Category Archives: Uncategorized

My 2018 donations

At the end of 2018 (specifically, on December 22, 2018), I donated $3,000 USD with the following allocation:

  • $2,000 to GiveWell for discretionary regranting to top charities, by check
  • $500 to the $500,000 EA Donor Lottery by check, on behalf of Issa Rice. This means that, in the 0.1%-probability case that I win the lottery, he will choose the allocation of the funds
  • $500 to the Machine Intelligence Research Institute by ACH, on behalf of Issa Rice. The donation qualified for donation double from Raising for Effective Giving’s Double Up Drive

This is my first set of normal charity donations since 2012, when I donated $500 to GiveDirectly. In this post, I describe:

  • My reason for donating now, rather than earlier or later: Till now, I was first building up a saving buffer and then funding contract work heavily. With recent reduction in contract work spending and increase in income, I am better positioned to donate.
  • My reason for choosing this amount to donate: It’s the minimum amount that is substantive yet unlikely to materially affect me; I would donate more if I saw more compelling opportunities, and want to keep a reserve that will allow me to donate more in the future
  • The selection of recipients: For the money directly allocated by me, GiveWell top charities won as a default baseline because the alternatives were not yet compelling enough.

General background:

My reason for donating now rather than earlier

After my last donation of $500 to GiveDirectly on December 24, 2012, my personal focus shifted toward building a good saving buffer and acquiring personal financial stability. For these reasons, I deferred donating, with the expectation that I might potentially donate larger amounts later, as I had done prior to 2012. You can see some related thoughts from me in a post to the Effective Altruists Facebook group in March 2014.

By 2015, I had a decent saving rate and I was quite interested in exploring philanthropic opportunities, but felt that, rather than directly allocate money to charity, I should invest more in getting a deeper understanding of various domains that might be relevant. This was part of the motivation for me spending a fair amount of money financing contract work, mostly by Issa Rice and Sebastian Sanchez, which picked up in 2016 and has been proceeding at a fairly steady pace since then. The total money I’ve paid for contract work since I started on it has been around $115,000. This is a significant amount of money that has taken precedence over donating money to charity.

A few weeks ago, while taking stock of my finances, I realized a few things:

Together, these factors mean that I have a little more money available to donate to charity now than I did till recently.

My reason for choosing this amount to donate

My current financial situation is such that I could donate a reasonably large amount if I felt there was a compelling opportunity that I was uniquely suited to fund. Indeed, some of the contract work I’ve funded has fit that bill. Outside of the contract work I’ve directly funded, however, I have not yet found such opportunities. I have found a fair number of opportunities that meet some of the attributes, but most of them tend to get funded through other means, which is good for them and for me.

In the absence of compelling opportunities to fund, my default is to donate some minimum amount. Some reasons to keep my current donation to a minimum amount:

  • The desire to keep funds available to spend on more compelling opportunities later, which could involve scaling up contract work (if it’s a project I can fund or direct) or donating to another person, project, or organization.
  • The potential benefits of a larger financial buffer to support transitions to “direct work” careers in the future (though I don’t anticipate immediately making such a transition).
  • The potential personal benefits of a larger financial buffer, outside of altruistic reasons.

To figure out the amount to donate, I used a few heuristics:

  • It’s been 6 years since the last standard charity donation, and I feel like $500/year is a reasonable “minimum” — it is substantive enough but still small enough that it does not foreclose opportunities for me. Multiplying $500 by 6 gives $3,000.
  • It also roughly works out to about 1% of my after-tax income for the 6 years, which is another way of seeing that it is nontrivial but still not enough to meaningfully impact me.
  • The total amount that’s being donated is big enough that the financial transaction overhead doesn’t eat too much into it. This is another advantage of bunching across years.

The selection of recipients

For each of the years 2017 and 2018, I had given Issa the option of assigning $500 of my money to charitable causes of his choosing (with no strict requirement that these be recognized as charities). In 2017, Issa deferred the use of the money, so he had $1,000 to allocate.

Issa ultimately decided to allocate 50% of the $1,000 (i.e., $500) to the $500,000 EA Donor Lottery, and another 50% to the Machine Intelligence Research Institute (MIRI). For the donor lottery, the understanding was that although the entry was under my name, Issa would make the decision as to how to allocate the winnings in the (0.1% probability) event that I win the lottery.

More of Issa’s reasoning can be found in the 2018 section of Issa’s donation history page.

Since my overall giving budget was $3,000, and $1,000 was to be determined by Issa, that left $2,000 to allocate. When thinking about this allocation, I considered a few different options.

  • GiveWell top charities
  • Effective Altruism Funds
  • Interesting “rising stars”

GiveWell top charities

I’ve been observing GiveWell for quite some time. For a while, I’ve found their research quite good, but until a few years ago, I still had significant uncertainty around how well their top charities and their vetting process would stand up to the test of time. In the last few years, I’ve seen that the top charities seem to stand up quite well, the research process has continued to improve, and the funding cycle has a clear rhythm I can understand. I believe the problems that are being addressed by GiveWell top charities are large enough that there are few risks of room for more funding disappearing. I also think GiveWell’s approach to discretionary regranting is a good one.

Donating money to GiveWell for reallocation to top charities seems like a reasonable baseline choice. The question for me was whether there are better choices out there.

Effective Altruism Funds

I’ve been following the Effective Altruism Funds for a while, and in particular I’ve been keeping track of the grants they’ve been making. I am impressed with what I’ve seen so far, particularly from the Animal Welfare Fund, and I think the Meta Fund and the Long-Term Future Fund have also made some interesting grants.

However, I have a lot of uncertainty around how these funds will continue to operate over the years, and what sort of funding landscape they will be operating in. It could be the case that the money pouring into the funds significantly exceeds the amount of good places they can donate to. It could also be the case that we discover over the next year that the grants that seem good now turn out to be not so good, or even actively bad. Another source of uncertainty to me is how EA Funds will play with other financing initiatives like EA Grants: it could be the case that as EA Grants scales up, there are fewer of these interesting projects for EA Funds to fund.

I ultimately decided to watch the EA Funds (as well as the general landscape in which they operate) for a little longer before making a decision to donate to them.

Interesting “rising stars”

Some organizations start becoming donation targets in the effective altruism community after word of mouth from a few people. These could be:

  • Organizations that have been around for a while and have no other direct connection to the movement, but that get picked up by a few people in the community. Examples: Tostan in 2016, StrongMinds (2016 to present), No Means No Worldwide. I thought a bit about donating in this space, but decided that these organizations have plenty of funding sources, and I also wasn’t convinced of the thoroughness of the evaluation of them.
  • Organizations that started out from within the effective altruism community, and are still in a startup stage where they have enough funding to survive but maybe not enough to grow fast. Potential examples: AI Impacts,Alliance to Feed the Earth in DisastersLessWrong 2.0. While I might have been interested in funding specific organizations in this reference class, none of the ones I know fit the bill for me; I believe the funding flowing to these organizations is adequate and the constraints hampering their growth are not primarily those of funding. However, this is just with my current state of knowledge; I’m keeping an eye on a lot of these and might change my mind about specific ones.
  • Fledgling, not-quite-started organizations. Their very survival is unclear. At present, I don’t have a clear idea of which of them might be worth funding; the data I’ve seen didn’t point to a clear example of something that would be a compelling use of funds.

As you can see from the numerous links I’ve included above, I have (with help from Issa Rice) been collating a lot of donation information and discussion documents on the donations list website, and this has helped me get a sense of the landscape. I hope to continue to learn more and might find good opportunities.


The upshot is that GiveWell top charities “won” as the default baseline, and that’s where I ended up donating. But I also ended up making this donation smaller, opening the possibility for larger donations later.

I hope to continue working on projects like the Donations list website on my personal time and through contract work. I also hope to get a clearer picture of a number of things that might lead to more interesting donation decisions in the future.

Despite not finding anything immediately worthwhile or exciting to donate to, I did see a lot of potential opportunities. And in many cases, the limiting factor was that there were already many other donors who had snapped up the best opportunities. To me, this is a good problem to have :).

PS: This post is cross-posted to the Effective Altruism Forum and is also available in my GitHub working drafts. The versions may differ slightly due to formatting adjustments and limitations.

Why I stopped quarterly reviews, and what replaces them

Recently, I stopped publishing quarterly reviews. In this blog post, I explain the reasons, and describe what I’m doing instead.

Quarterly reviews previously served to provide periodic updates of three things:

  1. My recent activity.
  2. Total pageviews and other reach and impact metrics of my recent and lifetime activity.
  3. Updates to my beliefs and habits.

Unfortunately, they weren’t an ideal vehicle for any of the three. Among the problems:

  • The updates were too infrequent to provide insight into what I’ve worked on and where, and to allow others to track my work.
  • There was too much number-crunching involved in collating data from different sources.
  • The quarterly period was an unnatural way of subdividing my work and reporting incremental progress, and therefore created extra load on me in terms of efforts to summarize progress.
  • The things that people liked most (my belief and habit updates) received too little attention, as I (rightly) believed in the importance of reporting objectively on activity and reach/impact measures.

My solution going forward is as follows:

  1. My recent activity: With a few stated exceptions (as described below), all my activity can be tracked by following me on GitHub. I explain this in more depth below. This is near-real-time, and provides a lot more granularity and insight.
  2. Reach and impact metrics: I am working to make these available in publicly accessible locations, updated at approximately daily frequency, so that people can check them out any time.
  3. Updates to beliefs and habits: I will publish these occasionally, at logically natural points, just as I occasionally publish blog posts or articles summarizing work I’ve been doing or key insights I’ve had. I don’t try to pressure myself to do these; I write them quickly when the ideas are in my head.

Note: As I hint at right above, I try to write these “updates to beliefs and habits” posts quickly, without thinking through them too much. Otherwise they wouldn’t pass a cost-benefit analysis for me. So I basically wrote the current post in one straight sitting of about three hours (which included a bunch of other website updates that were triggered by this draft). That might explain why this doesn’t look like it has undergone editorial review and improvement.

#1: Following my activity

My activity is of roughly three kinds:

  • Self-contained projects, that usually involve a mix of code (to render results, allow for programmatic querying) and data. Each such project has its own GitHub repository, as well as a public-facing website. Currently, there are three projects I am working on: contractwork (website:, donations (website:, and Wikipedia Views (website: I also have a project on BART (intended website: that does not yet have a public website, but some of the backend code is written.
  • Miscellaneous written content I create or improve, in the form of wiki pages, blog posts, articles, Facebook posts, and comments. Some of it is related to a larger project of understanding a specific aspect of the world, but it isn’t a self-contained project. This content is scattered across numerous venues. To name a few: my blog (the one you are reading right now), the Timelines wiki, wikiHow, Market subwiki, the Effective Altruism Forum, LessWrong. I generally keep a copy of the content I create (or, in case of significant updates, the content I update) in the working-drafts repository mirrored on GitHub. In case of updates, I usually keep a copy of the entire content, though if my updates are modular enough, I might only keep a copy of my updates. I also use this repository to maintain copies of drafts for content not yet published, or background material I want to link people to without “formally” publishing it as a finished piece of work. In some cases, I also keep copies of comments I leave on blog posts, Facebook posts, Facebook comments, survey responses, and (drafts of) emails I send to people.
  • Other forms of activity: The bulk of my activity on Facebook, and almost all of my activity on PredictionBook, is not mirrored in working-drafts. To keep track of these, you’d need to follow me on Facebook and check me out on PredictionBook.

The upshot: following all my public activity on GitHub (for instance, by cloning all the GitHub repositories and pulling updates), as well as on Facebook and PredictionBook, should give a reasonably comprehensive picture of what I’ve been up to and how I work. Note that these are not comprehensive ways of seeing all work I’ve ever done; I’ve only started mirroring my edits/updates in the working-drafts folder in the last twelve months or so.

I have included more detail on the first two bullet points on my website page about GitHub.

EDIT: Starting in October 2018, I started using daily-updates issues to provide more information on my activities at daily granularity (outside my day job and daily sustenance activities). See my post on the why and how of daily updates for more.

#2: Keeping track of total pageviews, reach and impact metrics

For my Wikipedia content, you can keep track of it through Wikipedia Views, as described on my site page about Wikipedia. Specifically, here’s the links to get data for pages I created:

For websites that I own or manage, I am trying to make the website data public in three ways:

  • Quantcast Measure (QM), with details on pageviews, visits, uniques, and demographic and geographic distribution of visitors: QM has a limit of ten sites, and does not show sites with too little data, so I am restricting its use to relatively high-traffic sites. Currently it applies to the five top subject wikis (you can see the list of subject wikis with links to their Quantcast pages, at In addition, I’ve also activated QM for QM numbers are usually a little lower than those of Google Analytics, but in the same ballpark, and they help capture the relative trends reasonably well. The main downside is that QM works only from the point in time that it was turned on. I turned QM on for,, and on November 27, 2015, and for,, and around April 2, 2017.
  • SimilarWeb: I am trying to connect SimilarWeb to Google Analytics, so that it uses certified metrics from Google Analytics when displaying data. I have submitted information to SimilarWeb; as of the time of writing this (April 8, 2017) it does not seem to have started incorporating Google Analytics data.
  • Using the Google Analytics API to make key metrics from Google Analytics publicly visible: I expect to do this with some help from Issa Rice some time in the next few months. This will give everybody access to key metrics from Google Analytics that are currently visible only to people I’ve given access (we probably won’t make all metrics available, to maintain simplicity and preserve privacy). EDIT: This website is live at

Beyond this, I will also update annually with total Google Analytics pageview counts for subject wikis, as I have been doing.

I have not worked out a suitable solution yet for websites that I do not own or manage, although once we have the setup to make Google Analytics data publicly available, I can apply that to websites I do not own but have (or can get) Google Analytics access for.

#3: Updates to my beliefs and habits

This is the aspect of my quarterly reviews and other writings that has attracted the most positive feedback. Freed of the need to publish quarterly reviews (given the superior alternatives that now exist), I can devote more time to writing up updates in my thoughts. Broadly, there are two kinds of posts I expect to write, and both will be fairly occasional:

  • “Wrapper” posts that provide a summary of a project I am working on, and of key findings. These wrapper posts could be to advertise the project, to checkpoint an accomplishment milestone (because the project per se never ends), and to convey key conclusions that might be of interest to people who aren’t interested in the project per se. The “wrapper” jargon is borrowed from, which says “Non-traditional contributions often don’t get credit unless authors wrap them with proxy papers. Unfortunately, this multiplies effort and divides attention.”
  • General posts about my thought processes, beliefs about the world, ways of working, etc. that are one level “meta” and are not tied to any project I am working on. My three guiding principles and Debugging my apparent 2016 stagnation blog posts are examples.

My goal with both these kinds of posts is to, essentially, write them only once the ideas are all in my head and reasonably clear, so that it’s just a straight exercise of transcribing from my head to the computer. Given that I have no particular pressure to “publish”, I believe it does not make much sense to artificially try to put in custom, “hacky” effort to push out posts of either of the above kinds in a way that interrupts the flow of the larger projects I am working on.

To give an idea of what I used to do, and am now choosing against: there is this strategy where I would decide to work hard on pushing out a wrapper post about a topic I’ve been learning about, even if I didn’t feel like I was fully ready to write it, or knew all the relevant facts. Examples of the kinds of topics I am talking about: understanding trends in Wikipedia pageviews, or understanding the history of immigration enforcement in the United States since 1986. My past strategy was: I would just draft it, pull in a fact from here and a fact from here, revisit, redraft, rewrite, and soon get something that looked okay. And there was a time, early on, when I found that this kind of effort helped me focus and collate information that I would not otherwise have interest in systematically grasping. And I still respect this approach.

However, my current belief is that in this sort of situation, it’s better to just keep collating background information in accessible formats. such as continued work on the Wikipedia Views website, to make it easier and easier to look up Wikipedia view trends, or work on timeline of immigration enforcement in the United States. And then to start working on the wrapper post only when I feel I have enough to say that I can just sit down and say it.

(To be honest, I’ve always followed my current approach in some areas — including group theory, my Ph.D. subject — which also incidentally led to me spending too much time on general learning and too little on cracking Ph.D.-worthy problems. But I’ve tended to give a lot of weight to my approach of “push out at all costs” in some domains, and I am now giving it less weight. I still think that approach can be quite good when I am just starting out in an area, and don’t have the general landscape knowledge and wherewithal to embark on the right sort of larger, more long-term projects.)

There are times when I feel like I can’t do more background work on a project (i.e., I don’t know how to proceed next), but I don’t have a wrapper post in my head either. In that case, I might start an exercise of faux-writing the wrapper post: just starting to draft it, and notice the pain points that are stopping me from writing it. And then use those to decide where next to continue to make progress with the project. The wrapper post thus serves as a diagnostic rather than a goal.

Now moving to general posts about my thought processes, i.e., the ones that do not correspond to any specific project, and are one level meta. I generally feel the need to write these when there is a huge dissonance between things I have said or believed in the past, and the way I am behaving right now. But even then, I don’t actually write them unless I feel I have some sort of explanation, or at minimum decent hypotheses, that I can quickly articulate. For instance, the current post (that you are reading right now) was born of a dissonance between my past intent to publish quarterly reviews and frequent wrapper posts, and the current reality of no quarterly review. I had been experiencing the dissonance for a while but did not write the post until now, because it was only now that I felt I could articulate things well enough. Unlike many people in the rationality community whom I know (and know of), I don’t engage in goal-setting and introspection exercises of my own volition, but rather, I wait for things to boil over before I decide to write about it.

Concluding thoughts on transparency and ease of understanding

I want others, including my future self, to be able to easily figure out what I am doing, where it is headed, and what impact it is having. At the same time, I want this “transparency” to be low-friction and effortless for me. I have moved away from an approach of periodically summarizing things, to one of continuously revealing information as part of my modus operandi (and therefore requiring no special effort).

Is this style of transparency actually more helpful to others? Yes and no. Because there is more information available in real time, in a more “computable” format (commit histories on GitHub, public website data at daily granularity), people can dig in much deeper than having to wait for a quarterly review. On the other hand, a quarterly review provides an important function of summarization, making it easier for a casual follower to get the big picture (as sumarized by me).

In the longer run (and I thank Issa Rice for this insightful observation), I expect that the disadvantages of lack of hand-crafted summarization will be overcome by the advantages of more granular, computable, real time data. That’s because more tools will become available for analysis and summarization of activity. These include some I might create myself, through code improvements in the projects, and general ones such as GitHub’s activity graphs.

My three guiding principles

Epistemic status: stream-of-consciousness rant

In the past few months, I’ve been somewhat depressed at my output outside of my day job. This led me to write the post Debugging my apparent 2016 stagnation last month. Some of the depression also reflects general frustration at the slow-paced nature of progress in the world, something that I mentioned when describing my background beliefs three months ago.

I’ve been thinking about how to think about whether what I’m doing is important. Fundamentally, I believe that, given my current constraints (a day job and not a lot of money to spend), my best bet is to continue to do “social science” construed broadly: investigate questions about the world and communicate my thoughts clearly. Thinking through, I’ve identified three guiding principles to what I should be doing outside of work.

  1. Pick important topics for investigation, based on criteria including value of better understanding for the world at large, and for my own model-building.
  2. Investigate them rigorously, combining the best ideas of academic rigor with the practical nature of the real world.
  3. Communicate what I’ve found clearly, comprehensively, frequently, and to the right audiences.

I believe that, on the whole, I’ve done a decent job with (1). Two of my focus areas have been migration (chosen for its broader importance) and web analytics/content creation (chosen as something of instrumental importance for me, and something I have more in-depth knowledge of). My relative focus between (2) and (3) has shifted. In previous years, I was focused more on (3), often to the detriment of (2). In other words, I would be eager to share results of partial research without fully understanding the context and details. In 2016, on the other hand, I moved significantly toward (2), and have generally neglected (3).

I think it’s really hard to get (1), (2), and (3) right together. The main way (1) creates a constraint is that anything that’s important is bound to be hard enough that (2) and (3) are difficult to do together.

Growth through outreach, borrowed against the future

In my earlier years, I focused more on outreach, to the detriment of deeply understanding the things I was talking about. For instance, when working on Open Borders: The Case, I would do a little bit of research, then write a blog post or a Facebook post, publish it, get feedback, and move forward. The effort that went into the post was nontrivial but still small. This led to reasonably rapid growth. This growth was borrowed against a future time when there would be more substance to back up the initial moves.

This sort of growth happens because a lot of messages are “waiting to be heard”. People hear the message, they like the direction in which the person conveying the messenger is striving to go, they see that there’s nothing out there right now that quite matches that, so they give it their attention. But staying at the same level won’t keep their attention; it’s important to transcend to a different level, otherwise they’ll get bored and move on.

One of the biggest mistakes with growth through outreach is that it can be a seductive local optimum: keep communicating material at the same level (reasonably well-researched, reasonably well-communicated) and expect a reasonably similar level of response. Decay, disillusionment, and fatigue will set in fairly slowly, and things will generally look like they run well. But the promise of qualitative change that originally attracted people isn’t fulfilled.

The trivia trap: when research is divorced from important questions

The trivia trap is one where I end up thoroughly investigating questions that are peripheral to the important questions I seek to better understand.

Fortunately, this isn’t a trap I have fallen into of late, but it’s something I should be wary of. Being cognizant of this trap also stops me from going on tangents on interesting side-investigations that come up when investigating an issue.

The rigor trap, and failure to communicate frequently

The “rigor trap” is when I just keep trying to investigate a question more and more thoroughly (based on my conception of thoroughness) without spending effort on communicating the findings frequently. I feel like I have come close to failling in this trap.

With respect to my contributions to Open Borders: The Case, the last 1.5 years or so can be thought of as succumbing to that trap. I’ve been busy researching a bunch of things related to migration (in the limited time I get from other things). I have also been slowly writing up my findings. Over the long term, I believe I’m communicating clearly and comprehensively. The part where I’ve failed is in communicating frequently.

There are two problems with infrequent communication:

  • Others lose interest in the underlying topics, and move on.
  • Alternative narratives, ones that are even less well-grounded than my hastily researched ones, become the dominant narratives.

Both of these are serious, genuine problems I’ve seen. Fortunately, in the case of Open Borders: The Case, other co-bloggers have stepped up to fill in some of the void I left. Nonetheless, I wish I could have avoided the rigor trap more intelligently, and I hope I can do so in the future.

The yoyo trap, and failure to communicate comprehensively

The yoyo trap is one of an inauthentic balance between (2) and (3). Here, on the one hand, I do deep research into important questions. On the other hand, I communicate things that are divorced from the insights I am gathering through my deep research, simply based on what I expect will get the most positive feedback from an audience.

The yoyo trap is seductive because at first glance it appears that I am getting everything done: I am doing thorough research and I am communicating stuff well. But the value of the work I do isn’t being used if I’m communicating something entirely different. And this hurts both sides. It makes me less motivated to really, truly dig into the research I am doing, because I know that won’t be that important to what I communicate. And it hurts my communication, because I know that doesn’t come close to sharing what I know.

Examining others through the same lens

This post was written in terms of reflecting on my own actions, and thinking of a framework for moving forward. However, I am also examining others’ actions through this lens.

I see rigor traps and yoyo traps all around me, and this concerns me a lot. What concerns me is that people often don’t even see these as traps, but even issue self-congratulatory statements announcing it. “What we’re doing is so complex, that we’re unable to share it except with elites who we can trust to be on the same wavelength” is a common self-congratulatory articulation of the rigor trap. People don’t even see the yoyo trap as a trap: organizations often take pride in, for instance, separating their “public” communications from their private ones, or hiring social media experts to handle their social media communications.

Acknowledging that social science is hard

True, legitimate social science is really hard. It involves using a wide variety of investigative tools, ranging from reading historical works to conducting online surveys.

In a Korean drama I recently witnessed (called Pinocchio) a man almost quit being a reporter after messing up the reporting of a major incident that cast a dead woman in an unfairly negative light. Narrating that to his adoptive father, he said something to the effect that he jumped into it, not knowing that it is hard. His adoptive father told him something to the effect: “So the problem was that you jumped into it without knowing that it was hard, and that’s why you ran into trouble. Now that you know it is hard, you can jump in with full knowledge of the fact and you won’t be scared.” (You can get the full episode details here, if you’re interested). Similarly, I need to know that it’s hard, but that it’s still really important to bring together important questions, thorough research, and clear, frequent, and comprehensive communication.

August to October 2016 in review

This is my fourth quarterly review, and it covers the months from August to October of 2016.

Things continue to be busy on the job front. There were a few big job-related projects that happened this quarter. I expect the next quarter to be more relaxed on the job front, partly because of a winter break slowdown and partly because some of the big projects are over.

Updates to Wikipedia Views

I made a number of updates to the code underlying Wikipedia Views, a tool I build to make it easy for people to look up views of multiple Wikipedia pages over long time periods. Many of these changes were motivated by work I’m doing on a Wikipedia decline blog post discussed later.

The chief changes I made include:

  • Added support for looking up mobile web, mobile app, desktop spider, and mobile web spider data from July 2015 onward, using the Wikimedia API. Also, to make the desktop data consistent with the other sources, I switched desktop to use data from the Wikimedia API starting July 2015 (previously, it had used the Wikimedia API only starting January 2016). This has retroactively changed some pageview counts on Wikipedia Views.
  • Added support for cumulative Facebook shares starting October 2016, and drawing on the Facebook API. This data only includes pages that were successfully captured at the end of the month.
  • Added automatic graph generation when viewing data for multiple months or years. The graphs, shown below the table, help give a bird’s eye view of the numbers in the table.

Other changes I hope to make in the next few weeks are listed below.

  • Re-enable CSV export, a feature I had disabled while adding new drildown-based features.
  • Complete documenting and fixing bugs in the existing features.
  • Highlight pages that currently redirect to other pages.
  • Add support for showing data on breakdown by referrers for a few months in 2015 and 2016 when Wikipedia clickstream data is available.
  • Allow for comparison of data across multiple languages in the same query.

Wikipedia editing

For background information, see my site page about Wikipedia

I created two new pages this quarter: Alan C. Nelson and Gladys Block. Both of them were created in the latter half of October, after a long hiatus from new Wikipedia page creation.

I expect to resume Wikipedia page creation after I finish work on Wikipedia Views and on the “Wikipedia decline” post I discuss later in this review.

While reviewing my early Wikipedia page creation history, I realized that a little over 100 pages I had created prior to 2010, that have survived deletion, were not included in the Wikipedia Views tag for pages I created. After adding them, the total count of pages I have created (including a few redirects) stands at 417.

In total, pages I created over my lifetime got 584,534 desktop pageviews, 354,365 mobile web pageviews, and 8,081 mobile app pageviews. In addition, they were viewed 141,778 times by desktop spiders and 14,885 times by mobile web spiders.

You can see the lifetime data for the tag as a whole here. I have also attached a graph from the link; the highest line is the total and the second highest is the desktop pageviews. You can also access a gigantic table (with 223,095 cells of which 71,307 are legitimately filled) of all pages I created across all months and all device types here (cautionary note: this is a large page and may take several seconds to load).

Liftetime pageviews of pages created by Vipul Naik
Liftetime pageviews of pages created by Vipul Naik

Now that I’ve started recording cumulative Facebook shares, I have data that pages I created over my lifetime had received a total of 2904 likes+comments+shares on Facebook as of the end of October 2016. You can see the data by page here. Couple of caveats:

  • This excludes 106 pages I recently realized I had created a while ago, so the total is an underestimate. The excluded pages show the message “cannot retrieve this data” indicating that data on cumulative shares was not captured during the month.
  • The counts are probably underestimates since we have evidence that Facebook has misplaced likes, comments, and shares that occurred in 2013 or earlier.

Sponsored Wikipedia editing

For background information, see my site page about sponsored Wikipedia editing

Sponsored Wikipedia editing operations continued, though at a lower overall level than previously. Many of the people recruited either left permanently, or have significantly reduced output because their academic term (high school or college) has begun.

Issa Rice (pages and payments) and Sebastian Sanchez (pages and payments) have continued to work at a similar rate as before.

Sebastian in particular has continued to work on pages in two broad themes: disease-related timelines (18 pages so far, on which I’ve spent $1466) and timelines of healthcare by country (14 pages so far, on which I’ve spent $1381). He has also created four other pages in themes broadly related to healthcare and global health, and done one translation.

I plan to more comprehensively review sponsored Wikipedia editing at the end of the year. By then, I expect to enter the data into a database so that I can generate a variety of summary reports.


For background information, see my site page about WikiHow

I wrote one new wikiHow article: How to Understand Your Website Traffic Variation with Time. The article got a lot of views in the first week (getting to about 800 pageviews), likely because it was featured prominently on various landing pages and lists of top articles. Its growth has subsequently slowed down to between 5 and 10 new pageviews a day. The draft history (before I pushed it to WikiHow) can be found by looking at the history for the version in my personal Git repository.

This post is the second in a series of posts I intend to write on understanding website traffic. My previous post, published last quarter, was How to Understand Your Website Audience Profile.

I also began drafting a new article in wikiHow style: How to Master Online Surveys. I don’t know if I’ll try publishing it on wikiHow (as there are many very similarly named articles there right now).

Personal server migration

I finished migrating a personal server (that hosts a variety of websites I manage). I had begun this migration last quarter.

Wikipedia decline blog post

I worked together with Issa Rice on a blog post that continues to explore questions originally raised in my post The great decline in Wikipedia pageviews. The new post is far thorougher than the older post. It uses new, more reliable pageview data, Internet surveys, and better graphing and visualization tools.

The first iteration of the post was done by Issa Rice over about ten days in September, and can be seen here.

Since the post was turning out to be fairly involved and we realized that a lot of pieces were missing, I decided to take over the completion of the post, since I could coordinate that better with improvements to Wikipedia Views that would help make research and exploration for the post easier. A copy of the current (very incomplete) draft can be found here.

Writing the post has been a challenging exercise of going back and forth between simplifying the content for presentation, and diving even deeper into the empirics and data in order to obtain clearer conclusions. Realistically, I expect that we will have the post ready for publication in January 2017.

Market wiki updates

Some posts by Buck Shlegeris linked to the price bundling page on the Market Wiki that I had written a while back. Buck raised a few questions (not specific to the page, but related to the topic of price bundling). Inspired by these, I worked to significantly improve coverage of price bundling. I also spent part of some weekends and some BART rides on improvements to the articles Effect of sales tax on market price and quantity traded and effect of price ceiling on economic surplus.

I believe that this wiki is one of the more promising things to edit when I’m in the mood for doing something that is intellectually interesting but doesn’t involve too much of the “arbitrary real world” i.e., doesn’t involve too much research on specific current events or history. There are times when I want to take a break from the huge amount of arbitrary information the real world entails, but I still want to do something in the economics or sociology realm rather than deal with pure mathematical abstractions. The Market Wiki fits this mood niche, and I believe that it’s one of the most high-impact things I could do that fit the mood niche.


I watched a few more Korean dramas:

I also watched two Japanese dramas, Mischievous Kiss: Love in Tokyo and its sequel, Mischievous Kiss 2: Love in Tokyo. They were based on the Japanese anime Itazura na Kiss. In addition, I finished watching The Flower in Prison, a semi-historical Korean drama based on events in the mid-16th century Joseon era.

Debugging my apparent 2016 stagnation

If you look at my quarterly reviews over the past few quarters (here, here, and here) it does look like there’s been stagnation in a number of respects, including production and consumption. As somebody who’s frequently raised the issue of stagnation in others’ output, I’m concerned about my own stagnation.

Stagnation has been seen in the following respects:

  • Reduction in the number of Wikipedia pages created: For instance, August and September 2016 were two consecutive months where I didn’t create any Wikipedia pages at all, after a track record of having created at least one page a month for the past two years.
  • Reduction in total pageviews for the Wikipedia pages I’ve created: This has two aspects: first, total pageviews for Wikipedia pages I’ve created over my lifetime fell year-over-year in 2016. Second, the new pages I’ve created in 2016 have received much fewer pageviews in 2016 compared to what pages I created in 2015 received in 2015 (even on a per-page basis, i.e., even after controlling for the lower volume of page creation). And both are way less than the corresponding counts for 2014. To be concrete, the 24 pages I created in 2016 got 4,235 desktop pageviews in September 2016, compared to 36,940 for the 61 pages I created in 2015 in September 2015, and 79,562 for the 155 pages I created in 2014 in September 2014.
  • Reduction in the total number of blog posts published: In total, across various fora (including Open Borders: The Case, the Effective Altruism Forum, and LessWrong), I’ve published very few posts this year. The respective numbers for the first nine months of 2016 are 2, 2, and 0. The corresponding numbers for the entirety of 2015 were 18, 4 (3 of them in December), and 1.
  • Reduction in other public online activity: My rate of Facebook posting has declined significantly. I’ve become basically inactive on Quora.
  • No significant increase in number of WikiHow articles: I’ve created 5 pages in 2016, compared to 6 in 2015. My activity on WikiHow began in May 2015, so considering the lengths of the respective time periods, activity in 2016 seems lower than that in 2015.

This is interesting to me because observing stagnation in other things often causes me some concern and puzzlement. Seeing the same (apparent) phenomenon in myself gives me a unique opportunity to dig deep and understand what’s going on. That can help me better guess what might be going on in other contexts where I don’t have inside information.

I believe there are a few different things going on. I describe them below for a better understanding of the phenomenon of (apparent) stagnation.

Note: I talk a lot about people I pay to create content on Wikipedia. Although I use the first-person singular (“I”), I should clarify that a lot of the management of the editors is done by Issa Rice, so “we” could be more appropriate. However, since there is a lot of comparison with the past, and Issa’s involvement in managing editors has been significant only recently, I’ve stuck to first-person singular to make the comparisons easier.

Significant shift to producing longer and much more thoroughly researched content

One way I’ve shifted gradually since 2014 is toward producing longer pieces of content. There are three levels of hierarchy:

  • At the highest level is the overall project, e.g., “understand 19th century United States migration” or “understand website traffic analysis”. At this level, the project is essentially unending.
  • At the mid-level is an original research/synthesis deliverable, usually in the form of a blog post or WikiHow article. The piece could be several thousand words long, include many pictures in some cases, and combine many topics into a coherent whole. An example under “understand 19th century United States migration” is the blog post How did we get here? Chinese Exclusion Act buildup (1848-1872). An example under “understand website traffic analysis” is the WikiHow article How to Understand Your Website Traffic Variation with Time.
  • At the lowest level is mini-contributions, such as Wikipedia articles or Facebook posts, that fall out as a “trail” of the research. Note that Wikipedia articles don’t make sense in all cases, but they do make sense in cases where Wikipedia’s notability criteria are satisfied. For instance, in the process of writing up the post High-skilled hacks: a (very) bref overview of H-1Bs (more to follow), I wrote a number of Wikipedia articles related to the subject (as summarized in a paragraph in the post itself). The same is true of my Chinese Exclusion Act post.

One of the implications of this is that the Wikipedia articles I create have a higher ratio of (research complexity) to (expected pageviews) than the ones I used to create. There are two aspects to this phenomenon.

  • The topics themselves are more obscure, and less likely to be of interest to many people. So regardless of the content of the article, they are going to get less pageviews.
  • My goal isn’t just “create a Wikipedia page that’s good enough for casual readers of the topic” but “create a Wikipedia page and in the process research the topic throughly enough (even if all details don’t make it to Wikipedia)” because I want it to be a building block to my mid-level goal.

For instance, the Wikipedia pages on Consular nonreviewability and Systematic Alien Verification for Entitlements took me a lot more time than they would if my goal was just to create something that would pass muster with Wikipedia. The latter in particular took me a lot of time as I was pulling in from many diverse sources.

The mid-level content I produce generally signals its own thoroughness reasonably well, but the extent to which quality can be signaled does not grow in direct proportion to the effort expended. I could probably have spent about half the time I did on my WikiHow articles and still produced something that, to a casual reader, would look approximately as good. Of course, if they were looking for the specific additional things that came out of the extra time spent, it would affect their impression more. But a casual reader just surveying the article may not fully see what exactly the return-on-investment was on the effort spent. Part of this can be combated by looking at the GitHub revision history, that helps better explain the step-by-step addition of content.

Here are a couple of examples of comparisons between posts I wrote in 2016 and 2015 that were part of the same big project, so you can understand more clearly the differences in thoroughness:

  • The project of understanding migration history actually began in late 2014, with a blog post back in January 2015: How Did We Get Here? Chinese Exclusion Act — Implementation (1882-1910). That’s actually illustrative for contrast because, viewed in isolation, it seems like a decent, fairly well-researched post. However, my April 2016 post How did we get here? Chinese Exclusion Act buildup (1848-1872) took more than double that time to write, and is much more heavily researched. And yet, it’s hard to say that most readers will be able to judge the difference in isolation, without a careful side-by-side comparison of the posts.
  • The project of understanding high-skilled migration began late in 2014. My early 2015 post High-skilled hacks: the case of Optional Practical Training was fairly well-researched, but mostly piggybacked on my existing knowledge rather than involving fresh research. It also had a number of minor errors and omissions that I fixed as I understood the topic better over subsequent months of studying it. In contrast, my June 2016 post High-skilled hacks: a (very) brief overview of H-1Bs (more to follow) required well over double the research effort, and again it’s unclear that a casual reader would experience a significantly different affect on the second post than the first.

Net effect: Fewer articles created and fewer pageviews per article created.

Heavier outsourcing of “pop” content creation and thematic content creation

I’ve cut down quite a bit on creating “pop” content on Wikipedia, i.e., creating articles about popular topics. I still believe they are valuable to create, but they don’t require my skills and expertise. Part of my sponsored Wikipedia editing focuses on such content. For instance, last week I asked Alex K. Chen to make the Wikipedia page for Michael Seibel, which he did.

There is another category of content that I have successfully outsourced, which is “thematic” content creation. Here, we identify a particular type of article, with some variable parameters (for instance, timeline of healthcare in country X, where X can vary over the 200+ countries). I closely supervise and provide feedback on the first few articles with that theme, but then the person creating those articles basically knows the procedure and can keep doing it for other countries. I’ve had success using this approach with Jesse Clifton (who’s made pages on animal welfare and rights by country), Sebastian Sanchez (who’s made pages on timeline of healthcare by country, and timelines of diseases), and Alex K. Chen (who has covered timelines for tech companies and has recently expanded to other companies).

Even with thematic content creation, there’s a difference between my focus areas in 2015 and my focus areas in 2016. In 2015, when I was trying to establish the tractability of thematic content creation, my focus was on topics I expected to be more popular. Hence, I focused more on timelines for technology companies, with timelines created by me (in 2014) and later by Alex K. Chen (starting 2015). In 2016, I picked more exotic themes, such as animal welfare and rights by country, timeline of healthcare by country, and timeline of diseases. Why the switch? Partly, the topics selected in 2016 were ones that formed part of a bigger picture of interest. Specifically, healthcare and diseases are important because a lot of people believe that health interventions offer very high value-for-philanthropic-money. I wanted to get deeper background on healthcare and diseases, and funding Sebastian’s work on these timelines is helping with that process. Animal welfare and rights by country was similarly motivated. As a result, if you look at total pageviews for thematic content creation I’ve funded in 2016, it doesn’t come out that different than 2015, even though the number of such pages funded has increased significantly.

Net effect: The changes from 2014 to 2015 shifted article creation away from me to the people I funded. The changes from 2015 to 2016 caused a decline in the incremental pageviews for new articles, despite an increase in the number of articles created.

Outsourcing of some other low-level mini-contributions

After Issa Rice created the timeline of global health, he and I identified a number of missing Wikipedia articles on related topics. Some of these articles were subsequently created by Issa. Others were given out to other contract workers. Many of them are interesting mini-projects for whoever chooses to do them. The goal is to improve our understanding of global health, for reasons discussed earlier in this post: namely, a lot of people thing health interventions are very good value-for-money in philanthropy.

Other areas where we outsourced some content creation include philanthropic foundations and the United States welfare system. You can go through the GitHub revision history of the new article pool to see how articles entered the pool, were reserved for specific editors, and then got removed as the article got completed.

Net effect: Even adding in articles whose creation I funded does not boost pageviews significantly, because many of these were low-level mini-contributions to grander projects. They require effort to create and don’t get a lot of pageviews.

Switch away from social media and public solicitation

This year, I’ve switched a lot to paying people to do stuff. In previous years, my model was more along the lines of doing stuff myself (or paying a couple of others, specifically Issa Rice and Alex K. Chen). I relied on public solicitation of feedback and thoughts, plus discussion with Alex and Issa and occasionally relevant subject-matter experts.

Now, a lot of the discussion happens privately. Issa and I discuss the topic and potential sources with the editor in question. We might sometimes seek thoughts from the general public (through Facebook) but that’s often not the case. This is mostly because the topics are more obscure, and we don’t expect most people to know things about them that are hard to Google.

Net effect: Less activity on social media, giving appearance of less work.

Some other measurement confounders

In January 2016, I switched counting of Wikipedia pageviews to the Wikimedia API, excluding bot views. Views till 2015 had included bot views. The transition to no bots led to a decline in estimated pageviews. Also, since it affects pages with fewer overall pageviews hardest, it makes the performance of my 2016 pages look worse compared to 2015. With that said, even adjusting for this still shows much less traction achieved in 2016 by pages I created.

Overall, are the changes “good” changes? Are they working out well?

For the most part, the apparent decline is more a story of a change in strategy than an actual decline. But is the change in strategy an improvement? Unfortunately, at this point it’s hard to be conclusive. My intuition is that a lot of these changes are improvements. However, you can differ in terms of values and empirics. For instance, if all pageviews are weighted equally, then the switch to the sort of pages I’ve created and funded recently is worse than sticking to pages like the old ones. On the empirics side, you might have a low opinion of the quality of the content that I’ve created and funded of late, or at least, think it is not of better quality than past, much more quickly produced, content.

Was my past approach wrong for that era?

Although I believe that my present approach to content creation makes more sense now, I do not mean to say that my past approach to content creation was wrong. In fact, I’d say that my original approach to content creation was probably a good fit for that stage of things. It provided a proof of concept in a relatively low-stakes environment (one-off page creations without undertaking a larger project). Also, by selecting more popular pages, I was able to achieve pageview counts that would come across as reasonably impressive to casual observers. But the current approach makes more sense as part of a plan to drive long-term value. Or so it seems to me, right now.

Implications for understanding apparent stagnation in others

One lesson is that increasing depth somewhat can result in a substantial reduction in apparent output, without the increases in quality being blatantly obvious to a casual observer. Whether the increases in quality are still worthwhile can depend on context, but casual observation won’t be a very reliable guide to that.

I believe one implication is to look more closely at revealed preference, especially preference in terms of financial investment and depth of engagement. The fact that I’m now paying significantly more (funding others’ work in addition to doing stuff on my own) is some evidence that I have good grounds for believing I am making progress. Similarly, if evaluating others’ progress, the extent to which they are able to draw people to work for and engage with them, and the extent to which they are able to raise money, can be stronger indicators than just looking at raw numeric counts of the amount produced or consumed. I’m not too satisfied with this answer, primarily because there is also a flip side: optimism bias pushes people to incorrectly believe that things are improving and that they are getting better, regardless of whether they actually are (and that’s a bias I might be under the influence of, too).

May to July 2016 in review

This is my third quarterly review, and it covers the months from May to July of 2016.

Things continue to be busy on the job front. I also wrapped up some of my personal bureaucracy by the end of April, so this quarter was focused more on personal projects. However, the main personal project I worked on in the quarter was working with Issa Rice to scale up sponsored Wikipedia editing.

Wikipedia editing

For background information, see my site page about Wikipedia

I created 1 page in May, 2 pages in June, and 2 pages in July. In total, I created 5 Wikipedia pages.

As I mentioned in my previous quarterly review, the pages I am creating right now are not as optimized for pageviews as pages in the past have been. Rather, they are pages in domains where I want to acquire a deeper understanding and believe the research process involved in creating the page will help me. All the pages I created in the quarter were related to migration.

In total, pages I created over my lifetime got 474,772 pageviews over the quarter. You can see the data for all of 2016 (only three of the seven months are in this quarter) here. This is between the 25th percentile estimate (460,000 views) and 50th percentile estimate (520,000 views) that I made in my previous quarterly review. So it was a little lower than my median estimate but well within the usual range,

Forecast for future impact: I expect the pageview counts to be roughly similar to the previous quarter: a baseline of around 155,000 pageviews per month, plus a possibility of minor spikes because of some pages being topical. The estimates are a little lower than those I made last quarter, and the variance is a little higher.

  • 2.5th percentile: 310,000 views.
  • 10th percentile: 375,000 views.
  • 25th percentile: 445,000 views.
  • 50th percentile: 500,000 views.
  • 75th percentile: 635,000 views.
  • 90th percentile: 720,000 views.
  • 97.5th percentile: 800,000 views.

Sponsored Wikipedia editing

For background information, see my site page about sponsored Wikipedia editing

With help from Issa Rice and Ethan Bashkansky, I significantly scaled up sponsored Wikipedia editing operations. You can see a full list of people I have sponsored, along with links to their contributions I have paid for, here.

It has been (and continues to be) an interesting set of challenges in recruitment, training, motivation, figuring out payment, and negotiating the Wikipedia bureaucracy. Issa or I might write more about the experience in the future. The jury is still out and things will be clearer by the end of the next quarter. A few people who have contributed significantly to the project, and deserve special shoutouts, are Issa Rice (pages and payments), Sebastian Sanchez (pages and payments), Jesse Clifton (pages and payments), Ethan Bashkansky (primarily for recruitment efforts) (pages and payments as well as recruitment efforts) and Alex K. Chen (pages and payments).


For background information, see my site page about WikiHow

I wrote one new WikiHow article: How to Understand Your Website Audience Profile. The article got a lot of views in the first week (getting to about 800 pageviews), likely because it was featured prominently on various landing pages and lists of top articles. Its growth has subsequently slowed down to between 5 and 10 new pageviews a day. The draft history (before I pushed it to WikiHow) can be found by looking at the history for the version in my personal Git repository.

The post is the first of many I intend to write related to understanding and interpreting trends and patterns, with an initial focus on the web.

I also edited some of my older WikiHow articles, including How to Gauge the Popularity of a Topic Online. I added and fixed up pictures for some other articles.

In total, pages I have created on WikiHow over my lifetime have about 9000 pageviews, and are getting between 1000 and 2000 additional pageviews every month. I had started editing WikiHow in May 2015.

Other written content

In response to charity evaluator GiveWell’s May 13, 2016 report on their annual traffic and money moved in 2015, I wrote a post reviewing my own past estimates of money moved, and discussing some implications.

I finally got around to publishing the blog post High-skilled hacks: a (very) brief overview of H-1Bs (more to follow). The post built on my reading and research into high-skilled work visas. I hope to do many other posts in the broad area.

Personal server migration

I began but have not fully completed the migration of all my websites to a cleaner Linode installation. The change will make it faster to quickly spin up a new Linode and transfer the key content to that. Basically, I am moving my server away from the snowflake server it had become in the last few years, and standardizing some of the procedures.


Two Korean dramas I have been watching in the last few months are The Flower in Prison (Viki) and Doctors (Viki). I’m watching both on Viki. An older drama that I binge-watched was Neighborhood Lawyer Jo Deul Ho.

I can now understand much of the Korean vocabulary used in the dramas, though these are probably unrepresentative of real-life Korean. Often, I watch an episode without subtitles and get most of the gist of it, but I miss a few nuances that I later fill in by watching with subtitles.

If you are interested in getting some drama recommendations, check out Evelyn Lee’s mid-2016 review.

I am also cancelling my ErosNow subscription since I don’t get a lot of time to watch Bollywood movies, and the smaller chunks of Korean dramas (1-hour episodes) make them fit better with my time.

Some background beliefs (July/August 2016)

I have a number of background beliefs about the world that, while not particularly rigorous, provide a sort of lens through which I understand the world. I thought it might be an interesting exercise to articulate those beliefs.

Some of these are general beliefs about the world, while others are beliefs about how things will unfold in the next few years (over timescales ranging from a year to a couple of decades). In the jargon of this post of mine, I am talking of the near but not very near future, as well as the medium-term future.

A slow-but-steady and low-variance near future

My overall view of approximately the next 10 or 15 years is: no major disruptions in either direction, but very slow and very steady improvement in the developed world, and faster catch-up growth in the developing world. I expect growth rates in the developing world to steadily decrease but still stay high enough for meaningful catch-up to be experienced along a number of dimensions where the developed world is more developed (and probably some where it isn’t!). Some of the recent sources of rapid growth are running out, but consolidation and utilization of existing advances still has some ways to go.

Probably the most exciting thing that will happen in the next few years is continued rollout of autonomous technology in contexts ranging from cars to trucks to factory to managing other machines. Some of it will rely on particularly impressive technological breakthroughs, but most of it will be a result of just more data and people getting around to implementing all aspects of something they already sort-of knew how to do. Raw effects of population growth and increased adoption of existing technologies will be negligible in developed countries, and reasonably solid but not revolutionary in developing countries.

End of Moore’s Law

Moore’s law and many of its variants are naturally ending. In a LessWrong post, I discussed the importance of both demand (at the right level) and supply in maintaining the pressure on technological progress. I think that we are running out on efficiency gains on raw hardware improvements, squeezed from both sides: a lot of hardware is good enough and cheap enough now, so people aren’t demanding the next release that much. And the technical challenges to maintain continued hardware improvements are getting harder. The story holds even for Internet speed. The Internet in developing countries is now fast enough to play streaming video, and most people have sufficiently reliable connections, so there’s not a lot of pressure to improve (increasing Internet access and reliability in far-off locations are continuing challenges).

Progress will continue in hardware, but the progress rate will move down to the progress rate we see in other kinds of gadgets, such as washing machines and refrigerators. New ideas and tweaks will continue to be applied, but people will have fewer and fewer compelling reasons to upgrade just because it’s been a couple of years since their last purchase.

While end users may be close to achieving full satisfaction with their computing devices, those who run servers at large scale will continue to push for improvements. I suspect many of these improvements will come from steady improvements in the software, as well as more automated management of existing server infrastructure, through smarter application of things ranging from autoscaling to wider use of spot markets to more tiered storage so that frequently accessed stuff is readily available and less accessed stuff is stored less accessibly. I expect that the server side will see larger year-on-year growth through these mechanisms.

End of Internet growth

I’m a big fan of the Internet, but I’m not bullish on it experiencing the kind of explosive growth it has averaged over the past two decades. I’d say there were three drivers of Internet growth.

  1. People adopting the technology through peer effects and increased familiarity, as it became better and more comprehensive.
  2. People becoming wealthier and therefore more able to afford the technology.
  3. Younger generations who grew up with the technology replacing older generations.

In developed countries, (1) and (2) are almost wholly exhausted. You just won’t find a 30-year old in the United States who would be a great fit for using the Internet but just hasn’t yet gotten around to trying it out! (3) probably still has some juice left, but it’s a very slow-moving process and in any given year it may be barely discernible. In fact, home broadband use has declined in the last couple of years in the US, as people realize they can get by just with a mobile Internet connection for all their needs.

In India and China, all three continue to be strong effects, though it’s possible that things will peak in China pretty soon and in India in a couple of decades. Mobile usage in Africa will also continue to grow and probably lead to huge growth in overall Internet use, perhaps like a lagged version of the growth seen in India.

Population growth not a big driver

The role of population growth in pushing up gross world production just through force of sheer numbers is going down. Population growth rates are declining as fertility in some of the most populous countries is below replacement. Previously, workforce populations were growing due to a lagged effect of fertility 15 years ago, but now it’s time to start living with a close-to-constant population size. This is pretty much true in most of the developed world, and a large part of the developing world.

Investment slowdowns

Investment will be slower in the next decade or so, as some of the drivers of growth in raw numbers (automatic technological progress, population growth, etc.) are slowing down.

Internet + real world connect better

Just as the restructuring of the economy around the automobile in the United States really started at about the time that autombile technology growth slowed down to a slow-but-steady pace (sometime in the 1950s and 1960s) we’ll see something similar with the Internet. Even as people’s raw time on the Internet will basically plateau, the efficiency with which that use is connected to the meatspace world will improve. People won’t be spending more time on social networking and work networking and email, but it will probably be more connected with and decisive in the real-world things they accomplish. On-demand services will continue to grow, albeit they probably won’t be the dominant consumption mode for stuff (for instance, people will use things like Uber and Lyft, but these will still form a small share of transit options).

This kind of qualitative progress will be hard to quantify, except from the vantage point of a specific service (e.g., if you are at Uber, you can measure the increase in ride volume). Generally speaking, it will be slow.

Autonomous vehicles

I expect that autonomous vehicles of some sort will become more important, but adoption will be slow as costs are initially high. Autonomous trucks, partly-autonomous cars, highway fleets, etc. might be some things that will try to gain traction. Explosive growth probably won’t occur in the next ten years, but perhaps by 2025 it’ll be clear that it’s poised to take off soon.

Epistemics around political and social issues will remain dismal, and the Internet will let good and bad ideas alike proliferate

The quality of average political discourse on the Internet will continue to fall, and this is probably a side-effect of the Internet mirroring real life more. Previously, gatekeepers controlled political discourse a lot more on the Internet. Ordinary people will feel bolder, and ideologically diverse movements will rise and ebb quickly.

Those who seek out high-quality discourse will be able to find it, and high-quality discourse will continue to grow in raw quantity and in the network of people who are aware of it and plugged into it. But the openness of the Internet will lead to frequent corruption and difficulty in maintaining it. On the plus side, this means memes from high-quality discourse might occasionally contribute a little bit to mass movements (by helping push them forward, or stopping them).

Cross-region cultural consumption will grow slowly and steadily, aided by greater content availability and broader social networks

Global cultural consumption, whether it’s cuisine or TV shows, will continue to grow. Not explosively, but slowly and steadily. Hopefully, this will lead to a world where people are more inclined to peace (it’s harder to support invading a country after watching a TV series or movie created there). The overall effect will be small, since the biggest creators of global culture are probably not on the invasion target list anyway (they are either highly advanced countries like South Korea, Japan, and the United States or big countries like India).

Healthcare and education will stay pretty much the same in developed countries, and most improvement in developing countries will be centered around infectious disease and sanitation

No surprises here, I guess!

February to April 2016 in review

Last quarter, I switched my review frequency to quarterly, in the hope of being able to provide more substantive updates in each review. Here I am, with my second quarterly review, this time covering the months of February to April 2016.

Things continue to be busy on the job front, but they are also more stable. In addition, I attended to some personal bureaucracy, including submitting my tax return and getting my passport reissued. With most of these items out of the way, I have started focusing more on various long-term projects, and hope to have a lot to report next quarter.

Wikipedia editing

For background information, see my site page about Wikipedia

I created 8 pages in February (the somewhat increased page creation count is because I took a total of four weekdays off that month, in addition to weekends). I created 2 pages in March and 2 pages in April. In total, I created 12 pages in the quarter.

As I mentioned in my previous quarterly review, the pages I am creating right now are not as optimized for pageviews as pages in the past have been. Rather, they are pages in domains where I want to acquire a deeper understanding and believe the research process involved in creating the page will help me. As was the case last time, my focus remains on issues surrounding migration. In fact, of the 12 pages I created in the quarter, only one, namely the page on Amazon Underground, was unrelated to migration.

In addition to creating pages, I also significantly edited and expanded some existing pages, most of them in the domain of migration. Some of the pages I significantly edited in this quarter were: F visa, I-20 (form), H-2A Visa, and Optional Practical Training.

In total, pages I created over my lifetime got 574,352 pageviews over the quarter. You can see the data for all of 2016 (including January, not included in the quarter) here. This is between the 50th percentile estimate (495,000 views) and 75th percentile estimate (640,000 views) that I made in my previous quarterly review. The main reason they were higher than my median estimate was that one page (Park Yeon-mi) got a lot of traffic in March due to related news coverage.

Forecast for future impact: I expect the pageview counts to be roughly similar to the previous quarter: a baseline of around 165,000 pageviews per month, plus a possibility of minor spikes because of some pages being topical. I have reduced the variation in my distribution since I am now more confident of the stability of pageview counts. The distribution is skewed to the right because of potentially huge traffic as a result of some events being highly topical.

  • 2.5th percentile: 350,000 views.
  • 10th percentile: 400,000 views.
  • 25th percentile: 460,000 views.
  • 50th percentile: 520,000 views.
  • 75th percentile: 645,000 views.
  • 90th percentile: 720,000 views.
  • 97.5th percentile: 800,000 views.

Sponsored Wikipedia editing

For background information, see my site page about sponsored Wikipedia editing

Through the quarter, I sponsored Wikipedia page creation by three people:

  • Alex K. Chen (updated page list here): Alex has worked on a large number of pages, including technology company timelines, as well as angel investors and venture capitalists and their funds. In addition to the pages I pay him to create, he also creates a number of other pages.
  • Issa Rice (updated page list here): After completing some pages on taxes and migration, Issa has shifted focus to global health and poverty.
  • Ethan Bashkansky (updated page list here): Ethan has continued to do sporadic work on tax forms.

In April, I announced on the Effective Altruism Forum that I was looking for people interested in Wikipedia editing. As a result of this announcement, I got two new people who have contributed significantly and expect to continue contributing for the next few months:

  • Jesse Clifton (updated page list here) is primarily working on topics related to animal welfare and rights. His first article was a timeline of animal welfare and rights. He has subsequently been working on articles on animal welfare and rights by country. I hope to share more exciting details of his work in my next quarterly review.
  • Sebastian Sanchez (updated page list here) is currently working on topics related to global health. I hope to share exciting details about his progress next quarter.

There are also other editors who are getting started with editing, and I expect to see some of them (and/or other recruits) make further progress during the quarter.

Subject wikis

For background information, see my site page about the subject wikis

I did active work on the subject wikis for a few segments of a few hours, but not in any systematic way that would affect the overall level of content or change the experience of typical visitors. The subject wikis have been doing well traffic-wise, but this is not due to any additional effort on my part.

Groupprops got a total of 261,975 pageviews this quarter compared to 244,136 for the corresponding quarter last year. The seasonal pattern in traffic was very similar between quarters, so much so that it is visually striking when you look at the graphs in Google Analytics. Market declined (47,781 compared to 65,367) and Calculus stayed stable (60,241 versus 58,499).


For background information, see my site page about WikiHow

I wrote one new WikiHow article: How to Avoid Crowds in Mass Transit. The article has received 491 views so far. I estimate that about 200 of those views were driven by my Facebook share of the article, and another 50 were a result of initial publication, plus my own views. The remaining views are probably organic.

Two of the WikiHow articles I have created over my lifetime now have over 1,000 views: How to Use Git Effectively (at 1,467 views) and How to Ride BART from the East Bay to San Francisco (At 1,362 views).

Other written content

I finally got around to publishing the blog post How did we get here? Chinese Exclusion Act buildup (1848-1872). The post summarized my understanding of the historical backdrop that built the momentum for the Chinese Exclusion Act in the United States.

I also worked on a few other drafts. I expect to publish some of these drafts this quarter, as well as start on (and possibly finish) some new posts.

I also posted to the Effective Altruism Forum to recruit Wikipedia editors.


I finished watching a few Korean dramas on Viki as they wrapped up: Madame Antoine, Descendants of the Sun, and Marriage Contract. In addition, I watched one older Korean drama: City Hunter.

I expect to be spending less time watching Korean dramas going forward, as I start picking up the pace on some more long-term projects during the weekends, now that some of the bureaucratic activities are completed.

November 2015 to January 2016 in review

After October 2015, I decided to switch away from monthly reviews to quarterly reviews. Depending on whether I feel I have clearer, more substantive updates at a quarterly level, and whether my readers agree, I will maintain this review frequency. This post reviews the past quarter (November 2015 to January 2016) while also offering a preview into the current quarter (February to April 2016).

Things have been busy on the job front in a variety of ways, but are now getting more stable (our company hired a bit more). I expect to not get a lot of time for the other activities but I do expect that the time I get for these will come in a more steady stream than it has in the past few months.

Wikipedia editing

For background information, see my site page about Wikipedia

I made one page in November, two pages in December, and four pages in January. You can see the lists of pages I created by month using the tags in Wikipedia Views. In particular, you can get the data at these respective links for the three months: November, December, and January.

I intend to create somewhere between 3 and 10 pages a month for each of the months of February, March, and April. So far in February, I have already created 5 pages, and have some more drafts ready to be pushed out.

The pages I am creating right now are not as optimized for pageviews as pages in the past have been. Rather, they are pages in domains where I want to acquire a deeper understanding and believe the research process involved in creating the page will help me. At present, my focus is on key aspects of migration laws, migration forms, and migration procedures in the United States. I intend to branch out more to other countries and to return to migration history (that I worked on in October and November) after finishing up a few more articles on migration procedures.

I do still hope to facilitate the creation of pages that are generically popular and underprovisioned. In recent months, I have successfully done so through sponsored Wikipedia editing. The people I have been paying include:

  • Alex K. Chen (updated page list here): Alex worked on timeline pages in January and has been working on pages about investors starting February.
  • Issa Rice (updated page list here): Issa created many great pages on important forms related to taxation in the United States, and is now transitioning to work on migration forms. See for instance the pageview counts in January for pages Issa created in January.
  • Ethan Bashkansky (updated page list here): Ethan is starting out with tax forms.

Measurement changes: With some help from Issa, I switched my data source for Wikipedia Views from (which had stopped updating) to the Wikimedia API, which is probably a better and more stable option going forward. One consequence of this change is that it is now possible to exclude obvious bots and spiders from the view counts, and data from January 2016 onward excludes those numbers. Hence, there is an artificial drop from December 2015 to January 2016 arising directly as a result of that exclusion. The drop mostly affects the “long tail” pages that get very few monthly views.

Impact: Here are the pageview counts for pages I have created over my lifetime in the three months of the quarter: 180,352 in November, 196,227 in December, and 172,159 in January. The total across the three months was 548,738. New content I created has gotten under 1000 views per month. The drop from December to January is explained by the data source change. The low pageview count of newly created pages is explained by the focus on maximizing my learning rather than creating pages with a huge amount of interest.

Forecast for future impact: I expect the pageview counts in the next quarter to come out a little lower than the previous quarter, primarily due to the data source change. My estimate range for pages I created is as follows:

  • 2.5th percentile: 275,000 views.
  • 10th percentile: 340,000 views.
  • 25th percentile: 410,000 views.
  • 50th percentile: 495,000 views.
  • 75th percentile: 640,000 views.
  • 90th percentile: 725,000 views.
  • 97.5th percentile: 880,000 views.

The estimates are a little lower than the past because of the bot-filtering, as well as the shorter length of the forthcoming quarter: 89 days as opposed to 91-92 days for most quarters.

The dispersion of the estimates is similar to that for a month. This is because the greater centrality arising from summing up over more data points is balanced out by the greater uncertainty arising from looking farther out into the future.

Subject wikis

For background information, see my site page about the subject wikis

I did active work on the subject wikis for a few segments of a few hours, but not in any systematic way that would affect the overall level of content or change the experience of typical visitors. The subject wikis have been doing well traffic-wise, but this is not due to any additional effort on my part.

Groupprops got a total of 254,100 pageviews this quarter compared to 232,786 for the corresponding quarter last year. The seasonal pattern in traffic was very similar between quarters, so much so that it is visually striking when you look at the graphs in Google Analytics. Market declined (47,429 compared to 58,823) and Calculus stayed stable (56,258 versus 54,466).


For background information, see my site page about WikiHow

I wrote two new WikiHow articles: How to Use Amazon Route 53 and How to Use Git Effectively. The latter of these articles appears to be the fastest of my WikiHow articles in acquiring pageviews, though it’s still early days to know just how far it will rise.

EA Forum blog post making forecasts about GiveWell money moved

In December 2015, I published a blog post in the Effective Altruism Forum with an analysis of the amounts of money moved to GiveWell top charities and forecasts about money moved for 2015, based on the information available at the time. I also published a follow-up blog post and some comments on my own posts. The actual money moved turned out to be at around the 90th percentile of my estimate, as I noted in a later update at the bottom of the post.


I have continued to watch the Korean TV series Remember: War of the Son on Viki, which is finishing up in February. Other Korean TV series I have been watching include Madame Antoine. Thanks to these TV series, I have picked up many words and phrases in Korean, and can generally understand a decent fraction (excluding some nuances) without subtitles if I am also watching the video alongside (so I can see people’s facial expressions and other contextual cues).

I also watched the Bollywood movie Jazbaa, which I later discovered was a remake of a Korean film Seven Days (something that is not quite surprising, given that some of the themes seemed distinctly reminiscent of the Korean TV series I had watched).

I also found a blog, E Drama Clouds by Evelyn Lee, that offers high-quality commentary on Korean TV series.

On the music front, I bought a few MP3 songs and listened to some more. My recent purchases include Hello, Wake Me Up, and Dark Horse.