A knowledge blogger shared his experience. In 2019, he set a goal to read 100 books a year. In the end, he read about 60 to 70.
His former boss later gave him feedback: It's not that you have too little cognition; it's that you have too much. You need to go do things. Since then, he has read less and worked more.
When I saw this, I felt it described the state of many people.
You've read the books and taken the notes. When discussing the ideas in the book, every word is familiar. But when you actually have to use these things to handle a problem, by the third sentence, you might not be able to continue.
When reading, it's easy to nod along. One book makes sense, and another article makes sense. Even two opposing views might both seem reasonable when placed together. This is because reading doesn't require you to choose one on the spot.
Output forces you to choose.
Writing an article, making a PPT, or preparing a lesson all feel this way. Before starting, you think you've figured it out in your head. Once you actually start, you realize the second and third parts don't connect, the examples don't prove the conclusion, and there are several common terms you can't even clearly define.
This isn't a problem with expression. It's that you haven't thought it through yet.
I also know a writing teacher. He conducted an AI writing experiment, wanting AI to mimic his writing style. He first wrote a set of system prompts, let the AI output a large amount of text, and then picked out flaws sentence by sentence.
Each time, he had to answer three questions: Which sentence was bad, why was it bad, and under what circumstances could it be written that way? Each round of feedback entered the next version of the rules.
After more than a month, he wrote: In this process, the one being trained wasn't the AI; it was me.
Because he originally knew what he didn't like but couldn't state the standards clearly. Only when the AI produced the text could he point out the problems. Pointing it out once wasn't enough; he had to explain why. As these feedbacks accumulated, a set of judgments originally hidden in habit slowly turned into rules that could be articulated.
This is what output does.
It turns "I roughly know" into something that can be inspected. Only when an article is written can others point out what they don't understand. Only when a product is made can users tell you where they are unwilling to use it. Only when a research plan is written can a mentor see where your causal chain is broken.
Negative feedback only becomes useful at this point.
A business friend once told me that negative feedback is positive feedback at the information level. Declining revenue, poor data, and opposition from others at least indicate there is something here that needs to be handled.
But feedback doesn't automatically make people progress.
If someone says an article is incomprehensible, you can feel the reader's level is insufficient. If no one uses a product, you can continue to explain that the market doesn't understand. As long as behavior doesn't change, feedback is just another uncomfortable piece of news.
So the focus of output is not the single action of "publishing." It includes making something, getting feedback, judging whether the feedback is correct, and then deciding what to change.
Every step in this process requires the author to choose for themselves.
This also explains why daily updates don't necessarily form judgment.
Some people post content continuously for a year or two, but when you look at their articles together, it's still the same few sentences. Change to a specific problem, and the original content can't answer it. One of my students suggested a test: Put all the content a person has output in the past into an AI knowledge base and see if this knowledge base can answer specific questions.
I think this is a great test.
If hundreds of articles are put in and it can only generate some correct but useless words, the quantity of those articles is large, but there aren't many judgments inside. If it can answer a specific question and provide past cases, conditions, and different choices, it shows that this person has indeed left something behind.
Therefore, I don't agree with the saying "as long as you output in large quantities, you will eventually change your fate."
Without input, output quickly becomes thin. Without facts, cases, and new experiences, you can only repeatedly use judgments you've already made. I have a relative who writes all the time, and he has an observation: Some people are not hollowed out by output because their input speed is greater than their output speed.
This "input" doesn't just refer to reading books. Doing projects, meeting clients, handling failures, and discussing with others all add material. When a person has truly experienced and handled more things, the content they write will naturally have details. Without these things, relying only on increasing publishing frequency will quickly make the writing empty.
Reading and output are not an either-or choice.
Reading provides concepts, facts, and others' experiences. Output requires you to reorganize these materials and provide your own causal relationships. Feedback lets you see if this set of explanations still holds up for others. Revision leaves the final choice.
Since the emergence of AI, these steps have become more distinct.
In the past, being unable to write might have been due to a lack of organizational ability or typing too slowly. Now, generating a structurally complete article is not difficult. Titles, cases, and conclusions can all be there. Thus, "writing it out" itself is less indicative of ability than before.
What truly creates a gap is what materials you gave the AI, what you deleted, what facts you added, and finally, which judgment you kept.
AI can generate ten versions. It cannot decide for you which viewpoint you truly agree with, which example can withstand scrutiny, and which sentence you are willing to take responsibility for under your name.
This decision still has to be made by yourself.
Output doesn't have to be public. Research notes, code, unpublished drafts, and course outlines can all complete this round of processing. The key is whether you are clearer about the problem after finishing than when you started.
If not, you've just changed the format of the materials.
If so, reading finally begins to turn into your judgment.
