One thing I think could be improved is how the schedule adapts when plans change during the day. My schedule isn't always predictable, especially when I'm doing research, since experiments, writing, or meetings can sometimes take much longer than expected.
It would be really helpful if unfinished tasks could be rescheduled more intelligently based on the remaining time and priorities, instead of requiring me to manually reorganize everything.
I also think it would be great if TimeLift could learn from my actual completion times over time. For example, if I usually spend 1.5 hours on a type of task that I initially estimated at 1 hour, it could use that history to suggest more realistic time estimates in the future.
I think these kinds of adaptive features would make TimeLift even more useful for people whose schedules change frequently.