This site is my personal science blog, dedicated to making complex things simple:

• Blockchain — consensus, wallets, smart contracts, explained in plain language
• Web3 × Real Economy — RWA, NFT ecosystems, the bridge between on-chain and the physical world
• AI — large models, agents, and how AI truly lands in enterprises
• Essays — occasional industry observations and reflections

My writing principle: give the conclusion first, then the principles, then a one-line summary. No jargon piles, no mystification.

If there’s any concept you’d like me to explain clearly, leave a comment or let me know — I’ll work it in.

A blockchain is essentially a public, decentralized, extremely hard-to-tamper ledger. Think of it as a ledger book placed in a town square: everyone holds a copy; if someone wants to record an entry, most people must approve it, and once recorded, secretly changing it is nearly impossible.

Understanding it through everyday life

Imagine a village’s bookkeeping: Traditionally the village chief kept the books alone — whatever he wrote stood. That’s centralization. If the chief erred, or was bribed to alter records, no one could challenge him.

Blockchain changes the game:

1. Everyone has a copy of the ledger — one posted at the bookkeeping office door, one in every household’s drawer.
2. Recording must be public — to record “Zhang gave Li 10 yuan,” you must announce it; the whole village checks, and only after confirming does each write it in their own copy.
3. Pages are bound in order — the bottom of each page prints the “fingerprint” (hash) of the previous page. Try to alter one page and every page after it stops matching — you’re caught instantly.

These three together are the core mechanism of blockchain: distributed ledger + consensus + hash chain.

What problem it solves

Traditional trust relies on “authority endorsement” (banks, governments, platforms). Blockchain moves trust onto math and rules. As long as most participants are honest, the ledger is trustworthy — no need to know who the village chief is.

The cost is obvious too: slow, expensive, redundant storage. So it’s unsuited to high-frequency micro-records like “bought a bottle of water at the supermarket today,” but ideal for records needing strong trust: asset ownership, contracts, provenance.

In one line

A blockchain = a public ledger anyone can read but no one can secretly alter. It isn’t magic; what’s magic is that it made “strangers collaborating without trusting each other” possible at scale for the first time.

RWA (Real World Assets — tokenization of real-world assets) uses blockchain to turn real-world assets like property, gold, bonds, even receivables into on-chain, fractional, tradable, programmable digital certificates (tokens). In one line: it lets real-world assets flow like stocks, at lower cost and across wider borders.

Understanding it through everyday life

Imagine “digitizing” a ¥5M apartment into 5 million tokens:

• You don’t have to sell it whole: need ¥100k? Sell 100k tokens — no need to transfer the whole property.
• Global investors can all buy: buying a Shanghai apartment used to mean cross-border wires, title transfers, hefty agent fees; now a few clicks and you hold it.
• Everything recorded on-chain: who holds it, how many times it traded, how dividends split — public, transparent, tamper-proof.

That’s the core imagination of RWA: turning assets that were “illiquid, high-barrier, opaque” in traditional finance into “liquid, low-barrier, fully transparent” assets.

Why it’s hot only now

RWA isn’t new (“asset securitization” has long existed), but three conditions only matured recently:

1. Compliance channels opened: the US SEC approved tokenized funds (e.g. BlackRock BUIDL); frameworks are landing.
2. Stablecoins became the “bridge”: on-chain, dollar-pegged stablecoins give assets a unit of account and settlement rail.
3. On-chain infrastructure matured: custody, auditing, and bringing off-chain data on-chain (oracles) gradually closed the trust gap of “how an on-chain certificate maps to off-chain physical assets.”

Typical scenarios

Three real-world constraints to remember

1. The trust gap of “on-chain certificate ≠ off-chain asset”: needs custodians, audits, and legal frameworks as backstops — this is RWA’s hardest engineering problem. Not tech-hard, trust-hard.
2. Regulation is just starting: countries haven’t unified how they define security tokens (STO); compliance cost isn’t low.
3. Not every asset fits: standardized, verifiable, easily-valued assets go first; “digitizing a painting” is more marketing narrative.

In one line

RWA = giving real-world assets a “digital ID + fractional stock,” letting them flow — the direction is set, the hard part is compliance and trust; the first to work will be standardized assets like government bonds and gold.

A large model (like the GPT behind ChatGPT) is essentially a machine extremely good at playing “word chain”: give it an opening, it predicts the most likely next word based on “experience,” then the next… until a full answer forms. Its “experience” comes from the massive human text it has read.

Understanding it through everyday life

Imagine a child who read a hundred thousand books. You give them an opener: “The weather is nice today, let’s go to the park,” and they’ll most likely continue with “for a walk” or “to play” — because that’s the most common pairing in the books they’ve read.

A large model does exactly the same — just at staggering scale:

• The books it read: nearly the entire public internet (Wikipedia, papers, books, code…).
• Its “word-chain” skill: not rote memorization, but learning the statistical patterns and deep structures between words.
• Parameters: hundreds of billions of tunable knobs that together decide how the “next word” is chosen.

Why it seems to “understand”

Play word-chain well enough and the illusion of “understanding” emerges — no, more precisely, genuine capability emerges: it can translate, write code, reason. Like a Go AI that only learned “which move wins” but ended up “knowing” Go.

But it has three essential limitations:

1. It confidently makes things up (hallucination): it is never responsible for “facts,” only for “plausibility.”
2. Its knowledge has a cutoff date: it only knows what it read, not what happened after.
3. It doesn’t truly “think”: no goals, no intent — it just computes the most reasonable continuation given an opener.

In one line

A large model = a machine that read the entire web, pushed word-guessing to the extreme, and thereby gave rise to understanding and creativity. Used well, it’s productivity; but know this: it never guarantees what it says is true.