4 ideas
A blockchain is a slow, append-only ledger that buys resistance to censorship by paying for it with enormous inefficiency. If a use case already has a trusted party, like a company, regulator or consortium, an ordinary database does the same job faster and cheaper. The only thing the blockchain uniquely adds is removing trust in a central party, and almost no enterprise use case actually needs that.
Read crypto history as libertarians rebuilding an unregulated financial system and then painfully rediscovering, one by one, why each regulation exists. Exchange collapses, Ponzi schemes, pump-and-dumps and unregistered securities offerings are the same frauds that securities, banking and consumer-protection law were written to prevent. Removing intermediaries did not remove the need for trust. It just removed any recourse when that trust was betrayed.
Mt. Gox began as a trading-card exchange and became the dominant Bitcoin exchange. It was run with amateurish code and no auditing, and it held customers' coins in commingled wallets. When it collapsed in 2014, around 850,000 bitcoins were missing after years of undetected theft and mismanagement. The case shows that 'be your own bank' in practice meant handing custody to unaccountable operators who functioned as unregulated, uninsured banks.
Smart contracts put agreements into immutable code that executes automatically. But real contracts depend on human judgement to interpret intent and fix mistakes, and any bug in immutable code is a permanent, exploitable hole. The DAO hack showed this. When an attacker drained funds through behaviour the code allowed, Ethereum hard-forked to reverse it, which conceded that 'code is law' was abandoned the moment real money was at stake.