extra integrations


i wonder if there’s a relationship between likelihood of abandoning instapaper and queue size or queue-to-read ratio. not that i would ever overload a to-do list on day four and give up on it.* i’m sure they already know it, but it’d be interesting to read about!

*i have done that literally every time i started a to-do list. it happened last month


Hi All,
I never tried instapaper as pocket works really well for me. Is it better than pocket?
Since I starting this project my trello todo list shrunk substantially and now I’m experimenting with exponential weights on cards ages like @philip suggested. There is also couple of more metrics can collect from pocket such as total word count and total word ages in days.
I also made some changed to the UI that might make it easier to use.
Let me know how it works for you.


I’ve been using the trello neglect metric for about a week now –

I noticed today that the integration is sending over data hourly, and the count seems to be growing throughout the day. I find that surprising. I did add two new cards today, but the count has creeped upward from 1024 to 1040 over the the course of the day – perhaps it’s counting days from the absolute time the card was last updated, as opposed to the day? It would probably be nice though if that number mostly didn’t increase at random times throughout the day, by rounding all the edit timestamps to midnight…

But it’s got me plucking out some of the oldest cards to get them cleared out! So it’s definitely helping me work on the neglect problem. I’m thinking of implementing the same for my email…


Interesting. I’ve written my manual version of something similar to deliberately use the card-update-time, exactly so that it increases at random times throughout the day. Ever more reason to get back on the road earlier in the day, because I never quite know when the value is going to jump…


Yes, The algotithm counts seconds and devides by the number of seconds in
a day. This leads to gradual increases throughout the day. It isn’t obvious
what is going on but the alternative might be even more surprising. Say you
add 20 cards at 11pm. Then each day will see 20 points bump at that time.
The current way of extrapolating seem to be less surprising. Maybe i should
round it with a digit or two of fraction so the linearity will be visible.


Another question: is it possible to have more than one Trello neglect goal? I just did what seemed like the obvious thing to set up a second one tracking a different board, and it seems to have just changed the config for the existing one.

New mobile app (Android, iOS, Windows), plus 7 new integrations!

The UI doesn’t support it at the moment. I was afraid it might add some complexity to an app that might be confusing enough already. As a workaround you can use the exponential version with 1.0 as exponent, that makes equivalent to the reuglar version.


If anyone is interested, I’ve added an option to track the average age of the cards instead of the sum.
This method seem to be more reasonable as having many cards is not necessarily bad.


I still love the original trello neglect metric of just sum of ages in days. I’m game to try this (average age instead of total) but here are my worries:

  1. Adding new cards will bring the average age down.
  2. Minding the mean makes it harder to dispatch a beemergency by focusing on one or two of the oldest cards.

Imagine a flat yellow brick road that you were right on the edge of yesterday (my motto: consider the akratic steady state). Then today the average age will increase by 1 day (+1 on all cards) and you’ll be 1 above the road…

… Wait, now I’m wondering if #2 is just wrong, and maybe #1 is fine if adding lame-o cards to bring down the average is not a loophole you’d abuse. Maybe adding more cards is even something you’d like to add incentive for.

Ok, so maybe simply minding the average age is best after all! Eager to hear what others think!

(Being a math nerd I’m now also wondering about things like geometric instead of arithmetic mean, which might incentivize greater focus on older cards. And of course then you can generalize that with a parameter…


(I had that backwards. Geometric mean would skew the mean lower. What we might want is higher power means, like root-mean-square which is a power mean with p=2. And in the extreme, p=infinity means simply minding the max age.)


Sounds like an interesting experiment. You can try it here,


I do this for my personal backlog-scorer scripts. I quantize the time per item into a customizable unit, and then have Ax^Y, where A and Y are configurable, and then sum the results into a score.

(I am glad I am not the only constant-happy backlogger out there.)


Sweet! I think the term is just “power mean”. And might be cool to have an “infinity” option for minding the max. And maybe worth highlighting the special cases like p=1 is the normal average, p=2 is root-mean-square, p=0 is the geometric mean.

I did think of one advantage of minding the simple sum of ages: If I see the “last modified” date I can (sort of) tell in my head whether a card will be enough to dispatch my beemergency. Minding the mean that’s much more opaque.


So there is no chance of fixing the Anki plugin or integrating it to


I couldn’t find API documentation for Anki so I have no reasonable way to integrate with them.


The traditional integration was from anki’s desktop client, rather than to ankiweb.

Related: anki sync broken


@galtsubery since the anki sync worked as an addon from anki’s desktop, is there a chance to make a new one that actually works for everyone and integrate it to I’m willing to pay for this.



Sorry, I can’t accept more work at the moment. Anyway, my experience with Python is limited so it wouldn’t be good use of your money to have me work on it.


The desktop plugin accesses the local database, not the one in the ankiweb cloud, so it’s not as straightforward as that… :frowning2:


The anki project on github has a source file which sounds like a good place to start if anyone wants to write something against it. Sounds almost as good as an API to me :slight_smile:

Disclaimer: I’m not currently into Anki, but I may if someone tells me it is kickass at Japanese.