Researchers can pursue whatever projects they want, subject to the constraint that every project must make exciting demoable progress every X weeks (e.g. 2). A project which ends up not demoable after X weeks must be shelved for at least X weeks.
The only exception is for “infrastructure” projects, which are allowed to consume more time if they expand the set of projects which can be achieved within those X week cycles.
Conversation with David Holz, 2020-01-01