Dev Log 3: Why Can't Skills Be Planned? How Do Skills Actually Grow?

Yesterday I was still using "imagination" to try to bridge the known present and the unknown goal, but after just one night, when I saw the book "Why Greatness Cannot Be Planned?", I suddenly had a doubt: is the life philosophy of being driven by a "sense of purpose" really always correct?

As the book says, just imagining a life without goals makes us shudder, which shows the position "sense of purpose" holds in our life and work. Not to mention that since childhood, we've always been asked "what kind of person do you want to be when you grow up."

But what does this have to do with "growing skills"? The skill referred to here, in English it is "skill" — in simple terms, a modular, multi-step prompt template that contains a fixed process (or fixed logic). So it belongs neither to life, nor does it seem grand — at most it carries a hint of geek flavor — why can't it be "planned" either?

Why can't "greatness" be planned?

Admittedly, goals can become a faulty compass; blindly persisting can step on the wrong stepping stone; natural selection is also random, unordered, and cannot be formalized. But I think the most important thing, as long as I'm still willing to insist on what I said yesterday—"imagination is indispensable"—is that greatness cannot be planned because greatness cannot be "imagined."

Why "the goal may become a faulty compass"? Because greatness cannot be imagined. Why "blind persistence will step on the wrong踏脚石"? Also because greatness cannot be imagined.

Why "Greatness" Cannot Be Imagined?

What is imagination? Imagination is the moment when your left foot is still on the ground and your right foot steps forward but hasn't landed yet. Someone might say, isn't that wrong? Isn't "a horse soaring through the sky" (天马行空) just the "castle in the air" (空中楼阁) you mentioned yesterday? But I think that's also imagination, isn't it? Let me ask you: what does "a horse soaring through the sky" actually refer to? Isn't it a horse that was originally running on the ground and then flew into the sky? And doesn't that horse on the ground represent the initial, reality-based, ground-stomping left foot?

So "greatness" cannot be imagined, because it is so "great" that it is too far from any real basis to be connected to, even by some kind of leap beyond our known understanding. Borrowing the book's concept of "stepping stones," if greatness means "finding another habitable Earth-like planet," these stepping stones are the waypoints the spacecraft passes through on its journey from Earth. Once we have exhausted the known galaxies where we can stop, beyond this boundary of understanding, even imagination has nowhere to land—so how can we "plan" to cross this unknown expanse to another unknown?

Coming back to today's topic: why can't skills be planned either? Because every feature point they contain is like a transit stop along a spacecraft's journey — some are known, some need to be imagined on top of those known ones. And those feature points that exist only as a vague concept are like galaxies not yet explored; we just call them "that place" or "that thing," without any real idea which "thing" that thing actually is, let alone how to build it.

Actually, letting a "skill" grow isn't a particularly new way of putting it. When OpenClaw was still all the rage, whenever someone stood up to share their experiences, they all ended up talking about the word "grow". What they meant to express, as I understand it, is that "there's no one-step solution"—a skill must accumulate and iterate gradually, allowing it to adapt to your needs over time, rather than starting with a big, all-encompassing "one-stop solution" right out of the gate.

Take a dinner fork as an example. Since we're talking about eating (and using the most typical cutlery scenario — steak and pasta), the requirement is very clear: "something that can pick up noodles and meat and put them into your mouth". Some people will jump in: "Then just make a fork, that's it!" I'd like to point out that if the concept or tool of a "fork" already existed, the requirement in this example would no longer hold. The fact that the requirement holds means that at this moment, no tool — or even concept — called a "fork" yet exists.

With this assumption, our steps would look something like this. First, find a relatively sanitary material, since it will be used as a tool for eating. Second, sharpen one end of this tool, because we need to try to stab the meat with it. Third, realize one point isn't enough, because a large piece of meat, if you don't stab its geometric center, will wobble around, affecting manners—so split the first point into a fork. Fourth, the handle, being just a stick, is easy to slip on, so flatten it to make it easier to grip.

Then came the "fork." A fork alone wasn't enough — the chunks of meat were too large, and gnawing on something speared was tiring. So why not make "something that can separate the meat"? Thus the "knife and fork" set was complete.

At this point, we can understand one thing. If a "skill" requirement, like the "fork", also needs a step-by-step development process, then through this clear analogy, we can — reluctantly but seemingly irrefutably — admit: "skills are grown".

But even I myself feel a bit dazed—when we do software development, the requirements are very clear: a button goes here, that page allows going back, all of that is written out in the documentation. That logic of "once you have the concept of a 'fork,' the need for a 'fork' no longer exists" just doesn't hold up.

But is that really the case?

Is it that this logic doesn't hold, or is it that what you've previously participated in is just a particular specific link in this "fork requirement" with the clearest cause and effect? Like "split another fork at the pointed end"?

Or, more precisely, it depends on whether you treat "skills" as a multifunctional collection with a complete logical chain — a product with a full life cycle — or just as the externalization of "one" feature point. If it's the latter, then what you are making is not a "skill" at all, but a shortcut for a (multi-step) complex prompt — which, in Claude Code, is also called a "command."

So, excluding the editing of simple "commands" with a single, specific function, for a skill in the true sense, it requires a continuous iterative process — a progression from "functional requirement" to "tool-like fulfillment", from the abstract to the concrete.

And in this process, during the development of a "fork as an example," every step has a clear, verifiable boundary of expectation. The first step is "the material must be hygienic"; the second is "the sharpened tip must be able to pierce the meat"; the third is "after piercing, it doesn't wobble"; the fourth is "it doesn't slip in the hand". Throughout the process, there will inevitably be products that don't meet these expectations, and these products will have their own characteristics. Some of those characteristics may happen to meet other "random" and "serendipitous" needs, and so they also survive. Doesn't this process of constant forking and constant evolution strongly resemble another process — that of "evolution"?

At this point, I can say: "A skill, when we are manifesting a need, gradually evolves and is selected from seed to sprout, from coarse to refined, forever being polished—a continuously changing, intermediate state that will only partially satisfy needs, and is intertwined with them, one that will never be, but approaches, perfection."

So why are skills grown rather than taught? Because it's a dynamic approximation, not simply a "single decisive stroke."