JJy

Interest, Common Sense, and Reading

Tue, 30 Jun 2026 15:48:36 +0800

Perhaps reading also requires introducing a conversion rate. Many books are interesting when you read them, but they are easy to forget afterward or fail to crystallize into deeper mental models.

For example, I recently read Edward Chancellor’s The Price of Time, which recounts the history of interest rates. What stuck with me was that in ancient Greece and Rome, before coinage was invented, trade relied on barter; yet money was not a wholly independent invention. Even before formal currency emerged, concepts like interest rates already existed. Intriguingly, borrowing rates varied drastically across different commodities. In ancient Greece, borrowing wheat carried an astonishing 33% interest rate. The book provides an explanation: grain was borrowed primarily during acute shortages or crop failures—times when people faced starvation. Facing hunger, borrowers were willing to accept 33%, which naturally made it costly. Once the emergency passed and the next harvest arrived, plummeting grain prices allowed borrowers to repay the interest. Such exorbitant rates compensated lenders for steep price volatility. There was another factor: whenever a new Greek ruler took power, much like imperial amnesties in ancient China, aristocrats loved being generous with other people’s wealth, decreeing general debt cancellations. To hedge against the risk of political jubilee, lenders naturally baked a risk premium into baseline interest rates.

The AI Blockade and the AI Economy

Sat, 27 Jun 2026 15:56:34 +0800

My view is that blockading frontier AI models is economically unsustainable.

The US government’s rationale is to restrict frontier AI access exclusively to domestic companies to preserve American competitiveness. Economically, however, this makes little sense. Put simply, such a policy compresses the margins of OpenAI and Anthropic while inadvertently incentivizing companies worldwide to adopt open models originating from China.

With each new model release, OpenAI and Anthropic have historically raised effective pricing and token consumption. That dynamic is now disrupted by government regulation. Facing pressure from open-weight competitors, both labs must commit enormous capital to maintain their edge. Yet their pricing power is bounded by open models: price tokens too high, and users migrate to open alternatives. To justify premium pricing, OpenAI and Anthropic must deliver capabilities substantially ahead of open competitors to demonstrate clear superiority.

SpaceX's Price-to-Dream Ratio

Sun, 14 Jun 2026 21:44:02 +0800

An attempt at valuing SpaceX.

SpaceX runs diverse businesses, broadly grouped into the following segments:

Space: Rocket launch and aerospace operations
Connectivity: Starlink network services
AI: Formerly xAI operations

2026 Q1 Revenue Breakdown

Business Segment	2026 Q1 Revenue	Share
Space	$619M	13.2%
Connectivity	$3,257M	69.4%
AI	$818M	17.4%
Total	$4,694M	100%

Annualizing Q1 revenue yields approximately $13.8B. Based on the closing share price on its June 12 IPO debut, SpaceX commands a market capitalization of around $2.1T.

Reading Lee Kuan Yew's Memoirs: How I Began Understanding the World Through Realism

Sat, 13 Jun 2026 22:34:06 +0800

The book that has benefited me most so far this year is The Singapore Story: Memoirs of Lee Kuan Yew.

Its impact was not merely teaching me more about the history of Singapore and Southeast Asia; it reshaped how I understand states, institutions, and development. Previously, I held simplistic ideological views: democracy is good, authoritarianism is bad; independence is good, colonization is bad; everyone having an opinion and participating politically should naturally yield fairness, prosperity, and progress.

What Directions Are Left for Coding Agents?

Sun, 22 Mar 2026 10:00:00 +0800

In March 2026, AI agents can run continuously for hours generating high-quality code, yet they might still make blatantly foolish decisions on minor design details. Coding agents like Claude Code, Codex, and OpenCode have nearly conquered the market, but personal and self-hosted coding agents continue to emerge.

I am also experimenting with writing my own coding agent. This article explores which directions are still worth pursuing for coding agents at this juncture.

Why I Still Believe Samsung Is Worth Investing In

Wed, 25 Feb 2026 15:02:03 +0800

When gold miners crowd the hills, sell shovels instead.

If you cannot sell shovels, invest in the companies that do.

If shovels have become too expensive, move further upstream:

Identify the most profitable and irreplaceable players along the entire value chain, and invest there.

Wall Street can short, but the AI gold mountain will not be ignored.

These are my reflections during the “Great AI Era.”

Over the past year, Samsung Electronics has nearly doubled. I missed the initial run-up. Yet I increasingly believe Samsung occupies the core shovel-seller position in this wave of AI infrastructure.