I was thinking about a problem with the current beeminder structure where you can be disincentivized to make goals more fine-grained, because then your monetary liability is higher (since a “bad day” could mean derailing on, e.g. ten goals instead of three).
One idea I had is being able to opt-in to a global charge cap, so you can be confident you won’t be charged more than the cap on any given day (you could also combine this with a mechanism to ensure that you aren’t hit with seven days of at-cap charges, by allowing a user who has hit the charge cap on a particular day to also derail on any other goal even if it’s not a beemergency).
A downside to this setup is that it can turn your bad days into catastrophe days, in the sense that you might decide that once you get near/at the charge cap there’s no reason for you to put any further effort into anything beeminder-related for that day/week. But that’s a tradeoff that you’d be choosing by setting a charge cap, and might consider it worth it.
I’m sure there are other solutions to this problem; you can implement it today in userspace by putting “If I get charged more than $N in a day then I can add fake data to this goal” in the fine-print for every goal, but I like to avoid fake data as much as I can.
I suppose another structure that gets basically the same effect is, instead of a charge cap, you have a “pay $N to temporarily remove the akrasia horizons on all goals” feature (where the $N is user-selected, and changes to it are subject to the akrasia horizon).
I dunno, just a small pile of ideas.