The joke's on me, I guess. And I'm okay with that.

scope/life thought/consciousness

I'm dictating this into my phone right now. Not typing—speaking. My voice becomes text through an app, and it also tries to fix my grammar. It's absurd, really. I'm writing a blog post by talking to a machine, which then talks back to me through corrections, and somehow this feels more honest than sitting at a keyboard pretending I know what I'm doing.

None of this would have started without a Korean researcher named Suh.

Meeting Suh

A few weeks ago, I met him through a group meeting for our research lab. He came to China for a research project, working with my advisor's team. We then had dinners, and talked. A lot. The problem was obvious: he doesn't speak Chinese, I don't speak Korean, so English became our only bridge. Not by choice, exactly. By necessity. Which is terrifying when you're someone who has spent most of his life knowing English but never actually using it.

This is my confession: I've been a fraud.

For years, I scored well on English exams. Tests, competitions, standardized assessments—I crushed them. But ask me to have a conversation? My brain short-circuited. The knowledge was there, locked away like software installed but never executed. Now, suddenly, there's a real person expecting real responses, and I can't just sit there repeating "I don't speak English" in English like some absurdist theater piece. Except—I already did that once. Multiple times.

Past Failures

Three times I've spoken English to actual humans. Let that sink in.

Basketball Court Panic (Age 14)

A group of men approached me. They looked foreign, spoke English, wanted me to take their photo. I panicked. The word "No" felt impossible. Instead, I went with the ironic script: "I don't speak English, I don't speak English."

They laughed. One of them said: "You're speaking English right now."

I wanted to die. Actually, I wanted to call my father. But here's the cruel joke—my father's English was worse than mine. I was trying to escape the situation by asking for help from someone less equipped than me. That's how broken I was.

They got their photo. Conversation over. I survived.

Movie Chat Freeze (Age 17)

An American exchange student came to eat with us. She lived with my sister's friend's family, which is how she ended up at my table. My father, knowing I was good at English exams, kept nudging me: "Go talk to her. You're good at English."

The pressure was excruciating. We made conversation. She asked what kind of movies I liked. I tried to describe something like Fast & Furious—a car racing movie. But the word "racing" evaporated from my mind. I had "car" and nothing else.

"I like cars... some cars... car movies..." I probably said. Something horrifyingly stupid.

She asked, "Do you mean car racing?"

My brain stopped. The gears literally stopped turning. I nodded, said something noncommittal, and we moved on. That was the end of that conversation.

Safe Intro (Last Year)

A professor from Europe visited to give a talk and meet with our research group. My advisor asked me to introduce myself. I was a brand-new graduate student, fresh meat, with nothing to say. So I gave a self-introduction. Prepared. Nervous. Safe.

I spoke English. No one died. Nothing went wrong because I said almost nothing of substance.

Talking Now

Now I talk to Suh regularly. In English. There's no escape route. No script. No "let me call my dad" option. Just two people trying to communicate across a language barrier where I'm the one who supposedly knows the language.

And it's... working?

Badly, unevenly, with awkward pauses and grammatical wounds, but it's working. I'm saying things. He's understanding me. Sometimes I'm even saying things correctly.

This is when I noticed something about myself.

AI in English

I use language models constantly. For work, for learning, for prompts, for everything. But I realized something recently: when I write to these AI systems in English, they respond better. They understand more. The training data is predominantly English, so of course they do.

So I started speaking to them in English too.

I record voice memos and feed them into the machine. The machine transcribes. Sometimes it gets it wrong—my accent, the background noise, the weird way I describe things—but it tries to make sense of it.

Here's where it gets interesting: the polishing sometimes changes what I meant. It smooths out the edges where my actual thought lived. It's like my broken English meant something that proper English erases. But overall? Talking to AI in English is becoming natural.

I barely use Chinese with these systems anymore. Even with Chinese models—DeepSeek, Kimi— I suspect they work better in English too. The whole world's AI infrastructure is built on English. We're all just walking around in that gravity well whether we know it or not.

Message Bloat

Last night I sent Suh a message on WeChat using voice input. It turned into a massive, rambling, verbose thing. Like an email. A real email. Pages long.

I found it hilarious and embarrassing. In Chinese, I can send fragmented thoughts, emoji, single characters. The medium allows for incompleteness. But with English? There's this psychological weight. Maybe it's because I'm not a native speaker. Maybe it's because I'm hyperaware that every word needs to mean something. So I keep adding, explaining, clarifying. I keep packing meaning into each message until it becomes this bloated thing that no one asked for.

I turn casual messages into obligations.

Blog i18n Puzzle

Here's what I'm actually working on now: I'm adding English to my blog. The interface is easy—strings and translations, a technical problem. But the content? That's a choice.

I have years of articles in Chinese. Do I translate them to English? Here's my dilemma:

If I write in Chinese first and translate to English, I'm basically doing word-for-word conversion. I'm not thinking about how English speakers actually talk. I'm not reaching for idioms, cultural context, the rhythm that makes English feel right. My translations would be technically correct but emotionally flat. That's the honest assessment.

But if I write in English first? Then translate to Chinese? That's different. The English version would have the consistency it needs because English-language technical writing is a universal standard. Then when I translate back to Chinese, I'm working with material that's already solid. And since I'm a native Chinese speaker, I have the freedom to adapt, to find the right way to say it in my language, not just the literal way.

It's backwards, maybe. But it makes sense.

Tech Fluency Paradox

Here's what terrifies and fascinates me equally: when I talk about code, keyboards, AI systems, anything technical— my English becomes natural.

I can't explain this completely. I live in Chinese. I work in Chinese. But when someone asks me about my tech stack or my keyboard setup or the weirdness of LLM agents, English just happens. Fluently. Without thinking.

Why Concepts Stick in English

I think it's because the first time I encountered these ideas, they were in English. GitHub is English. Technical documentation is English. Stack Overflow is English. I didn't learn these concepts in Chinese and then translate them. I learned them directly in English, in their native linguistic environment.

So when I need to talk about them, my brain goes straight to the original language. "Agent" sounds right. "代理" (dàilǐ, the Chinese equivalent) sounds wrong— it means "representative" or "proxy", which isn't the same thing at all. The concept is baked in English. To translate it is to betray it.

Code-Switching Reality

This explains something I used to find frustrating: why so many Chinese people code-switch between Chinese and English, especially in technical contexts. It's not affectation. It's not trying to sound smart. It's that the concepts genuinely live in English in their heads. The word "agent", for them, doesn't have a home in Chinese. So they say "agent". It's the path of least cognitive resistance.

And I'm understanding now that I do the same thing. I just didn't realize it until I started being forced to speak English with an actual person.

Response Cache

There's another layer: I've absorbed so much English-language technical thinking— through reading, through projects, through GitHub discussions— that I have a precompiled response system for these topics. I don't need to compose thoughts in real time. They're already there, already organized in English, waiting to be retrieved.

When someone asks me about keyboard switches or React patterns or how language models work, I can feel that cache activating. The language flows. I'm not translating from Chinese. I'm not constructing English from first principles. I'm just... speaking the language that was always there, dormant, waiting for the conversation.

It's like muscle memory. Except it's vocabulary memory.

Sync Anxiety Hack

This connects to something darker: my fear of synchronous communication.

Messages are safe. I can edit, revise, rewrite, delete. But real-time conversation? That's terrifying. With Suh, I have to be present. I have to respond. There's no Cmd+Z, no mulligan, no way to unsay what I've said. Maybe that's why my messages to him become these endless, obsessive things. I'm trying to compress all my editing capacity into asynchronous text. I'm trying to make the message so complete, so thoroughly considered, that it performs the function of multiple drafts.

It doesn't work, obviously. It just makes messages longer.

But here's the hack: Prefer written communication for important things. Use async channels where I have control. Use voice calls and real-time chat for low-stakes things where perfection doesn't matter. Restructure my communication to play to my strengths instead of my weaknesses.

Path Ahead

I don't know where this ends. I'm not a native English speaker and I'll probably never feel entirely comfortable with it. But I'm starting to understand that "comfortable" isn't the point.

The point is: I'm doing it now. Badly, inconsistently, with constant doubt— but I'm doing it.

I'm speaking to a Korean guy in English. I'm dictating my thoughts into voice recorders and letting machines clean them up. I'm writing blog posts in a language that doesn't live in my bones. I'm sitting in the discomfort and typing anyway.

Is it absurd? Yes.

Is it working? Also yes.

The contradiction doesn't need to be resolved. It just needs to be inhabited.

And maybe that's the most honest thing I can say about learning anything: the humiliation is part of the process. The broken attempts are the point. The moment you realize you're actually speaking, even badly, even to a machine, is the moment you've already won.

The joke's on me, I guess.

And I'm okay with that.