every login, every task, every conversation accretes onto one permanent digital self. one browser, one set of sessions, one memory that never forgets. for people, that is a privacy tax. for autonomous agents, it is a liability: an agent that browses as you inherits your sessions, your logins, your history, your reach. we handed software our whole identity and called it convenience.
the danger is not that identity is stored. it is that identity is permanent and shared by default. one self, one context, everything linkable, nothing disposable. as agents act on our behalf, "operate as the user" becomes the default and the exposure. that default has to end.
identity should be disposable. it should be something you provision for a purpose, scope to exactly what it needs, and destroy when the work is done. not a profile you manage forever. a compartment you create and end. every task, every agent, gets its own: its own browser, memory, files, permissions, and a finite life. when it ends, it leaves a receipt and nothing else.
every guarantee mortal makes is labeled by what it actually enforces. what is real is marked enforced. what is intended is marked advisory or coming. mortal does not make you anonymous, and says so. the whole product is a refusal to lie about what isolation means.
mortal is a programmable identity runtime. local-first. no account. no telemetry. it launches real isolated browsers, not simulations. it destroys them on a schedule and proves it. it is built so an agent can be given an identity that is not yours, do real work inside it, and disappear.