Note

This English version was translated and polished with the assistance of an LLM (Large Language Model). The structure, ideas, examples, and final editorial decisions remain my own.

I used to keep wanting to write something, but I never actually wrote it, so all that remained was the idea that I wanted to write something. It was not because my mind was empty. It was because I had too many thoughts, too tangled, like a ball of yarn someone had kneaded into a mess. Threads were everywhere, and that somehow made it harder to know where to begin.

Many times, driven by that impulse, I would open a document and then stare at the blank page for half a day. There were clearly plenty of things in my head: a few opinions, a few examples, a little bit of emotion, and sometimes even a few sentences that I secretly thought were pretty good. But on the screen there was only a blinking cursor. My hands rested on the keyboard, and after a long while, I still could not squeeze out a single word.

Or I would finally write two sentences and immediately feel that something was off. It was not that anything was exactly wrong. It just felt insufficiently precise. So I would backspace, rewrite, delete again. After wrestling with it for ages, I would be left with an untitled document and one exhausted soul.

Maybe you have had a similar experience too.

Talking with other people is clearly not like this. You say one sentence, I say one sentence, and many things that were originally vague slowly become clearer in the conversation. Sometimes you get more and more excited, and before you know it, you have talked through the whole night. But once you try to turn it into an article, the whole person suddenly jams up, as if one word is too much to produce.

Only later did I realize that the biggest difference between writing and conversation may not simply be whether one is formal.

It is not that there are no ideas, but that you do not know where to start

The first thing that blocks people in writing is not a lack of ideas or material. It is more like the disorder of thoughts and content.

Often, ideas happen in the mind all at once. They crowd together there, sometimes without clear boundaries between them, but an article asks them to line up in a single sequence. Worse still, an article is not a chat. From the beginning, it seems to demand that you hand over something already formed: it needs an opening, a structure, a focus, and preferably it should not be too boring.

It is like putting a pile of books scattered across the floor back onto a shelf. The annoying part is not only picking the books up. You first have to know how the shelf is divided: which books belong together, which one should go in front, and which one you merely happened to flip through and should not put there at all.

So when writing, people often get stuck on the very first sentence.

It is not that they do not know what to say. It is that every thought seems like it could become the opening, and every angle seems like it could grow into an article. The more possibilities there are, the harder it becomes to begin.

In other words:

The difficulty of writing is often not “blankness”, but “messiness”. It is not that you have no knowledge, no opinions, no ideas. It is that there are too many things, so many that they squeeze into a knot.

Before it lands, it has already become distorted

Even after you finally manage to start, you soon discover that there is a second gate. The discomfort shifts from not being able to write to feeling, after you have written something, that it does not quite have the right flavor. It seems to have lost many things that are hard to name.

The thought in your head is three-dimensional, complex, floating high in the sky. It has emotion, tone, background, causes and consequences, and it feels complete and abundant. But once it lands on the page, it is as if it has fallen from the sky and lost some of its texture.1

That is because thoughts in the mind are not pure text. They may be feelings brought by an experience, images from a certain scene, the tone behind a sentence, and a lot of background that you know but others may not. An article, however, can only move forward one word at a time. It has to compress these mixed, three-dimensional, warm things into a linear expression.

So writing itself is a kind of processing, and also a kind of translation.

And as long as something is processed, there will inevitably be loss. That is why we often feel that we clearly understood something in our own minds, but once we write it down, it seems to lose a layer. It is not that the meaning completely disappears. It is that the sense of wholeness, vividness, and closeness has been shaved away by language.

Then if both are forms of expression, why is talking easier than writing?

Because conversation is streaming.

Conversation does not require you to get everything right in one shot. You can say a rough version first. If it is unclear, you add another sentence. If the other person does not understand, you explain again. More importantly, conversation has feedback. The other person’s expression, questions, pauses, and responses all help you judge: “Should I say this another way?”

This also shows that understanding in conversation does not rely entirely on the words themselves.

You might begin by asking, “Uh… have you ever had this feeling?” Then you express that feeling in fragments, and the other person may still understand. You say, “You know what I mean, right?” and they nod. For the moment, that shared understanding is established. Tone, expression, pauses, and shared experience all become part of the context. They allow expression to be incomplete at first, and they let thought calibrate itself gradually while it is being transmitted.

But an article has none of these things.

An article has no instant follow-up questions, no facial feedback, and no person sitting across from you saying, “Is this what you mean?” Once it is written down, it has to stand on its own. The reader will not wait for you to slowly add more context. They can only understand what you mean from the words in front of them.

So you have to translate many private feelings into expressions that are more public and more generally understandable.

Originally, you may only be describing a very specific stuck feeling that you have experienced. In conversation, you might first ask, “Have you ever run into this kind of situation?” and then slowly test, slowly organize your words, until the two of you arrive at a shared recognition.

But when writing an article, you have to find a way for more people to understand that stuckness. You know why it feels bad, but the article cannot stop at “I just feel something is wrong”, “I just feel uncomfortable”, or “I feel like this is how it is.” You have to break down that “feeling” and turn it into something other people can read and experience.

And it is precisely in this process of breaking down and translating that many originally vivid things begin to distort.

So writing is not merely putting thoughts onto the page.

It is more like a translation: taking the mixed, three-dimensional things in your mind, things with private context, and processing them into words that a stranger can understand independently.

While writing, you start fighting with yourself first

Precisely because what we write always differs from what is in our heads, writing easily enters another state: hesitation and self-consumption.

You write a small paragraph, then hold it up and examine it again and again, staring at every sentence you just wrote.

Is this opening not good enough?

Is this example too wordy?

Should this paragraph be deleted?

Is this sentence accurate?

Is this word too ordinary?

Should this character be added or not?

Writing is no longer just writing. It becomes a tug-of-war of correction and calibration.

One side says, since you finally managed to write it, “good enough, keep going and fix it later.” The other side says no, this sentence is still not accurate enough, something is still missing, and it feels unbearable not to change it. One person is responsible for moving forward, while another points fingers and steps on the brake. The paragraph staggers along; every two sentences, you drag yourself back to revise once more.

While writing, you start fighting with yourself first.

This is also why many people fall into a strange loop when they write: write one sentence, delete one sentence; try another opening, still feel it is wrong; finally produce a paragraph, then feel the whole paragraph is not it either. Then they keep choosing between words and sentences, making decision after decision.

And when there are too many choices, mental pressure increases. Attention and patience are limited. You are writing, judging, filtering, denying, and imagining how others will read it all at once. Is that tiring? Of course it is.

More importantly, the medium of an article is naturally heavier. The same sentence may look pretty good in a chat box. Posted as a short status, it may seem fine. But once that sentence becomes part of a full article, you suddenly feel that it is a little awkward, maybe even a little annoying.

The sentence itself has not changed. What changes is the container that carries it.

Compared with words casually spoken, the article as a container is too “hard”. It makes people unconsciously raise their standards and look at their own sentences with harsher eyes. So before the thought has truly unfolded, the picky self has already stepped forward.

In the end, you discover:

You are not only expressing an idea. You are also constantly repairing the gap between that expression and the idea in your mind.

If we put all these problems together, we can see that the real difficulty of writing is not just one specific link.

It is not simply a lack of ideas, nor merely poor expression. It asks one person to complete too many tasks at the same time: organize messy thoughts into order, compress multidimensional feelings into text, and while writing, keep checking what is inaccurate, what needs to be deleted, and what still needs to be added.

Mermaid
flowchart TD
    A[Why is writing hard?]
    A --> B[Hard to begin]
    A --> C[It feels distorted once written]
    A --> D[Self-conflict while writing]

    B --> B1[Too many messy ideas]
    B --> B2[Articles require linear structure]
    B --> B3[The medium feels too much like a final product]

    C --> C1[Thoughts and text are made of different things]
    C --> C2[Lack of instant feedback]
    C --> C3[Private context must become public expression]

    D --> D1[Fear of inaccurate expression]
    D --> D2[Too many choices create decision fatigue]

This is why many people clearly have many things in their heads, yet get stuck as soon as they arrive at the writing stage.

Because writing is not simply pouring thoughts out. It is organizing, refining, translating, and proofreading those thoughts before delivering them as something that can exist independently.

The problem is that the brain is not necessarily suited to doing all of this from the very beginning.

So instead of forcing yourself to write an article directly, a more natural approach may be to return to conversation first.

So, do not rush to write the article

If writing is like compressing thought into a block of stone, conversation is more like letting thought become a small stream.

Mermaid
flowchart LR
    A[Conversation] --> A1[Streaming expression]
    A --> A2[Instant feedback]
    A --> A3[Allows additions and corrections]

    B[Writing] --> B1[Block-like finished draft]
    B --> B2[Lacks feedback]
    B --> B3[Must stand independently]
    B --> B4[High-pressure packaging]

The advantage of conversation is that it does not require you to be complete at the start. It has guidance, feedback, and tolerance for mistakes. You can first give the other person a very rough thought, then be led by their questions to add, correct, and reorganize. In this way, many things that were only feelings and chaos begin to take shape as you say them.

This is also why I feel that turning a saved conversation into an article may suit many people who have similar difficulties better than starting directly from writing.

But by “returning to conversation”, I do not mean moving a chat transcript into an article unchanged. I mean first using conversation to let the idea settle into shape.

Mermaid
flowchart TD
    A[Say it out first] --> B[Review the whole conversation]
    B --> C[Build the writing structure]
    C --> D[Make the language clearer and easier to understand]
    D --> E[Final polishing]

    A --> A1[Low-pressure generation]
    B --> B1[Pull out keywords and loose threads]
    C --> C1[Cut the stream into blocks]
    D --> D1[Turn private context into public expression]
    E --> E1[Trim edges, prune branches, close the voice]

Say it out first

At first (kidding) If you now have a somewhat unique thought or insight, maybe we can start by closing the document.

Find a close friend, or someone who may not fully understand but is still willing to listen to your thoughts.2 Ideally, find someone who is also highly interested in the topic. Whether face to face or through text, go talk with them first.

Or, if you feel awkward, socially anxious, or simply cannot find anyone, a capable large language model with decent reasoning ability can also work.

As of May 2026,3 models suitable for this kind of deep conversation roughly fall into two groups: proprietary models, and open-weight or more open models.

Proprietary models include:

  • ChatGPT 5.5 Thinking
  • Claude Opus 4.7 Thinking (personally, I feel Sonnet 4.6 Thinking is also okay)
  • Gemini 3.1 Pro

Open-weight or more open model ecosystems include:

  • DeepSeek Expert Mode (DeepSeek-V4-Pro) + deep thinking (try not to enable intelligent search4)
  • Zhipu/z.ai GLM 5.1 Thinking Mode

Or if all else fails, you can learn from programmers and talk to a little yellow duck,5 explaining everything you are thinking to it.

It sounds a little silly, but it really does work. You can start with a very loose sentence, such as:

Recently I have been thinking about something…

Then follow that opening and keep going. Say whatever comes to mind. It is fine if it is messy. It is fine if it is wrong. It is fine if you wander off halfway through.

If it is an oral conversation, I suggest turning on recording first and converting it to text afterward. If it is an online chat, keep the chat history.

Because when the conversation ends, the valuable part is not only “I finally said it out loud”, but that these words have been recorded.

In the best case, you may have an exhilarating conversation, moving from one question to another, from one field into another. While talking, you will find that many things were not planned in advance at all, but slowly surfaced in the process of speaking.

Review the whole conversation

But after the words have been spoken, do not rush to turn them into an article immediately.

Because the conversation that just ended is still fundamentally fluid. It may be excited, or it may jump around. It may begin with writing, suddenly drift into learning, and later turn toward AI, the desire to express, and self-examination. There are indeed many valuable things inside, but they are scattered in different places.

So the second step is to review the entire conversation.

At this stage, do not rush to polish sentences, and do not rush to judge which sentence is good. What you are doing is more like fishing things out of water: see which words keep appearing, which questions are repeatedly raised, which sentences may look like casual chatter but actually catch the key issue, and which places suddenly open up after wandering around for a while.

Or, if you find it tiring to read, or simply feel too lazy to read it all, a lazier method is to throw the scattered material to AI for summary.6 The material is still yours, but the structure suddenly becomes much smoother.

By this point, it is easy to see that the real thread of the article has already been hiding inside the conversation.

For example, this article itself began as a casual feeling: “Why can I talk endlessly like a chatterbox when discussing something with others, but once I start writing, I get stuck and cannot move a single word for ages?” But during the conversation, some fresh observations kept popping out, such as the differences between writing and conversation, where exactly writing becomes hard, and then a thought naturally appeared: why not use conversation directly as the preparatory step for writing?

So you discover that the large, scattered conversation that originally seemed impossible to understand is no longer a mess. Some loose threads have begun to appear inside it.

And what we need to do next is follow those threads and build the structure of the article.

Build the writing structure

At this step, you can no longer completely follow the order of the conversation.

The order of conversation is usually the order in which thought naturally flows. It is often jumpy and does not necessarily have a clear logic. An article is different. It needs some order so that readers can understand smoothly.

So these two things are not the same.

When chatting, you may first state a conclusion, such as “I feel like…” Then you suddenly remember an example: “For example…” As you keep talking, you add some background: “Like in daily life…” Later, you realize the real issue is not here but somewhere else. Conversation can be this loose because the other person is present. They can follow your thoughts, and if they do not understand, they can ask at any time.

But an article cannot.

An article requires you to rearrange the material: what should go at the beginning, what should go in the middle, what should be left for later, and what was merely brought up in passing during the chat, not very important, and can be deleted.

This step is essentially cutting the “stream” into “blocks”, then stacking them neatly.

For example, if this article were written strictly according to the original conversational order, it would have a problem: one moment it talks about writing, the next it discusses chatting, then it evaluates AI, then it wanders into psychology. Everything would be scattered. But if we reorganize it, it can become a clearer structure:

  1. Start with a real-life scene: why opening a document leads to staring blankly.
  2. Raise the question: why do we talk endlessly in conversation but get stuck when writing?
  3. Then break down why writing is hard: the beginning is difficult, the written result feels distorted, and we become tangled while writing.
  4. Then explain why conversation does not have these problems.
  5. Finally give a method: how to turn scattered conversation into an organized article.

In this way, the material scattered throughout the conversation is no longer just a pile of fragments. It begins to become the skeleton that can support an article.

So when building this structure, do not rush to make every sentence beautiful.

What you really need to do first is ask yourself several questions:

What problem does this article truly want to solve?

Which material should be used at the beginning to build resonance more effectively?

Which parts must be developed in the body?

Which parts were merely mentioned in passing during the conversation and can be deleted?

In what order should the reader understand this most naturally?

Once these questions have answers, the article is no longer a pile of raw material.

It begins to have an entrance, scenery along the way, and a destination.

Make the language clearer and easier to understand

Once the skeleton is built, the article has its spine.

But an article cannot consist only of bones. Gnawing on bones is not very interesting. It needs at least some flesh. And even then, readers may not be able to understand it smoothly.

Because sentences pulled out of conversation often carry a lot of private context that only existed at the conversational scene.

At this point, we can finally do something we have been wanting to do but have not allowed ourselves to do yet: make the language clearer and easier to understand.

But “clearer and easier to understand” does not mean making the article shallow, nor does it mean turning everything into overly plain and laughably loose wording.

What it really asks you to do is take the things that only worked in the live conversation and rewrite them into expressions that a stranger can understand.

For example, in conversation you can say:

“You know, it is just that feeling.”

But you cannot write that in an article. It will only leave the reader confused. The article needs to go one step further and tell the reader what “that feeling” actually is: staring at a document for half a day after opening it, having a mind full of thoughts but not knowing where to start, writing two sentences and feeling something is wrong, then deleting them and starting over.

Conversation can rely on tone, pauses, and shared experience to complete the meaning. It can patch itself through multiple rounds. An article has to bring back as much as possible of what was omitted and what was left half-said.

So making language clearer and easier to understand does not mean making deep things shallow. It means letting suspended things land, making vague content specific, and turning the parts only you understand into expressions that more people can read.

Final polishing

By this stage, the article has basically stood up.

It has material as its foundation, a structure to support it, and many things that originally only worked in conversation have been rewritten into forms readers can understand. What comes next is no longer starting from zero, but pushing the article further toward “feeling like an article”.

At this point, you can begin looking at which parts are repetitive and should be removed, patching jumps in logic, and adjusting sentences that are too colloquial, too complex, or too loose. You can also check whether paragraphs connect, whether headings and body text echo each other, and whether the question raised at the beginning receives a response later.

But one thing still requires care: do not turn polishing into another form of self-consumption.

Polishing is not about making every sentence flawless. That can erase some of your own style. Polishing is about making the article smoother and clearer so that readers can keep going. Many sentences do not have to be especially beautiful. If they are accurate, natural, and suitable for where they are, that is already enough.

You can focus on a few questions:

Are there repetitions or unnecessary verbosity?

Does any paragraph skip a logical step?

Are there expressions that only I understand but readers may not?

Are some sentences too long, making them hard to breathe through?

Are there places that clearly could be deleted, but I kept them only because I was reluctant?

After checking these questions, the article usually becomes much cleaner.

Finally, look back at the beginning and the ending.

The opening should bring people in and let readers know why this article relates to them. The ending should gather the whole article together and let readers take something away. It does not have to force a grand elevation, but at least it should make people feel that the journey was not for nothing.

So the final polish is not about rebuilding the skeleton. It is about trimming edges and pruning branches.

It turns the article from “readable” into “easy to understand”, from “having content” into “having a sense of completeness”.

And at this point, an article grown out of conversation can finally be considered formed.

Conclusion

So being unable to write does not necessarily mean that you have no ideas, nor does it mean that you are unsuited to writing.

Often, it is simply because we put ourselves into the position of “finished draft” too early, giving it expectations and goals that are too high. Before the idea has emerged, we rush to demand that it have a complete structure. Before the feeling has been explained clearly, we rush to demand that it “cash out” into a decent paragraph. Before the article has truly stood up, we rush to judge whether it is good enough.

The meaning of conversation lies in giving thought an easier option, making it easier for thought to break through.

It allows you to be a little messy first, a little clumsy first, a little vague first. In sentence after sentence of expression, you can see what you are actually thinking. Through repeated additions, questions, and reorganizations, you can slowly make clear the things that were originally only feelings.

So for those who always feel that they do not know where to start with writing, perhaps there is no need to force yourself to open a document and produce a proper article right away.

Say it first.

As you keep talking, the thing you truly want to write may already have begun to take shape.

  1. In psychology, there is a concept called the “Illusion of Explanatory Depth” (IOED). We often overestimate our understanding of a mechanism or causal process until we actually have to explain it step by step, at which point we discover many gaps and breaks in the middle. ↩

  2. If the listener really does not understand, this approach is also somewhat close to what people often call the “Feynman technique”: trying to explain something to an outsider often forces you to notice what you do and do not know. But if the other person also understands the topic, the advantage of conversation becomes mutual inspiration, and the topic can spread much further. ↩

  3. Specific model versions keep changing, so what matters more here is the selection criteria: long context, strong reasoning or thinking ability, the ability to ask follow-up questions, the ability to organize material, and low hallucination, rather than memorizing a specific version name. ↩

  4. Web search introduces more external information, and it may also bring unreliable sources and extra noise. So if the goal is to organize your own thoughts rather than research facts, it may be better not to enable web access at first. Besides, any AI may make mistakes, so what I emphasize is still conversation rather than interrogation: treat it like a friend who knows a bit of everything but is not always reliable, and keep your own critical thinking. ↩

  5. Programmers sometimes try to explain their code line by line to a little yellow duck or any other inanimate object, as if the other party could understand the logic of the code. The core idea is to clarify your own thinking through spoken explanation. This method is known as rubber duck debugging, and in practice it is also a form of the Feynman technique. ↩

  6. Large language models are quite good at pattern recognition, summarization, and reorganization, so they are well suited to extracting keywords, themes, and structure from a long, messy conversation. But summarizing is not the same as understanding, nor does it mean the model has completely understood your intent correctly. In the end, you still need to decide for yourself where the material should go. ↩