Hacker News Reader

The scary thing is the agent that could read the bank account information and the agent that posted the information were separate agents.

In what sense "separate" though? No shared memory? No IPC? No what?

No-one can ever deterministically test any of this - just expect the worst to happen.

"Don't write any bugs"

Not separate enough, it seems.

The failure IMO is trusting anything that has not proven itself first with a year or three of experience.

US privacy laws - has anyone gone to jail or massive impacting fines for violations?