3.3 Billion Tokens in a Week
This week, our AI application consumed over 3 billion tokens—more than ten times the previous level. I was deeply struck by the scale.
The sharp increase stems largely from official incentives—token subsidies that are unusually generous toward AI application developers.
This is a form of institutional support. Such policies accelerate real-world AI adoption across use cases.
Just recently, Baidu slashed the token cost for its ERNIE Bot by 90%—a highly credible move signaling strong commitment from major model platforms to push AI applications forward.
Subsidies work fast—like Didi’s early ride-hailing subsidies. They train users to reach for AI first when solving problems, turning theoretical capability into habitual practice.
For us, this incentive directly turbocharged token usage—and, more importantly, sharpened my own question: What other problems can AI solve here?
That habit of asking “Where else can AI help?” is itself a catalyst for deeper, more sustainable adoption.
Doing Hard Things
A senior healthcare executive messaged me recently: “Do hard things.”
That resonated deeply. Hard things are hard at the start—but once cracked, they tend to have little competition.
Building on that, I’ve clarified three guiding principles for how I’ll operate going forward:
First, leverage momentum and alliances. Partner with exceptional people or institutions—but only if you bring differentiated value. Without that, collaboration stays superficial.
Second, embrace AI proactively. Over the past year+, my hands-on experience has confirmed it: AI meaningfully lifts productivity and efficiency—not as hype, but as daily leverage.
Third, do hard things. That’s not just advice—it’s how he operates. Find a genuine pain point. Yes, it demands steep learning, solitude, and delayed validation. But precisely because it’s hard, few show up. That’s your edge.
Let’s hold each other to this.
Dynamic Focus
Distinguishing what’s urgent from what’s important is a foundational skill.
Each day brings a flood of tasks—some time-sensitive, others high-impact.
The rule remains: roughly 20% of tasks drive 80% of outcomes. So allocate ~80% of your energy there.
But crucially: what qualifies as “important” shifts constantly—with your stage, goals, and external conditions.
That’s why reflection isn’t optional. We must regularly reassess our work—not because we’re indecisive, but because reality is fluid. Our focus must be, too.
Dynamic focus isn’t just about saying “no.” It’s harder: continuously calibrating your current position, mapping your actual capabilities (not aspirations), and acting from that grounded clarity.
When Others Are Greedy, I’m Fearful
A-shares have been surging lately—so I revisited a classic.
Warren Buffett’s oft-quoted line—“Be fearful when others are greedy, and greedy when others are fearful”—is actually a distilled version. His original 1986 letter to Berkshire Hathaway shareholders is more precise and nuanced.
Here’s the full English passage:
What we do know, however, is that occasional outbreaks of those two super-contagious diseases, fear and greed, will forever occur in the investment community.
The timing of these epidemics will be unpredictable. And the market aberrations produced by them will be equally unpredictable, both as to duration and degree.
Therefore, we never try to anticipate the arrival or departure of either disease.
Our goal is more modest: we simply attempt to be fearful when others are greedy and to be greedy only when others are fearful.
The core insight isn’t bravado—it’s discipline. Not prediction, but posture. Not timing the market, but holding your ground while others swing wildly.
That same posture applies beyond markets: in product decisions, hiring, resource allocation—even in how we engage with AI hype. Stay anchored. Act deliberately.