Foreword
The urge to write hit out of nowhere this morning. I sat down, and this came out.
What follows is a stitched-together set of things I've been figuring out lately. If you stick around for it — thank you. Thank you for handing me this slice of your attention.
On identity narratives
The story you wrap around an object becomes its operating system.
The most efficient purchase I made this year is a MacBook. I'd had a heavy filter on Macs before owning one — Mac people were the serious-work people, the ones whose lives looked together. The moment the machine landed in my hands, I quietly slid into that identity without noticing.
Once that identity is in place, doing anything recreational on the Mac feels low-grade wrong. The machine became, by default, for work.
The strange part: switching from gaming on the Windows laptop to working on the Mac is now completely frictionless. Back when I had a single machine for both, switching modes was a small act of will every time. Two machines collapsed that friction to zero.
This isn't a buy-a-Mac pitch. Mac and Windows both have their merits, and any meaningful purchase deserves real thought. The point is the narrative. When I open the Mac, I've already silently agreed I'm a person doing work — and the behavior follows the agreement.
If you don't carry the "Mac = work mode" story, the hardware won't conjure one for you.
The flip side: if you can hold an identity firmly enough — I am the kind of person who does X — what hardware you use barely matters.
I'm not pretending I've cleanly cut games out of my life. The Mac has bought me a real productivity gain — I work, then naturally slide back to the Windows machine for games, and back to the Mac for the next round. The split has held for over a month, and Mac usage is way ahead. I'll take that as evidence I'm trending in the right direction (sort of).
On using AI
Use the best one you can get your hands on.
The gap between AI models is enormous. If you've only tried a relay-routed model, or a domestic stand-in that lags the frontier, and concluded "AI is fine, I guess" — you've quietly killed your own appetite to keep exploring. You'll stop tracking new tools, new capabilities. That loss of curiosity costs more than the price of any subscription.
There are guides everywhere. Search, hand-test, just put your hands on the real thing.
And if you are using a frontier model and only chatting idly with it, you're also missing the picture. To actually feel where the field is, you have to push the tool against a real project. Personal interest, market need, doesn't matter. The shape of an AI tool only reveals itself when you take something through it from zero to one.
A while back I planned and shipped a real project that way, and that's when I fell in love with Claude Code. I felt — concretely — what it was good at, and where the bottlenecks still were.
So: use AI on real problems, not toy ones. (And, side note — feedback from a senior collaborator on that same project taught me how much project work depends on tight iteration loops. Worth saying out loud.)
On building a thing in public
"Building a personal brand is just being yourself."
I never really got that line. Now I think I do.
The unlock came from two casual posts I made over winter break — game guides for Endfield, dropped onto Xiaohongshu. I wasn't thinking about traffic or monetization. I was just having fun, and at some point thought huh, maybe this'll be useful for someone. Posted it. Both took off — together close to 10,000 views — and I picked up a small payout from the game's creator program. The whole guide-uploader subculture suddenly made sense.
The cash was a happy surprise. The real thing was that the experience lit up my appetite to keep making.
Around the same time I picked up a piece of advice from a talk — enrich the parameters of your life and record them — and that became the angle. Now I post whatever I feel like on Xiaohongshu: game rants, daily fragments, builds, screenshots of weird AI moments, even old written reflections turned into image-and-text posts. Loose, unfocused, and I don't mind.
Most posts sit at three-digit views, two-digit likes. I genuinely don't care. Pure record-keeping. The backend dashboard actually looks fine — volume probably does the lifting.
There's a take I keep coming back to: if your post doesn't go viral, post 100. Then 1,000. Then 10,000. Quantity buys you the lottery ticket of one going viral.
The previous time I tried Xiaohongshu, I was forcing myself to stay within one niche. That's exhausting. Forcing yourself to ship within a single category is how creators burn out and quit. This time, just write what I feel, and the whole thing is light.
I'm honestly not sure if what I'm doing counts as "creator work". Right now it's closer to recording. Conversion rates, follower counts — none of that occupies me.
If you're a professional creator, none of this is news. Thanks for tolerating the ramble. If anything sprouts on Xiaohongshu down the road — that's a bonus.
On just keep buying
Keep dollar-cost averaging.
During the recent stretch when US stocks and gold were both doing their dramatic dance, I happened to glance at my account and got jolted — wall of green. Not gonna lie, the mood took a hit.
The thing that stopped the spiral was a book I'd read earlier: Nick Maggiulli's Just Keep Buying. The argument is airtight — even crashes that gut markets eventually recover. The current dips aren't even in that league.
That cooled me down instantly. I'm not selling next quarter. The money in there is on a three-to-five-year horizon. Small movements are noise. Just keep buying. Time does the rest. And — predictably — by now things have rebounded. The buys I made on the way down? They quietly lowered my cost basis.
Things will dip again. Just keep buying.
On charging for what you make
The hard part wasn't pricing it. It was admitting it had value.
I recently announced my Claude Code playbook to friends and was twisted up about pricing. Free? Paid? I couldn't decide.
What I was actually afraid of was that nobody would think it was worth anything. Working through it with Claude, I landed on the current plan — keep 80% open, gate the higher-leverage core behind a small paywall. Priced at ¥9.9 — about a coffee — to keep the friction low.
Real thanks to the people who DM'd to pay. The thing that surprised me wasn't the money. It was the feeling that someone valued what I'd made. That hit harder than any of the cash going through.
Trust is its own kind of inspire. We're social animals — those small exchanges between people matter more than the line items. The fact that someone trusted me enough to send money — that trust was the real thing. The cash was almost incidental. What mattered was the moment of recognition: this thing I made is something someone actually needed. That feeling is rare for me, and I want to remember it.
So the best version of any of this is always: earn money as a side effect of making something.
Thanks again to everyone who paid in. I've been chewing on some recent Karpathy notes — once I finish the next iteration I'll write up what I learn and ship it free to everyone on the list.
On family
Family is the topic nobody escapes.
When I was a kid, my home had its share of fights. Middle-school me went full rebellion, even ran away once (in retrospect — really dumb). Then high school and university smoothed things out. Partly: I just got less stupid. (Though looking back from now, plenty of the high-school me was also dumb.)
What's actually changed the relationship in the last few years is small, deliberate stuff I started doing in college. The cadence of weekly video calls slowly became every three or four days. The default of don't share anything with them slowly turned into actually tell them what I'm doing.
I used to resist sharing because I assumed they'd kill the vibe. The truth is simpler: an older generation has more locked-in mental models — they don't catch your point — and that's what made me angry. Not anything deeper.
Some habits I can teach them. Some worldviews I cannot move. So I stopped trying to move those.
I just talk to them on the channel they actually receive on.
And the relationship loosens. The closeness comes from somewhere different than I expected. Change is gradual, and it doesn't go in the direction I used to fantasize about. We can't demand our parents change. We can change our own posture, and the protocol we use with them.
A few honest questions to ask yourself:
- Do you actually want to change the relationship, or is it just something you say?
- Have you read anything serious on the topic, looked into it at all?
- Have you genuinely exhausted what you could try — or did you make a half-hearted attempt and then declare the relationship "unchangeable"?
I think you already know the answer.
The simplest thing that works: just video-call them more. Even if it's nothing-conversation.
I've hit 3000 字 on the Chinese side, and I think this might be one of the few pieces I've shipped without putting it through an AI for a polish pass. Usually I run a sweep — fix any awkward phrasing, swap a clumsy sentence for a smoother one, tighten the rhythm — because the read becomes nicer that way.
Today I just didn't feel like it. Some of the lines may not flow cleanly. That unevenness is mine.