Most of these were eventually not offered because we could not find an interested client. Many of them are either fun or feasible, and we're happy to consider using them in future.

2016

Message to the future

There are so many things that only happen once a year - birthdays, start of term, insurance renewals - if only your email system could remember what you did last year, and take care of it next time! Your job is to implement an email server @nextyear, that will interpret the content of any message it receives, work out which things will have to change next year, and then draft an automatic update. It would be safer if this went to the user for checking a few days in advance, and provided a web interface to review currently active processes, but everything should be achieved with the absolute minimum of mouse clicks or wasted user attention

Message to the future

There are so many things that only happen once a year - birthdays, start of term, insurance renewals - if only your email system could remember what you did last year, and take care of it next time! Your job is to implement an email server @nextyear, that will interpret the content of any message it receives, work out which things will have to change next year, and then draft an automatic update. It would be safer if this went to the user for checking a few days in advance, and provided a web interface to review currently active processes, but everything should be achieved with the absolute minimum of mouse clicks or wasted user attention

Liker-bot

Modern teenagers have to spend hours every day pressing like buttons on Facebook and Instagram. Your task is to automate this drudge-work. Of course, these companies really want to trap eyeballs, so they won't make it easy for you. You will probably have to emulate a web browser, which at least pretends to be looking at the posts before automatically "clicking" on the like-link. It will also be embarrassing if it likes the wrong thing, so some machine learning will be involved to monitor, learn and then anticipate the user's real preferences (kittens = like, parents = dislike etc).

2015

==