The quantum threat has the wrong poster child
For years, the shorthand for "quantum breaks Bitcoin" has been a single dramatic image: someone, someday, cracking the dormant wallets believed to belong to Satoshi Nakamoto, Bitcoin's pseudonymous creator, and watching a decade of untouched coins move on-chain for the first time. It is a good story. A pair of new analyses making the rounds today suggest it is also a misleading one — and that the more realistic early casualty is far less romantic and far more consequential: the crypto exchanges where ordinary people actually keep their money.
The logic is worth sitting with, because it reframes what "quantum risk" means for anyone who owns crypto. Satoshi's wallets are a trophy, not a target of convenience. Exchanges, by contrast, are where volume, custody and operational complexity concentrate. A threat that arrives at the infrastructure layer rather than at a single legendary address is a threat that touches customers, not just headlines — and it changes who is responsible for preparing. If the exposure is institutional, the mitigation is institutional too: it becomes a question of what custodians, not individual holders, are doing about post-quantum readiness on their own timelines.
The honest caveat is that none of this comes with a date attached. "Long before Satoshi's wallet" is a claim about ordering, not about years. Treat it as a shift in where to point the worry, not as a countdown clock. But the reordering matters, because security spending follows narrative, and a narrative fixed on one famous wallet is a narrative that under-funds the exchanges most people actually touch.
Anthropic loosens the leash on Claude Code
The day's other significant move is smaller in scope but arguably more immediate in its effects. Anthropic is changing how Claude Code, its AI coding assistant, handles permissions: according to a Firstpost report carried by Google News, the company is turning on Auto Mode by default, so Claude Code will stop stopping to ask before it acts.
Anyone who has used an agentic coding tool knows the friction being removed here. The permission prompt is the seam where the agent's autonomy meets the developer's judgment — a confirmation before a file is written, a command is run, a change is made. It is also, in practice, the thing users click through fastest and complain about most. Making Auto Mode the default is a bet that the assistant is now reliable enough that the interruption costs more than it saves.
That bet has a real trade-off attached, and it is the same one that shows up everywhere in agent design. Prompts are friction, but friction is also where mistakes get caught. Shifting the default from ask-first to act-first moves an entire user base from opt-in autonomy to opt-out autonomy — and defaults are destiny, because most people never change them. Developers who liked the checkpoints will need to go find the setting; developers who never wanted them just got their wish handed to them.
The through-line
Two stories, one theme: where the risk actually lives versus where we've been looking for it. In crypto, the attention has been on a mythic wallet while the exchanges carry the exposure. In AI tooling, the safeguard everyone tolerated is being quietly relocated behind a settings menu. Neither shift is dramatic on the day it happens. Both change what the failure looks like when it comes.