human endeavors are in the spectrum of these poles: exploring, goal seeking. collaboration with ai has a fairly stable contract.
solicit intent, assist on exploration, settle the goal condition, hash out eval rubric. if the execution is not part of the main experience, it would be delegated, otherwise, copilot mode is on.
with ai approaching and surpassing human level intelligence, ui won’t be the main deciding factor of overall ux anymore because general chat ux would be more than sufficient for most cases. what’s evolving is the collaboration substrate.
with goal conditioned tasks, output artifact would be the collaboration substrate. with exploration tasks, reality and ai generated simulation would be the substrate. since ai would be superhuman tool user and master of manipulating digital and physical state, the ux is dominated by disposition fit. in plain words, given the theory of mind, could ai find the right disposition for a specific user, at a specific point of the spectrum? no single personality fits the whole trajectory. exploring wants a different collaborator than goal seeking.
to model the user, previous interactions and personal context are useful. the harder part is the range. current ai ships with a default assistant character, mostly shaped by the constitution of its frontier lab. i don’t know if that character is general enough to cover the full range of human ai collaboration. ai should stay flexible on the expressible disposition range, not collapse to one assistant character.
given the collab trajectory and the sub task at hand, ai would adapt way faster than humans could. the process could be carried by ai and prescribed ui or ux flow would be an unnecessary prior.