I floated this in a daily beemail and got a lot of thoughtful replies and ideas and feedback.  Daniel Tedesco made the sage suggestion that we were juggling too many priorities so we've been trying to ruthlessly move things from "we need to do this right now" to "we'll consider this after the actual top priorities".
I've marked top 10 priorities with little 's and 's (for things we as users really care about and things we have to care about to keep Beeminder viable, respectively) and put 's on things that are not yet priorities or that are waiting on other things to be done.
For the priorities I identified a plausible first step in parentheses. I'm hoping that by starting with all of those first steps that we'll be able to keep up momentum along multiple important dimensions (growing, bug fixing, newbee-friendliness, features we power users are desperate for and that keep Beeminder exciting for us, infrastructure & long-term planning).
What do you think? I'm making this a poll, approval voting style, so vote for your favorites, or vote for everything except the ones you don't think we should care about:
- 1. Upgrade all the things (Ruby 2.x and MongoMapper->Mongoid underway)
- 2. Keep acquiring users (email capture on front page)
- 3. Lifecycle emails (better unsubscribe controls)
- 4. Fuzzy Buzzy and other premium plan experiments (Chelsea's 2-tier plan)
- 5. Support Paypal (support Alipay, since Stripe makes that super easy)
- 6. More autodata integrations (MFP if possible or Khan Academy which looks easy)
- 7. Road Editor and concomitant swarm of end-date bugs (auxiliary road editor via API)
- 8. Death to Freebees aka Freebees für Alles (JFDI)
- 9. Open-sourcing autodata integrations (zeno callback for beemind.me et al)
- 10. UI simplification/streamline/redesign, especially mobile-friendliness of goal creation & settings (settings revamp)
Goal creation wizard (goal cloning? after death to freebees)
Sponsor behavioral econ conference
Uncle Button (maybe important given how hard we make it to turn off zeno polling)
Tags and sorting options
Client-side graphs (after tini/vini, sadreset, auto-wide which we'll do as part of Road Editor)
True breaks (after Road Editor) 
Yellow Brick Halfplane (after Road Editor)
The ridiculous do-less loophole (after Yellow Brick Halfplane?)
Find/invite friends on Facebook
Follow friends, group goals
Integery and timey-wimey goals (post)
[UPDATE: Eep, we seem to have broken the tallies by editing the items in the poll. Sad face! Discourse's @zogstrip is kindly investigating for us. UPDATE 2: Looks like using 2 different list styles (bullets and numbers) confused it and there's no way to recover the true votes. Double sad face!]
And of course reply with other ideas we missed above!
Other things it may be bad to neglect till after 2016
- More useful stats in sidebar like actual average rate based on data, and streak tally
- Pledges in local currency
- More sophisticated deadmanning
- Account deletion
- IFTTT/Zapier channel improvements (zeno trigger, backdate data with polling freq window of deadline)
- Make the rest of the autodata integrations support arbitrary deadlines
- API version 2
 Thanks @bostonenginerd, @peppertoni, @dtedesco1, @apolyton, @nickwolf, @malcolm, @drtall, @insti, @anomalily, @chipmanaged, @yebyenw, @nepomuk, and @adamwolf!
 True Breaks is about having an actual gap in the road, for things like weight loss or inbox count where it's not enough to just have a flat spot. You want to be wholly unaccountable during a vacation and have the road resume at wherever your datapoint ends up at the end of the break.