What I Learned in China

Free Shanghai skyline, China photo

In early 2025, I took a 3 month (92 day) trip to Asia, visiting multiple countries – with China standing out as the most interesting1.

Screencap from polarsteps

I flew from Singapore (SIN) to Shanghai (PVG) on a $120 Spring Air flight, then exited through Busan (PUS). In China, I stayed mainly in Shanghai and Suzhou.

At my Shanghai hostel, I met a cool traveler who was doing a RTW trip before starting a finance job in Chicago. I also connected with a Shanghai local/business guy I originally vibed with in Jakarta at another event.

The vibe in Shanghai2 – and China broadly – felt like work hard: think 996 culture, eat-sleep-code-repeat, build-an-empire hard. Factories run all night, startups push updates daily, and engineers learn AI frameworks faster than we debate which one to use.

I wanted to visit Shenzhen – China’s Silicon Valley – but didn’t have time.

SV by MJ 🤣

In China, major US tech platforms like Google, YouTube, OpenAI, X/Twitter, Instagram, and Wikipedia are blocked so I had to use Baidu for search, WeChat Pay for payments, and Bilibili for video.

It’ll be interesting to see which model wins in the future: China’s authoritarian style or US/Western liberal democracy. History shows us that economic progress can stall without corresponding political progress e.g. Gorbachev’s policies of perestroika, and glasnost3.

From China, I learned that freedom of thought isn’t free.

The decision tree of choices grows exponentially, roughly \mathcal{O}(2^n).

Authoritarianism acts as a bounding function B that prunes branches, reducing the effective search space:

This pruning speeds convergence but risks trapping the system in local optima.

Liberal democracy maintains an unbounded, high-entropy search space with minimal pruning:
\text{Search space}_{liberal} \approx \mathcal{T}, \quad |\mathcal{T}| \sim \mathcal{O}(2^n),
enabling broader exploration but slower convergence.

The challenge is applying effective forcing functions F — like purpose or constraints — that guide navigation through this combinatorial explosion without paralysis:
F: \mathcal{T} \to \mathcal{T}', \quad |\mathcal{T}'| \ll |\mathcal{T}|,
turning exponential complexity into strategic advantage.

*Disclaimer: This blog post was written with the help of AI.

  1. Thanks to the 240 hour visa ↩︎
  2. https://paulgraham.com/cities.html ↩︎
  3. https://en.wikipedia.org/wiki/Perestroika ↩︎

Leave a Reply

Your email address will not be published. Required fields are marked *