Effort and Opportunity

A college classmate I hadn’t seen in over a decade came to Beijing on business, and we met for dinner that evening.

By the end of the meal—glasses raised, stories flowing—he shared several experiences that stuck with me.

He now serves in a local government role overseeing economics and technology, reporting directly to the city mayor. The pressure is real: his two departments carry roughly 70% of the city’s GDP targets.

He showed me his WeChat log of daily work reports—sent individually to both the mayor and the party secretary—dating back to the day he took office. He hasn’t missed a single day.

I asked why he does it. “First, it is important,” he said. “Second, if either leader gets asked by a higher-up, they can answer instantly—with up-to-date facts.” Alongside progress updates, he routinely flags key issues and outlines concrete solutions.

If I were his boss, I’d want exactly this kind of subordinate.

When the top two leaders in a region are aligned—not just in words but in action—the place moves fast. Internal friction drops. He described his mayor and party secretary as precisely that: united, pragmatic, and execution-focused. Working under them, effort reliably translates into visible results—and results open doors. At this pace, he’ll likely become deputy mayor within a few years.

“How do you make those results?” I asked.

“Treat yourself as a true servant,” he replied.

His main job is investment promotion—bringing quality enterprises into the region. But attracting them is only step one. What matters is delivering real support after they arrive. Everything else is noise.

So when he visits companies, he skips small talk. He asks: What’s blocking you? What do you need? How can we solve it—together?

That mindset turned many entrepreneurs into close friends.

He also recalled our brief time as dormmates—just a few months—but said it left a lasting impression. While others played games or watched movies after class, I was teaching part-time and diving deep into technical research. That intensity fascinated him.

Later, he spotted an emerging tech field with promise—and spent half a year teaching himself its fundamentals. That self-driven learning gave him an edge during campus recruitment. One year, a Xiamen-based company held a notoriously difficult coding test. Almost no one completed it. He told the interviewers, “I’m not from this major—but I’ve studied it independently. I can do it.” In 90 minutes, he finished the assignment—and landed the job.

Within months, he became a core technical contributor, earning a strong salary. He even designed a chip component used in the Xiaomi Mi 3 smartphone—his name etched onto the silicon.

A few years later, he walked away from that high-paying corporate role to enter public service—and has risen steadily ever since.

Learning. Opportunity. Effort. Rigor. Resources. Timing. Put them together, and light emerges.

AI’s “Overuse” Principle

Tokens have become absurdly cheap—practically free. So go ahead: generate freely. If a task can be handled by tokens, don’t default to human labor.

Here’s a working principle: overuse tokens. Even if it takes 100× more tokens to gain just 1% efficiency, it’s worth it. Why? Because it trains your brain to think natively with AI.

“Overuse” has another benefit: it surfaces invisible problems—things you didn’t know you didn’t know. It reveals odd opportunities. Sometimes, brute-force token volume does yield breakthroughs.

If a production factor improves 10× in cost-efficiency per year, it’s time to adopt it at scale.

Under those conditions, AI applications won’t trickle—they’ll explode like spring bamboo shoots.

When price and entry barriers fall this far, markets we once dismissed as “not yet viable” suddenly deserve serious attention.

Four Tiers of AI Adoption

AI isn’t one thing—it integrates at different levels:

  1. AI for the general public: AI + search → AI as a replacement for traditional search.
  2. AI for individual contributors: AI + tools → AI as a productivity amplifier (e.g., writing, coding, analysis).
  3. AI for managers/leaders: AI + roles → AI as a digital employee, handling discrete responsibilities (e.g., customer onboarding, compliance checks).
  4. AI for executives/strategists: AI + business functions → AI as a team of digital employees, orchestrating cross-functional workflows (e.g., end-to-end supply chain optimization, dynamic pricing engines).

Over dinner, I shared this framework. I said: “My goal for next year is to have 100 AI employees working for me daily”—meaning autonomous agents performing full, recurring tasks like human staff. Right now? I have three. That leaves 97 to build.

Second Half-Marathon

Today I finished my second half-marathon of the year—in 2 hours 13 minutes. That’s 14 minutes faster than my first. Hoping for more improvement next time.

I started running in June—four months ago. A few insights stand out:

  1. Running slow is harder than running fast—especially early on. You unconsciously speed up.
  2. Sustaining that slow pace is even harder. Here, “slow” means staying within your aerobic heart-rate zone—around 140 bpm for many middle-aged runners.
  3. Harder still is maintaining consistency: running every other day, without exception.

But once those three habits lock in, something shifts. You begin to enjoy the run itself—regardless of weather—and start reaping tangible benefits: better mood, deeper sleep, improved stamina, sharper focus.

IP First, Then Digital Twin

Douyin recently launched a digital avatar feature—and usage is surging.

I heard about a fitness coach (a modestly followed creator) who launched a subscription service: via WhatsApp, he sends daily check-ins, answers questions, and nudges users to complete their workouts—charged annually as a “health coaching” package.

That entire service? Now fully automatable with AI avatars.

But trying to build an IP using only AI avatars? That doesn’t work.

Why? Three reasons come to mind:

  1. At its core, IP is about trust—earned through authenticity, consistency, and human resonance.
  2. IP carries significant premium power—and while AI can reflect that premium, it doesn’t create it.
  3. An AI avatar can host an IP—but only if the underlying identity is already trusted. When users interact, they need to feel recognized, not processed. That requires more than mimicry: it demands intentional, human-informed design.

To deliver real value and engagement, copying a person into an AI agent isn’t enough. You need structure—grounded, deliberate, and deliberately not AI-idealized.

A Key Cause of Poor Decisions

In the years since China’s “Double Reduction” policy, many high-level decisions—looking back—contain glaring flaws and miscalculations.

One root cause stands out: treating uncertain situations as if they were certain.

That false certainty warps expectations. Reality diverges. And divergence breeds error.

The consequence? Losses. Failures. Wasted resources.