The right to skip your interface

Aslan Askarov
2026-09-08

There is a peculiar expectation in the software I use for teaching, grading, and fundraising: having decided what needs to be done, I must then explain that decision to a sequence of dropdown menus. Why must I click them in the first place?

In their 2021 EFF paper, Bennett Cyphers and Cory Doctorow describe delegability: users should be able to authorize software to interact with a platform on their behalf. “Every substantial part of the user interface should be available to automated access,” they write. This requires stable programmatic access on which other tools can depend. [Privacy Without Monopoly]

Their original argument, articulated five years ago, concerns dominant platforms. With agents now able to act on our behalf, it has acquired a new urgency for enterprise and institutional systems. I may have no influence over which software an academic organization or a funding agency uses. It does not follow that I should have no choice about how to enter data into it.

I want to guide my agent, inspect the proposed changes, and authorize their submission, without having to run a headless browser. The system can decide how to present the result to everyone else.

Of course, delegation needs limits. Authority to enter grades does not imply authority to disclose student records to whichever service happens to run the agent. Information security must be enforced, and changes must be accountable. None of this is guaranteed by requiring me to navigate five screens.

The vendor cannot anticipate every reasonable way I might organize my work. A documented API would let me supply that knowledge myself.

Just give me an API endpoint.