This chapter is the point of the book. Everything before it was mechanism; this is judgement, and it is the part that makes you useful in a room where someone is proposing one.

Enterprise blockchain has now had roughly a decade and a large budget. The results are public, and they are consistent enough to be worth stating plainly.

• TradeLens — Maersk and IBM's shipping platform, shut down in 2023. Its own post-mortem said it could not reach the level of industry collaboration required. Note that this is a failure of question 1, not of engineering: competitors would not co-write.
• we.trade — a consortium of twelve European banks, ceased trading 2022.
• Four of five major trade-finance consortiums have closed. This is a pattern, not a run of bad luck.
• Nasdaq Linq, the 2015 private-market share issuance pilot — widely cited for years afterwards, never scaled into anything.

Set against that, the things that did work: stablecoins, which move very large volumes for cross-border value transfer, and where the honest comparison is against correspondent banking rather than against a database; tokenised money-market funds, where BlackRock's BUIDL and similar products have reached real scale; and JPMorgan's Kinexys (previously Onyx), a permissioned network handling meaningful daily volume in intraday repurchase agreements and cross-border payments.

What those three have in common is instructive. Each moves value between institutions that genuinely do not have a shared trusted operator, on a settlement problem where the existing process is slow because of intermediaries rather than because of compute. They pass all four questions. Most proposals do not.

India's central bank digital currency, the digital rupee (e₹), is a live pilot and its numbers are published, which makes it a rare chance to see what adoption looks like without marketing.

• The wholesale pilot — for settling government securities between banks — collapsed by more than 99% from its early volumes and RBI later paused adding new wholesale use cases.
• The retail pilot reached about 0.003% of currency in circulation.
• The widely reported ‘one million transactions a day’ milestone in December 2023 was reached after banks began paying part of employee benefits directly into CBDC wallets. Volume fell roughly 90% once that stopped. The milestone measured the incentive, not the demand.

Not dismiss it. A technology can be badly oversold and still be the right answer for a narrow set of problems — and being able to say which narrow set, with the record to back it, is a far more valuable position than either enthusiasm or cynicism.

• If someone proposes one, walk the four questions. Ask who the distrusting writers are, by name. If the answer is departments of the same company, you have your answer without a single technical argument.
• If you are asked to build one, the security chapter is your checklist: multi-signature with signers on separate hardware, timelocks on privileged actions, aggregated oracles, OpenZeppelin rather than hand-rolled, fuzzing, and an explicit written answer to who can repoint the proxy.
• If you are considering it as a career direction, re-read the note at the top of chapter 1 — the corpus behind this site mentions it zero times in 4,556 job descriptions. Learn it because the mechanisms are genuinely worth understanding and it makes you the person who can evaluate a proposal. Not because the hiring evidence supports it, because it does not.

For the banking-specific record — the digital rupee's own published figures, what atomic settlement actually removes, and why four of five trade-finance consortiums died — see three chapters in the other book: The digital rupee, and what its own numbers say, Tokenisation, and the one thread that did not collapse, and The consortium graveyard.

The strongest thing this subject teaches is not the cryptography. It is that coordination problems are usually governance problems wearing a technical costume — and that recognising the difference is a rarer skill than implementing either.