Weekly Vol. 14

I'm at home in Chengdu, and I just finished watching the LPL Split 2 final. BLG swept TES 3-0, completing a run of five straight series wins from the losers' bracket to take the title.

About half a month ago I already wanted to rename "Xiluluke's Weekend Letter" to just "Weekly." Thinking it over, what I write every weekend no longer really matches the name. I'm going to keep writing in this weekly format, so I figured I might as well change the name sooner rather than later.

I didn't want anything fancy either, so I just went with "Weekly."

Posting

Starting in June, I stopped the daily posting I'd kept up for a while. Not because I don't want to write anymore, but because I want to rethink what I write.

More than one friend has told me that I'm nothing like my posts. What do they mean? In real life I'm about half funny, half serious. With friends I'm usually the funny one. But reading my posts, most people would probably think I'm a serious, stiff person.

I don't think that means the posts aren't the real me. But after hearing it enough times, I started to wonder: should I loosen up a bit in my writing too?

I have two friends who've been posting daily all along, Liang Mouyin and Yuehui. They post every day, and their content and style are each their own, really distinctive. Sometimes I think I rarely write posts like theirs. Is it that I can't? I don't think so. It's just that I choose to put certain kinds of things on my WeChat account. But why do I choose that way?

So I've been thinking it over lately, and I've decided to adjust my writing style and direction on my WeChat account. Partly so posting doesn't become a burden, and partly so my writing can show the real me more fully.

The Cold

I actually caught a cold last week, but I only got over it this week.

Sometimes a cold comes with no warning at all. I wake up in the morning and my throat starts to hurt. For me, a cold always starts with a sore throat, and by now I basically know how it goes.

The first two days, a sore throat. On the third day the soreness goes away and I get phlegm and a runny nose, and then it basically stays like that. Since I have chronic rhinitis, I always wait until the fourth or fifth day to go to the clinic for medicine.

I don't know if it's that I recover well or that the clinic's medicine works, but every time, after two days of medicine and two more days of rest, the cold just goes away.

That sounds quick, but start to finish it still drags on for over a week. It hits how I feel, and it puts my workout and training plans on hold too.

See? It's been almost half a month since I last worked out, again.

Wearing a mask really does lower your chances of catching a cold. A couple of years ago I still wore one on the subway, and thinking back, I really didn't catch many colds then. Now I've gotten lazy and don't wear one on the subway anymore. Who knows how many of my colds came from not wearing a mask.

I've had two colds this year. The first had a clear cause. One night over Chinese New Year I'd been drinking, woke up at 2 a.m. wide awake, and went out to sit on the balcony in just a thin shirt, catching the wind for two hours. The next day I had a cold.

This time I honestly can't figure out why. I didn't get cold, I didn't get overheated. Very strange.

The Entrance Exam

This weekend is Chengdu's high school entrance exam, and my brother is one of this year's test takers.

I took the exam in 2017, exactly nine years ago. A lot of my memories of middle school have gone fuzzy. Back then, at graduation, it was popular to fill in each other's yearbook pages and get friends to sign your school uniform. That uniform of mine is still hanging at my mom's place.

A couple of years ago I dug it out of the closet. Looking at the names on it, every one of them once so familiar, I ran through old memories in my head. One classmate's name with the surname Li caught my eye, and I couldn't for the life of me remember who that was.

In middle school I was kind of the social butterfly of the class and got along well with lots of people, though there were a few I was closer to.

But after middle school, every one of the classmates I was close to, except Zhang Jiahua, dropped out. Fresh out of school, I never would have imagined that we, who were basically best friends right up to the exam, would never be in touch again in the nine years that followed.

Now I only hear about them now and then and vaguely know what kind of work they do, but I really haven't seen them since.

Zhang Jiahua is the friend I've stayed close to ever since middle school, and the only one from my middle school class.

I also have this feeling that ever since middle school ended, time has flown. High school and college went by in a snap of the fingers, and suddenly I was a corporate drone.

Soon my brother will be entering this brand-new stage of life too. He's fifteen this year, and he's changed a lot in these three years of middle school. His personality has grown, he's started having his own ideas, and he's got things he likes and gets hooked on.

Looking back at myself in middle school, I felt lost and helpless, but from where I stand now, it was all worrying over nothing. Still, I have to admit that growing up just comes with lots of "growing pains." And when parents can't solve them, all that's left is fate and yourself.

Work

Work's been a bit busy lately. From last week through this week I've hardly had any rest (sure, sure, ambush me right while I've got a cold). That's how my job is: when it's busy it's insanely busy, and when it's slow it's deadly slow.

This week I happened to open 51job and saw a job posting from a company whose main business is pretty similar to my current company's. So without a second thought, I sent in my résumé.

I've been at my current company ever since I graduated, four years now. It's not that anything's wrong with the job. Honestly, I think it's pretty good. The workload and pay match up.

I sent the résumé for a few reasons:

• First, this company is in Shenzhen. I've always wanted to join a Shenzhen company and cover my home region, Sichuan. I think Shenzhen companies are, for one, full of life, with mostly young employees, and for another, they have more complete systems and better pay and benefits.
• Second, I could learn new things there. Besides the similar core business, this company has new lines of business my current company doesn't, areas I don't know. If I could go, I'd get to learn more.
• Third, sending a résumé is one thing, and whether I get a shot is another. Given the first two points, it doesn't hurt to apply.

The healthcare industry is shrinking and tightening right now. Lots of engineers in positions like mine have been laid off by big companies and either switched careers or are out of work. At my current company there isn't much room to grow, but at least it's quiet.

The day after I sent my résumé, I saw that over a hundred people had already applied for the position. Terrifying. It also says something about the job market in healthcare right now.

But my current position is stable, so I'm sitting pretty, not rattled at all. It's just that if a better chance comes along, I'd still like to grab it.

To me, any job has its pros and cons.

Take my current job. The company is pretty loosely run. A lot of the time things don't get handled properly, and the processes aren't clear. But on the other hand, employees aren't tied down much either.

The pay isn't high, but thanks to how well I handle problems, I still get over ten days off a month. I can use that time freely to grow, for working out, cycling, reading, vibe coding and so on. If it were nine to five with weekends off every week, I'd actually feel a bit out of place.

What's it called? You can't have your cake and eat it too. If you enjoy the good side, you have to put up with the bad side.

Reading

This week I'm still reading Google's AI Brain, the biography of DeepMind founder Demis Hassabis.

I read pretty slowly. Unless a book is really gripping, I can hardly finish one in a week. So far I've finished the first two parts and just started on part three.

The biggest impression so far is that Hassabis's life and his startups were all shaped by games.

As a kid he was a chess prodigy.¹ That's a game.

Later, as a teenager before college, he developed games at a game company.²

Then for his first startup, he founded Elixir Studios.³ That's a game company.

Later he founded DeepMind,⁴ and on the road of exploring AI there were lots of cases of beating games, including Atari's Breakout⁵ (a game that was designed by Jobs and Wozniak, no less).

After that came AlphaGo,⁶ research with Go as the breakthrough target. Another game.

Only after reading the book did I learn that after cracking Go, the DeepMind team also put work into StarCraft II,⁷ trying to beat the very best human players at it, but it seems that project ended up failing.

And later, when Hassabis cracked the protein folding problem and won the Nobel Prize in Chemistry,⁸ that too came from a game, one called Foldit,⁹ made specifically for simulating how amino acids are arranged in proteins.

"We believe that if something is structured like a game, DeepMind can apply reinforcement learning and make progress."

I have to say, Hassabis and his team are just too strong. Though reading this far, I think I get why Gemini isn't as good as Claude and ChatGPT now: while OpenAI was working on LLMs, Hassabis was still busy with games, and that put Google years behind (just kidding).

The book's okay overall, but the drama in part two about the company's growth, the industry's growth and the tug-of-war with business was way too boring. By the last few sections of part two I couldn't take it and skipped ahead.

There's a lot of technical content on AI in the book, and I learned a few things.

What stuck with me most was Google's Transformer technology and the paper called "Attention Is All You Need."
¹⁰

"In many of their waking moments, people dwell on yesterday or look ahead to tomorrow. Even in the present, they're often immersed in the past and the future. In other words, what governs their experience isn't the relentless ticking of the clock but where they choose to focus their attention."

"A 1960 talk by a Bell Labs scientist, 'I read it again and again as a grad student. One of the things he said was that if you've thought about a problem from every angle, you'll recognize the right solution the moment it appears.'"

Book of Truth and Wenchi NovelPilot

Book of Truth, my AI long-form novel writing tool, officially got a new name during today's development: Wenchi · NovelPilot.

When I started the project, Book of Truth was a name I came up with on the spot. I just wanted something bold and powerful. But as development went on, the name started to feel off.

It had a religious vibe.

Then one day I searched "Book of Truth" on WeChat, and besides my own past posts mentioning it, I found that it's the holy book of a cult.

That scared me into wanting to rename the product right away, and today the name is basically settled.

It's called Wenchi · NovelPilot.

The Wenchi project now has 150+ commits. Apart from the prose system, the other features are getting more and more complete.

In my product philosophy, the prose system and the outline system are Wenchi's core. The outline system is much more complex than the writing system. It's not just about writing the outline. The outline has to be laid out and planned well so the pacing works, and then the prose system can write more structured text.

The outline system also pulls in several subsystems, which is the most troublesome part of development. This week, because of work, the project basically didn't move, and I only made some progress today.

Also, I'm still stuck on the front-end design. It feels like the AI world has no good front-end designers.

I'm now planning to rebuild Wenchi's front end following the design philosophy of an IDE. No idea what it'll end up looking like.

Final Words

Looking at the WeChat Official Accounts assistant recently, I found out that posting your own WeChat QR code on your account isn't allowed.

So from now on I won't put my WeChat on the account. Not many people add me anyway.

References

Footnotes

1. Demis Hassabis was a strong chess player as a boy, reaching master level at 13 and ranking second in the world for his age group at one point. See his public biographical materials and DeepMind's website. ↩

2. As a teenager, Hassabis worked at the British game company Bullfrog Productions as co-designer and lead programmer of the classic management sim Theme Park (1994). ↩

3. In 1998, Hassabis founded the game studio Elixir Studios, known for titles such as Republic: The Revolution. ↩

4. DeepMind was founded in London in 2010 by Hassabis, Shane Legg and Mustafa Suleyman, and was acquired by Google in 2014. ↩

5. Breakout is an arcade game released by Atari in 1976, originally designed and built by Steve Wozniak with Steve Jobs's involvement. In 2015, in DeepMind's DQN research published in Nature, an AI taught itself to play dozens of Atari games, Breakout among them, from screen pixels alone. ↩

6. AlphaGo is the Go program developed by DeepMind. In 2016 it beat world champion Lee Sedol 4-1, and in 2017 it defeated world No. 1 Ke Jie. ↩

7. AlphaStar is the AI DeepMind developed for StarCraft II. In 2019 it reached Grandmaster level on the game's ladder. ↩

8. For AlphaFold's breakthrough in protein structure prediction, Hassabis shared the 2024 Nobel Prize in Chemistry (not Medicine) with John Jumper and David Baker. Source: the official Nobel Prize website, nobelprize.org. ↩

9. Foldit is a puzzle-style citizen science game launched by the University of Washington in 2008 that lets players help research by folding protein structures. It's a separate project from DeepMind's own AlphaFold. ↩

10. "Attention Is All You Need" is a 2017 paper by Vaswani and other Google researchers that introduced the Transformer architecture, the foundation of today's large language models. arXiv:1706.03762. ↩