One of the major skills / talents of seasoned instructors (and some new ones) relates to their live "run-time adaptation." This computer term refers to the operation of a computer program. As applied to instructors, this relates to how an instructor leads and supports a group of learners. This involves a fair amount of complex multi-tasking and the nuances of reading human behavior and meaning (verbal and non-verbal). This run-time adaptation also involves a deep body of knowledge about a particular field and then teaching and learning. In instructor-led online learning, with human-intelligence mediation of the learning, the instructor actually puts together and customizes the learning for the individual learners. The instructor collects real-time information about the learners and makes adjustments accordingly. The instructor may have also been the multimedia author and SME as well. There exists plenty of research on how various software developers are trying to build this personalization / customization piece into automated eLearning software. There are various multi-agent scenarios with "adaptive agents" functioning to create "personalization". There are uses of learner profiles. Some use learner-directed learning. Others use metadata-labeled digital learning objects with filters and agents culling the materials and then delivering these in original ways. Others use recommender systems. Such technologies move towards replacing the instructor's instincts and experiences with formulas and algorithms. So far, what I've seen has been quite basic. However, at some point, with the proper instructional design and maintenance of the contents, and more nuanced software development techniques, could this individuation of learners and the unique meeting of their needs be achieved totally by "thinking machines"?