often requires us to invest more cognitive resources—to perceive, analyze, and adjust. It's not like those pre-packaged open-loop recipes that you can just pick up and use, following instructions effortlessly, which feels so convenient and energy-saving.

In fact, in many scenarios, efficient open-loop execution is still very valuable. For example, for simple tasks we already master, or in stable environments with highly optimized industrial production, open-loop control is often the optimal solution for efficiency and cost. Our brains also naturally prefer this energy-saving mode (similar to System 1 thinking mentioned in "Thinking, Fast and Slow"), which is a deep reason why simple, direct routines and experiences are so popular.

Therefore, what's truly important is perhaps not to simply judge whether open-loop or closed-loop is superior, but to understand their respective application boundaries and cultivate the ability to consciously choose and switch between thinking modes in different contexts. Ultimately, open-loop seeks stability; closed-loop refines precision.

Returning to our initial coffee discussion. We all aspire to replicate that stunning, divine taste. But perhaps the real magic isn't hidden in those seemingly perfect parameters, but rather in every moment you perceive, taste, think, and adjust in a closed loop. It's these dynamic calibrations that transform us from mere executors of parameters into creators of flavor.

You cannot replicate every perfect moment, but you can build a system that continuously approaches your ideal. And that is the freedom that closed-loop thinking brings.