About 18 months ago I made a post here on my workflow. I've received a handful of requests for follow-up, so I thought I would make another post detailing changes since then. I expect this post to be less useful than the last one.

For the most part, the overall outline has remained pretty stable and feels very similar to 18 months ago. Things not mentioned below have mostly stayed the same. I believe that the total effect of continued changes have been continued but much smaller improvements, though it is hard to tell (as opposed to the last changes, which were more clearly improvements).

Based on comparing time logging records I seem to now do substantially more work on average, but there are many other changes during this period that could explain the change (including changes in time logging). Changes other than work output are much harder to measure; I feel like they are positive but I wouldn't be surprised if this were an illusion.

Splitting days:

I now regularly divide my day into two halves, and treat the two halves as separate units. I plan each separately and reflect on each separately. I divide them by an hour long period of reflecting on the morning, relaxing for 5-10 minutes, napping for 25-30 minutes, processing my emails, and planning the evening. I find that this generally makes me more productive and happier about the day. Splitting my days is often difficult due to engagements in the middle of the day, and I don't have a good solution to that.

WasteNoTime:

I have longstanding objections to explicitly rationing internet use (since it seems either indicative of a broader problem that should be resolved directly, or else to serve a useful function that would be unwise to remove). That said, I now use the extension WasteNoTime to limit my consumption of blogs, webcomics, facebook, news sites, browser games, etc., to 10 minutes each half-day. This has cut the amount of time I spend browsing the internet from an average of 30-40 minutes to an average of 10-15 minutes. It doesn't seem to have been replaced by lower-quality leisure, but by a combination of work and higher-quality leisure.

Similarly, I turned off the newsfeed in facebook, which I found to improve the quality of my internet time in general (the primary issue was that I would sometimes be distracted by the newsfeed while sending messages over facebook, which wasn't my favorite way to use up wastenotime minutes).

I also tried StayFocusd, but ended up adopting WasteNoTime because of the ability to set limits per half-day (via "At work" and "not at work" timers) rather than per-day. I find that the main upside is cutting off the tail of derping (e.g. getting sucked into a blog comment thread, or looking into a particularly engrossing issue), and for this purpose per half-day timers are much more effective.

Email discipline:

I set gmail to archive all emails on arrival and assign them the special label "In." This lets me to search for emails and compose emails, using the normal gmail interface, without being notified of new arrivals. I process the items with label "in" (typically turning emails into todo items to be processed by the same system that deals with other todo items) at the beginning of each half day. Each night I scan my email quickly for items that require urgent attention. 

Todo lists / reminders:

I continue to use todo lists for each half day and for a range of special conditions. I now check these lists at the beginning of each half day rather than before going to bed.

I also maintain a third list of "reminders." These are things that I want to be reminded of periodically, organized by day; each morning I look at the day's reminders and think about them briefly. Each of them is copied and filed under a future day. If I feel like I remember a thing well I file it in far in the future, if I feel like I don't remember it well I file it in the near future.

Over the last month most of these reminders have migrated to be in the form "If X, then Y," e.g. "If I agree to do something for someone, then pause, say `actually I should think about it for a few minutes to make sure I have time,' and set a 5 minute timer that night to think about it more clearly." These are designed to fix problems that I notice when reflecting on the day. This is a recommendation from CFAR folks, which seems to be working well, though is the newest part of the system and least tested.

Isolating "todos":

I now attempt to isolate things that probably need doing, but don't seem maximally important; I aim to do them only on every 5th day, and only during one half-day. If I can't finish them in this time, I will typically delay them 5 days. When they spill over to other days, I try to at least keep them to one half-day or the other. I don't know if this helps, but it feels better to have isolated unproductive-feeling blocks of time rather than scattering it throughout the week.

I don't do this very rigidly. I expect the overall level of discipline I have about it is comparable to or lower than a normal office worker who has a clearer division between their personal time and work time.

Toggl:

I now use Toggl for detailed time tracking. Katja Grace and I experimented with about half a dozen other systems (Harvest, Yast, Klok, Freckle, Lumina, I expect others I'm forgetting) before settling on Toggl. It has a depressing number of flaws, but ends up winning for me by making it very fast to start and switch timers which is probably the most important criterion for me. It also offers reviews that work out well with what I want to look at.

I find the main value adds from detailed time tracking are:

1. Knowing how long I've spent on projects, especially long-term projects. My intuitive estimates are often off by more than a factor of 2, even for things taking 80 hours; this can lead me to significantly underestimate the costs of taking on some kinds of projects, and it can also lead me to think an activity is unproductive instead of productive by overestimating how long I've actually spent on it.

2. Accurate breakdowns of time in a day, which guide efforts at improving my day-to-day routine. They probably also make me feel more motivated about working, and improve focus during work.

Reflection / improvement:

Reflection is now a smaller fraction of my time, down from 10% to 3-5%, based on diminishing returns to finding stuff to improve. Another 3-5% is now redirected into longer-term projects to improve particular aspects of my life (I maintain a list of possible improvements, roughly sorted by goodness). Examples: buying new furniture, improvements to my diet (Holden's powersmoothie is great), improvements to my sleep (low doses of melatonin seem good). At the moment the list of possible improvements is long enough that adding to the list is less valuable than doing things on the list.

I have equivocated a lot about how much of my time should go into this sort of thing. My best guess is the number should be higher.

-Pomodoros:

I don't use pomodoros at all any more. I still have periods of uninterrupted work, often of comparable length, for individual tasks. This change wasn't extremely carefully considered, it mostly just happened. I find explicit time logging (such that I must consciously change the timer before changing tasks) seems to work as a substitute in many cases. I also maintain the habit of writing down candidate distractions and then attending to them later (if at all).

For larger tasks I find that I often prefer longer blocks of unrestricted working time. I continue to use Alinof timer to manage these blocks of uninterrupted work.

-Catch:

Catch disappeared, and I haven't found a replacement that I find comparably useful. (It's also not that high on the list of priorities.) I now just send emails to myself, but I do it much less often.

-Beeminder:

I no longer use beeminder. This again wasn't super-considered, though it was based on a very rough impression of overhead being larger than the short-term gains. I think beeminder was helpful for setting up a number of habits which have persisted (especially with respect to daily routine and regular focused work), and my long-term averages continue to satisfy my old beeminder goals.

Project outlines:

I now organize notes about each project I am working on in a more standardized way, with "Queue of todos," "Current workspace," and "Data" as the three subsections. I'm not thrilled by this system, but it seems to be an improvement over the previous informal arrangement. In particular, having a workspace into which I can easily write thoughts without thinking about where they fit, and only later sorting them into the data section once it's clearer how they fit in, decreases the activation energy of using the system. I now use Toggl rather than maintaining time logs by hand.

Randomized trials:

As described in my last post I tried various randomized trials (esp. of effects of exercise, stimulant use, and sleep on mood, cognitive performance, and productive time). I have found extracting meaningful data from these trials to be extremely difficult, due to straightforward issues with signal vs. noise. There are a number of tests which I still do expect to yield meaningful data, but I've increased my estimates for the expensiveness of useful tests substantially, and they've tended to fall down the priority list. For some things I've just decided to do them without the data, since my best guess is positive in expectation and the data is too expensive to acquire.