Annoying iOS Copy-Link Bug


No space between text and URL in Copied/Shared links in any application

Steps To Reproduce

1) Open Apple News app
2) Go to any app and click the More Options menu
3) Click Copy to get a link of the article
4) Paste the link in Messenger/Telegram to share the link to someone
5) Verify the shared text commingles the URL and the last word in the shared text. Example – Missed out on the bitcoin boom? Try these virtual currencies instead – NBC News

Expected Result

Space between shared text and URL. Example – Missed out on the bitcoin boom? Try these virtual currencies instead – NBC News

How to Not Trade Crypto Like Jym

There are 3 ways to decide how you want to invest. Profitability and difficulty as well as risk is different for each one but worry not I shall guide you.

If your portfolio size is <0,5 BTC, you should invest solely on etherdelta, cryptopia and other tiny exchanges. Research coins that either JUST got added or got added recently, or were fomo’d and abandoned. What matters there is SHILL POWER. Look for coins that fill all these requirements:

  1. Small marketcap (below mil is the best, below ~3 is okaysih)
  2. Low total supply, majority must be in circulating supply (85%)
  3. Have a solid team or idea or advisors
  4. Have burn/exchanges announcements or rumors going around

Anything extra shillable is great. Once you buy, talk about the coin on telegram groups and shill it like your life depends on it. Post often with cashtags and talk about the burn/whatever there is to attract investors and potential marketcap. Don’t bullshit though or you’ll hurt both your reputation and the coin.

What’s the most important here is TAKING PROFITS before the price goes to shit.

Don’t expect a 500%, take the 250%.

I advise you to all in if your portfolio is below $1000, your portfolio is tiny, you need to grow, I started the same way.

If your stack starts getting too large to sell/enter trades there with, consider climbing up the second step. If  <20 BTC, you should diversify between multiple solid projects on Binance and by the time you go there you should have already learnt to do proper research. What you need to understand now is market demands (this is why we have a group) or you’ll get stuck holding forever with your portfolio growing at a slower rate than btc. Don’t hold your whole stacks, when you see a coin went 10-20% up and down multiple times, set sells and buys at these prices. Go with the market, not against it.

I strongly advise you divide your stack between 2 holds that you swing and 3 trading stacks at max that you use to go into pumps/breakouts.

If you entered a quick trade and it didn’t work out, don’t be a retard, CUT YOUR LOSSES, don’t trade emotionally. Don’t panic. Always have a strategy before you jump into a trade, always keep some spare BTC.

Always keep a few ethers to invest in possible 100xers on etherdelta/cryptopia for longs. 5% of your stack at most.

Keep in mind you don’t need to be an expert on TA, use trendlines, support, resistance, volume and bullish/bearish indicators.


I present you the most valuable material of what i stumbled upon which i’ve chosen to study trading/investing, mandatory watch, you need to see how successful people operate, act and think

An Aside…

Jim Rohn – dude’s gone from 0 to 100 by changing his perspective of life and surrounding himself with the right people, good watch

Anton Kreil this guy is massive. Nuts daytrader + teacher, my personal favourite, in some of these videos you will see him teaching complete newbs, you should focus on what mistakes they make and avoid it.

Anton Kreil v2

Anton Kreil v3

Anton Kreil v4 watch the whole series- 3 1hour videos of lex vandam and anton kreil turning noobs to pros, spoiler noobs end up outperforming their own hedge fund

Warren Buffet tons of gain and nothing to lose by hearing the thoughts of one of the greatest investors of our time, he did what we do now in crypto, find undervalued assets (hodlers relate but dont ill talk about it later)

Robert Kiyosaki more of an offtopic but dude’s gained wealth by being beneficial to others and taking smart loans, i reccomend you read all his books too, if you have children teach them what he teaches or give him his books, financial knowledge is absolutely MANDATORY for everybody who doesnt want to stay dirt poor trading or not

Why Zilliqa

Here I’d like to share with you my own thoughts to quickly differentiate Zilliqa from other blockchains as the first line of filters before you raise the comparison questions in our slack channel.

*Scalability does not only mean high throughput but also indicates the large scale of the network.* To compare a blockchain to Zilliqa, you can go through a few questions listed as following.

*1) Is it a permissioned or permissionless blockchain?* As we know, lots of blockchains or new consensus algorithms claim they can process over 100, 000 transactions per second (tps), but they are running in a permissioned setting. This implicates that only registered or privileged nodes are allowed to join their systems, and it does not support open membership. In the real world, lots of banks and big companies are working on this direction to meet their internal requirements, e.g., hyperledger. In contrast, Zilliqa is a public blockchain providing open membership to accept new nodes.

*2) How many nodes do they support?* Another important metric for scalability is the scale of their network. In the permissioned setting, high throughput is relatively easy to achieve, as it only requires a few nodes say a hundred nodes to reach consensus. However, it typically does not scale to a large network, as the underlying consensus protocol does not run effectively when the network size becomes large. The root cause is that different variants of BFT protocols can incur at least O(N^2) communication complexity. For example, assume a message size is 100 bytes, if the network size is 100, at least 1MB (100* 100^2 = 1,000,000 bytes) are transferred, but if network size is 10000, 10,000,000,000 bytes (10 GB) are transmitted to every node, which is unrealistic in the public setting. In contrast, Zilliqa is designed to be scalable. In Zilliqa, we first apply sharding to divide the whole network into relatively small shards say 600 nodes. Then we further employ Schnorr signature and multi-signing schemes to reduce communication complexity from O(N^2) to O(N), which significantly improve the performance of our consensus protocol. Therefore, Zilliqa can survive and scale in the public setting with thousands of nodes in the network.

*3) Does it use PoW or PoS?* PoW is energy-inefficient, but the Nakamoto consensus protocol derived from it can effectively resolve the consensus issue, i.e., choosing the longest chain as the confirmed one with agreement. PoS is energy-efficient and helps the network to elect a consortium of representative nodes. Then these representatives can run consensus protocol like variants of BFT to agree on the same set of valid transactions. Typically the consortium size is small, e.g., 21 in EOS and 100 in Ethereum Casper. We at Zilliqa believe that PoS is still a nascent idea. On the contrary, PoW is a very well understood mechanism which has survived the test of time. Having said that, the use of PoW in Zilliqa is very different from the way it is used in say Bitcoin. This ensures that PoW does not become a bottleneck for scalability in Zilliqa. For the detailed explanation, please refer to Q11 in our FAQ.

Yik Yak Auto Downvote Script

For research purposes… ¯\_(ツ)_/¯

window.scrollBy(0, 10000);

function sleep(ms) {
return new Promise(resolve => setTimeout(resolve, ms));
async function downvote() {
var L = document.getElementsByClassName(“downvote”);
var i = 0;
for (i=0; i<L.length; i++) {
await sleep(2000);
window.scrollBy(0, 10000);


I play the game for the end-game music. True story.

Overwatch OST - Play of the Game     

Installing Hadoop 2.6.X on Raspberry Pi B Raspbian Jessie

Configure Java Environment

With the image Raspbian Jessie image, Java comes pre-installed. Verify by typing:

java -version

java version "1.8.0"
Java(TM) SE Runtime Environment (build 1.8.0-b132)
Java HotSpot(TM) Client VM (build 25.0-b70, mixed mode)

Prepare Hadoop User Account and Group

sudo addgroup hadoop
sudo adduser --ingroup hadoop hduser
sudo adduser hduser sudo

Configure SSH

Create SSH RSA pair keys with blank password in order for hadoop nodes to be able to talk with each other without prompting for password.

su hduser
mkdir ~/.ssh
ssh-keygen -t rsa -P ""
cat ~/.ssh/ > ~/.ssh/authorized_keys

Verify that hduser can login to SSH

su hduser
ssh localhost

Go back to previous shell (pi/root).

Install Hadoop

Download and install

cd ~/
sudo mkdir /opt
sudo tar -xvzf hadoop-2.6.4.tar.gz -C /opt/
cd /opt
sudo mv hadoop-2.6.4 hadoop
sudo chown -R hduser:hadoop hadoop

Configure Environment Variables

This configuration assumes that you are using the pre-installed version of Java in Raspbian Jessie.

Add hadoop to environment variables by adding the following lines to the end of /etc/bash.bashrc:

export JAVA_HOME=$(readlink -f /usr/bin/java | sed "s:bin/java::")
export HADOOP_INSTALL=/opt/hadoop

Alternative you can add the configuration above to ~/.bashrc in the home directory of hduser.

Exit and reopen hduser shell to verify hadoop executable is accessible outside /opt/hadoop/bin folder:

su hduser
hadoop version

[email protected] /home/hduser $ hadoop version
Hadoop 2.6.4
Subversion -r 1503152
Compiled by mattf on Mon Jul 22 15:23:09 PDT 2013
From source with checksum 6923c86528809c4e7e6f493b6b413a9a
This command was run using /opt/hadoop/hadoop-core-2.6.4.jar

Configure Hadoop environment variables

As root/sudo edit /opt/hadoop/conf/, uncomment and change the following lines:

# The java implementation to use. Required.
export JAVA_HOME=$(readlink -f /usr/bin/java | sed "s:bin/java::")

# The maximum amount of heap to use, in MB. Default is 1000.

# Command specific options appended to HADOOP_OPTS when specified

Also, we need to edit “”. Uncomment


and write:
export YARN_NODEMANAGER_OPTS=”-client”

Note 1: If you forget to add the -client option to HADOOP_DATANODE_OPTS and/or YARN_NODEMANAGER_OPTS you will get the following error messge in hadoop-hduser-datanode-node1.out:

Error occurred during initialization of VM
Server VM is only supported on ARMv7+ VFP

Note 2: If you run SSH on a different port than 22 then you need to change the following parameter:

# Extra ssh options. Empty by default.
# export HADOOP_SSH_OPTS="-o ConnectTimeout=1 -o SendEnv=HADOOP_CONF_DIR"

Or you will get the error:

connect to host localhost port 22: Address family not supported by protocol

Configure Hadoop

In /opt/hadoop/conf edit the following configuration files:







Create HDFS file system

sudo mkdir -p /hdfs/tmp
sudo chown hduser:hadoop /hdfs/tmp
sudo chmod 750 /hdfs/tmp
hadoop namenode -format

Start services

Login as hduser. Run:


Run the jps command to checkl that all services started as supposed to:


16640 JobTracker
16832 Jps
16307 NameNode
16550 SecondaryNameNode
16761 TaskTracker
16426 DataNode

If you cannot see all of the processes above review the log files in /opt/hadoop/logs to find the source of the problem.

Social Web Readings and Commentary

Herein lies a text dump of the reflections I did for my Social Web class. Meh.


I found the article confusing at first because I lacked the context of what MOOs or MUDs were. Once I read up on those, I realized that what this article in summary was the shift in decision making power in the game to the majority, as well as the creation of tools to police its users.

This reminds me of the game Twitch Plays Pokemon, where players could determine what the game character does by pressing key combinations. The program would process each player’s key in sequence. The task then was to see if the internet hivemind could complete the game even with the chaos that ensued. It took some minor updates to the methodology before it was completed.

Considering this in the context of LambdaMOO, it could be said that the the viewpoints of the majority can be considered the most amenable in any dispute or decision-making process. This is why the final choice was made to implement a voting scheme in-game to carry out tasks which require administrator rights. The implied notion was that concentrating power in the hands of a few people was not desirable. In addition, the administrators did not want to arbitrate in any more disputes.

In sum, making a decision such as kicking out a player is difficult to do so in a personal capacity. However, once you have the majority of people agreeing to your decision, it is easier to do so. And this is what LambdaMOO became.

I find the tables in the paper a challenge to make sense of. However, from the graphs that follow, the main takeaways from the paper are that high editor concentration leads to greater improvements in article quality. However, as time passes and the article improves, more power needs to be relinquished to the masses because ostensibly, the latter does a better job in the maintenance of the article. This is new to me because I have always thought more people involved would always lead to better outcomes. I also like the longitudinal study done for the “Music of Italy” article as it shows the activity of certain groups of users during the infancy of the article.

This article was novel because it postulated a model for conflict detection and resolution. I think it is eye-opening to see how researchers think of quantifiable statistics as proxies for issues like maintenance work (like reverts) and conflicts (Controversial Revision Count).

Also, I think their work on the Revert Graph opens up a lot of possibilities for improving Wikipedia moderation tools. For example, such a tool can be run on each article, giving everybody an overview on the possible bias that they might subconsciously take on the article. Since Wikipedia prefers a nuanced perspective in article writing, extra caution might be given to users who fall on extremes of the Revert Graph. Also, an alert can be given to users who have a biased perspective (based on their reverts) before they commit any changes to the article.


week 5

The Effect of Online Privacy Information on Purchasing Behavior: An Experimental Study

The conclusion of the study is that “people will tend to purchase from merchants that offer more privacy protection and even pay a premium to purchase from such merchants.” However, as seen from Figure 4 on page 15, there is also a substantial number of people who ignore the presence of privacy cues. For the case of sex toys, with privacy information disclosed, more than 30% of participants bought from the lowest priced vendor, comparable to the most privacy-protecting (highest-priced) vendor. This shows that price is still a strong predictor if a buyer is going to purchase from a vendor, irregardless of privacy concerns. Finally, I feel that this study is at best, unrealistic in the real world. P3P is all but forgotten on the internet and only Microsoft Internet Explorer supports it. If you inspect the network data when you surf Facebook, there is a header that says ‘p3p:CP=”Facebook does not have a P3P policy. Learn why here:’ – Facebook says that the P3P standard is now out of date.

I regretted the minute I pressed share: A qualitative study of regrets on Facebook.

This study looks into the different types of posts that Facebook users are likely to make and then regret. The main types revolve around posts made in the heat of the moment, associated with undesirable activies as well as those that were not intended to be seen by a specific group of people. A useful extension to the study would be to analyze posts made on Facebook that were subsequently deleted. I cannot confirm but it is likely Facebook has a copy of all posts made, even though it was subsequently deleted, as some (undisclosed) information is deleted only when one permanently deletes one’s account. From this post history, users can be asked the reason why they deleted certain posts, and more accurate data can be gleaned.

Jason Hong: Smarter Phones

In his talk, Hong surmises that privacy concerns arising from the use of smartphones are mainly due to users not being aware of the things that smartphone apps are doing to their personal data such as location and device ID. To dispel such concerns, users should be made more aware of the details each app is requesting. Personally, I feel that a context-sensitive alert about the requests for data each app is using would be beneficial. For example, when downloading Google Maps from the App Store, a prompt that this app would be using location data would not be shown because most users know that this would be needed. However, if a bible app requests for location data, a visible prompt would be shown that such location data is probably not needed for proper functioning of the app. A possible way to get information on what app requires what kind of data to work properly would be to crowdsource information from users. For example, what percentage of users deny the app’s request for data – if the number is high, it is highly possible that users do not think that such information would be necessary.

Making Sense of Privacy and Publicity

In her talk, Boyd points out that privacy is alive in the sense that people do care about it. Privacy is about having control over how information flows. However, companies fallaciously think that people do not really care about privacy because the latter discloses their personal information on the online public sphere, often unintentionally. She then proposes that companies do not egregiously change privacy settings on behalf of users, or ask users to opt out of a possibly controversial feature rather than opt in. She also adds that people takes the public sphere into account when sharing information. When they feel comfortable, they tend to share more, often unknowingly. The onus is then on the company to ensure they are well-informed.

This brings to mind a feature that Facebook implemented a few years ago with the introduction of the Activity Feed called Beacon. Facebook intended to extend the social graph from the Facebook site to what users do on other sites. When someone gets a movie from Fandango, this activity would pop up on the Facebook feed. There was great controversy when users pointed out it would be problematic when someone buys a book called “Coping with Cancer” and the purchase was posted on Facebook for all to see.



Facebook’s privacy policy is located at and it is a refreshing change from the legalese that other websites have. Of course, Facebook has an offical one ( that overrules the more interactive version.

It is grouped into 3 main parts – what information is collected, how Facebook uses this info, as well as how they share such data

In the first part, the main types of data being collected are listed. The more notable ones are data about the frequency and duration of the user’s activities, which are possibly used to determine who the user is engaging with. This might be scare some people who are not used to such a granular level of tracking. According to one website (, Facebook is even tracking mouse movements and clicks, in a bid to improve user engagement via A/B testing. Another data of concern is those from third-party partners. Althrough phrased in a nuanced way, these can include data brokers who seek to link your offline transactional data with those that are conducted online via the myriad Facebook buttons on countless websites these days.

In the second part, they clarify that the information is used to improve their services. The one that is of note here is the part about info being used to show and measure ads and services. They have an entire page dedicated to ad privacy and explanation – which is also written for the layman. They also mention cookies, but users might not know what they are. A possible way to improve in this aspect is to call is “cookies or tracking beacons” to make know to the user the intent of said cookies, that is, tracking.

Finally, the longest and most detailed portion is the one on information sharing, and rightly so. It is grouped into two parts – sharing on Facebook and sharing with external parties. For the first part, it covers most of what we know about Facebook sharing. When certain Facebook features are used such as Facebook Login, your public info is shared with them – this includes your username or user ID, your age range and other information you choose to share. They are also allowed to share information with other companies owned by Facebook. The second part covers sharing with third-party partners which include advertising agencies as well as research partners. While the former has language that only “Non-Personally Identifiable Information” will be used for the purpose of advertisement targeting and tracking, the latter does not have such information. In the latter case, such partners “must adhere to strict confidentiality obligations in a way that is consistent with (said) Data Policy”. However, this means that this could be the weakest link in the chain that could deanonymize users. If said vendors or partners not adhere to Facebook’s policies, private information that could identify users could be leaked out. This is the most seriously lacking part of the whole policy page.

Overall, I think Facebook’s privacy policy is presented in a very succinct and easy to understand manner with great use of subheadings to delineate each portion. Links to interactive help tools are provided allowing the user to see how their public profile look like from a search engine, for example. Some of the information presented by Facebook is intentionally vague, possibly because it might be too technical for users to understand. Of course, a more conspiratory perspective is that Facebook is hiding something so that it can hide behind vaguely fleshed out terms.



One day companies might know you better than you know yourself.

It has already happened – Target is able to identify 25 products that can accurately ascertain if you are pregnant or not – and send you coupons before you even need them.

When you get a membership card from Target, they are not offering you discounts for free – they seek to link your real world purchases (credit cards used) with your online impulse buys. Some even sell or ‘exchange’ your anonymized information with so-called data brokers. However, research has shown that such anonymized info can pretty much identify a specific person with great accuracy.

One day, such tracking will be so pervasive and invasive that we will wonder if there was ever a notion of privacy.


Week 6

How to Ask for a Favor: A Case Study on the Success of Altruistic Requests

In summary, this paper concludes that showing gratitude, showing intent of reciprocity and an urgent need, as well as a higher status within the community, will increase the probability that one will have one’s request fulfilled. Mood, politeness and similarity were not significant factors.

Before, looking into the meat of the paper, I predicted that only gratitude, reciprocity and urgency were contributing factors. I did not find status to be particularly important because status indicators on Reddit require one to click through to the user profile; unlike forums where you can see how many posts or the ‘flair’ the user has. My guess is that because money is involved, people tend to see that the user was not just created to get free pizza from people. They would like to see an active user in Reddit – as a proxy for how much contribution said user has made in the community.

Investigating the Appropriateness of Social Network Question Asking as a Resource for Blind Users

I like this idea of crowdsourcing for social good. I downloaded the VizWiz app to see how it worked first hand but could not get it to work as the app keeps on crashing at some portions. As expected, the anonymity and speed of on-demand services like Amazon Mechanical Turk makes it a very compelling option to get crowdsourced services at a fee. I did some research on fraud rates on Amazon MTurk to see how much fraudulent completion of tasks were there. One possible reason why Facebook would not be effective is because Facebook’s EdgeRank selects what would be shown on a person’s newsfeed – if a user is not particular active on Facebook, or posts updates that are not particularly popular, their updates would be shown on the website as much. In recent months, there have been a shift towards contextual display of updates to people who demonstrate interest in a particular topic. For example, if I post an update about a new Pusheen Box, Facebook would show my post more to people who like Pusheen. Depending on perspective, this might be a good or bad thing to VizWiz – good because the posts would be more targeted to people with the knowledge to answer questions correctly – bad because the posts would see a more limited audience now. Personally, I feel that Facebook would be a less effective medium for VizWiz users because those users do not ask questions which require subject matter experts.


I had some experience with Tomnod – a volunteer crowdsourcing website that takes satellite imagery from DigitalGlobe to find things. One of their more notable ones that I signed up for was the search for MH370. They provided satellite imagery of the sea floor for people to identigy wreckage. For this task, I decided to try Tomnod again. I started off with the California Valley Fire 2015 campaign, which asks users to look for damaged buildings and impassable roads based on imagery before the fire as well as infrared imagery after the fire. I was pretty frustrated doing it because it seemed they hadn’t improved their infrastructure since I last used it. The site still lagged like crazy and the images refuse to preload, making for a very frustrating experience scrolling the huge expanse of land. My browser crashed a few times during this ordeal.

I switched to a different campaign – tagging Indonesia forest fires and burn marks. I spent around 1.5 hours on this campaign. It was easier to use than the previous one, probably because there is only 1 image to load for each tile. In my time there I explored 3455 tiles (~1700 sq km), tagged 245 objects and achieved consensus on 210 of them. Consensus meant the number of people who agreed with me on my tags. The other features of the site was progress updates disguised as a form of gamification. For example, when I explored 100 Km2, a notification pops up saying “100 Km2 Super Explorer!
You have explored through a full one hundred square kilometers!” I feel that the interface could be made better so users don’t feel irked by the clunkiness. Instead of showing the whole map, they could show the users one tile at at time. Also, they could potentially highlight areas where other users have highlighted to show examples of what is correct or wrong.

2nd Readings

Showing Support for Marriage Equality on Facebook

I read Malcolm Gladwell’s epic paper on slacktivism (Small Change) and his results are particularly relevant. Slactivism does not require much effort on the part of users, and that is why it is such an easy and popular way to support a cause. It is an effective way to get people to sign up for a cause. Moreover, such slacktivism harnesses so-called weak ties prevalent in social networks – people do not know each other personally on Twitter for example. To bring about real change, strong ties must be formed. This happens when we know someone personally taking part actively in a cause. When we know someone personally taking part, we are more likely to transition from slacktivism to actual activism.

In defense of “Slacktivism”: The Human Rights Campaign Facebook logo as digital activism

Slacktivism can be helpful if the people who ‘passively’ contribute from the comfort of their homes actually believe in it. However, as it is, many people jump on the bandwagon so that they are not left out, or they want to be seen as ‘liberal’ in the eyes of their friends. This isn’t particularly helpful. However, I surmise that companies treat slacktivism as a form of viral marketing – such word of mouth will further their organization’s aims through awareness – whether it makes a real impact there and then does not matter.


We Are Dynamo: Overcoming Stalling and Friction in Collective Action for Crowd Workers

The paper discusses the idea of galvanizing crowd workers to band together to further their rights, akin to a virtual workers’ union. Similar to how Uber labels its drivers as contractors, online workers have limited recourse. For example, on Mechanical Turk, requesters often mark work as unsatisfactory even though it might be done well, so as to reduce their cost. In their online community, stalling and friction served to undermine their efforts to establishing effective collective action. The solutions the authors propose is useful not only in collective action but serve as good general guidelines for a community.

The Future of Crowd Work

The paper discusses ways in which online crowd tasks sites could be made better. Much of the problems crowd workers face are that instructions from requesters are not clear, or they are not paid for tasks which they completed but the requesters declare to be not up to scratch. The authors propose creating a career ladder for such crowd work but I feel that this is not feasible. The state of micro-tasks are such that they are meant to be low-paying, non-skilled tasks. The fact that a lot of these involve liking Facebook posts, or even clicking on ads mean that such tasks are not very legitimate after all. 3 years ago, I actually did make my own micro-tasks website, modelled after, which aims to help workers out by reducing fraud performed by the requester, by implementing an API that automatically verifies the work done by worker, rather than leaving the requester to manually approve them.

2nd Readings

The Mirage of the Marketplace

I believe that Uber was so successful simply because it disrupted the taxi industry which had remained pretty much static for the past few years. Previously, taxis with their medallions were pretty much entrenched in the transportation industry in US as the medallions gave them a virtual monopoly. Moreover, in other countries, pricing tiers meant that taxis were not available at certain times of the day. For example, in Singapore, taxis would go into hiding 30 mins before the peak hour rates start, so that they can bide their time until those higher rates start. Instead of alleviating traffic jams, taxis increased congestion during those peak hour periods. However, Uber came in to solve this problem.

Having said this, Uber is simply using technology to faciliate transportation. Much like how eBay and Amazon used technology to get goods delivered from manufacturers to consumers, Uber works on the basis of delivering transport from drivers to riders. The ‘algorithm’ which the article says is a mirage, is simply used to better the rider experience by improving speeds and satisfaction, even though some cars were simulated.


I took a few Uber rides and asked them why they wanted to be an Uber driver. One of them said the flexibility appealed to him. He said he was a night person and liked driving passengers at night. He added that he earned more too, from fetching passengers on red-eye flights from PIT. When I asked if he minded not having any employee benefits, he shrugged it off and surmised that there had to be a trade-off. He added that Uber’s technology has indeed made connecting passengers to drivers easier. When asked about the future, he said he had no plans to change jobs unless legislation forced him to otherwise.

Another time, the Uber driver said that he intended to use his Uber earnings to get a foothold into the real estate business. His brother bought a few houses from his savings, rented them out (to sketchy people month-to-month) and has a stable income now. He said he does freelance web consulting work at home and whenever a potential passenger requests for a ride, he takes it up to get extra income. He also added that he does not like the rigidity a 9 to 5 job entails, and that was why he does freelance work – it even pays better. Finally, he said he hoped that Uber and Lyft instills some kind of community in its drivers – a lounge for other on-demand drivers to connect would be good.


Using social psychology to motivate contributions to online communities.

This reading discussses two experiments about overcoming social loafing – whether framing uniqueness and benefits, or goal setting are possible solutions. While uniqueness increases contributions, benefits do not, probably because users probably have subconscious reasons for doing so (ie. altruism) and feel that reminding them of benefits reduce their proclivity for contribution. I did not expect this result as from a web marketing perspective, one would remind potential users about the benefits of signing up. According to Kraut, users might have seen those messages as manipulative and acted opposite to their recommendations simply to preserve their autonomy.

The second study showed that specific goals increased user contributions. What I hoped they explored was whether a very low goal will reduce user contribution. Their lowest goal level was the median number – 8 ratings per week. What if they had intentionally put a very low goal of 2 ratings per week? Would it otherwise persuade people who just logged in and did not rate because 8 ratings were too effortful? Would it reduce the ratings of very active users? The conclusion of the study was that group work increased contributions (contrary to hypothesis and what we know about the collective effort model) and very high goals reduced the rate of contribution.

Understanding the motivations, participation and performance of open source software developers: A longitudinal study of the apache projects.

The paper talks about whether various factors influence use-value and status motivations, as well as level of contributions to the field of open source software. I was confused as to what use-value motivation actually meant (and still is) while I know status motivation refers to the motivation derived from knowing one’s status can be elevated in the future. The paper states that use value refers to the contribution to solve a problem of personal use benefit. Does this refer to a case where I contribute to a code project because a feature which I personally find useful was not present, hence my use-value motivation increases? In any case, a part which I found interesting was that being paid to contribute increases intrinsic motivation by virtue of increasing status motivation. This runs contrary to what I learnt in social psychology. When people are paid to do things, they attribute their continued action of that task to the money, and not an intrinsic reason. If one is not paid and continues doing the task, there is no extrinsic motivation (money), hence I will (perhaps falsely) attribute an intrinsic one subconsciously – I like doing it. Hence, it is posited that monetary incentive would reduce the motivation to participate.

Evidence-based social design: Mining the social sciences to build online communities.

Although a tad long, the paper was very insightful, with very actionable strategies for motiviating contributions online or offline. With the deluge of information online, it is easy to execute some of the ideas in the paper. For example, many marketing funnel pages employ such strategies by stating a limited time period (“you have 3 hours left to sign up”) or showing that other people have joined, or requests that target people’s interests. As a corollary to the previous reading, this paper states that reward increases contribution if it is a sufficiently high amount; it can be done away with if you are given a small contribution. One point to think about is how or why micro tasks websites are so popular given that the pay rate is so low. Do people unknowingly become self-absorbed into the tasks that they spend too long doing it? Or do they like contributing to the myriad research studies on MTurk?


Experimental Study of Inequality and Unpredictability in an Artificial
Cultural Market

This was an interesting study, but with a few weaknessses. It is strange that their subjects are mainly teenagers and not adults, yet they argue in their paper that even if they tested it with adults, the overall amount of inequality and unpredictability would be similar. It is a bad assumption to make, because it is possible adults are not swayed by weak cues like amount of numbers. This is something to think about. Additionally, in experiment 2, they ordered the songs by popularity of downloads, with less popular songs very far down the page. Did this choice of presentation influence the unpredictability? I am of the opinion that those music that appeared above the fold will be seen more, and thus downloaded more, while those down the page would not be, thus exacerbating the Gini coefficient. I believe an additional control case ordering songs with least popular songs at the top of the single column page would allow the researchers to control for page layout bias.

Everyone’s an Influencer: Quantifying Influence on Twitter

This paper concludes that a cost effective marketing strategy could entail normal users sending out sponsored messages, rather than a very popular celebrity. Also, in addition to their current methodology, one could extend upon this project by considering the speed at which a URL is shared. A link shared 100 times over a few minutes could be considered to have greater influence than one shared over a few months. Finally, an interesting point to note is that a significant amount of content on Twitter is actually spam ( with spam accounts retweeting the same set of links over and over again.

Reading, Writing, Relationships:

This study was pretty in-depth and dispelled some of the common misconceptions about how Facebook is disrupting traditional communication methods in a bad way. On the contrary, such social networks can complementing existing ones. Also, having more weak ties do not crowd out strong ties, so it can be said that having more friends on Facebook does not mean that one knows less people more closely. The second research conclusion was that direct communication like messages help increase tie strength. This conforms to what people would hypothesize – pokes and likes are not meaningful forms of expression. The final conclusion was that losing a job increases weak tie interaction and decreased satisfaction arising from communicating with strong ties, especially if the latter is not in the same pickle as the former.


1) I asked my friend to make a Wikipedia edit. He protested that Wikipedia is already comprehensive as it is and any further edits by him would not stand to scrutiny. I told him to make an edit anyway and he told me edited the article which expressed an amount in 1920 dollars in more recent amounts. I believe a bot came by and wrote on his talk page that his edit has been reverted, telling him to use the sandbox while he worked on citing his source.

2) I asked my friend to hwork on Galaxy Zoo and she felt frustrated. Some of the photos that were given were photos which had too many tiny galaxies, making it hard to classify them properly. I believe this was a problem faced by many of people in class as well.

3) I got one of my friends in Singapore to set up an OKCupid profile since he is a bit lacking in the love department. He had reservations – “What if my classmates see me there? They will think I am desperate”. He put a profile photo of a plate of noodles instead of his real self.

4) I got my friend to read the iTunes legalese. He said no thanks. I showed him this link ( that presented the same in the form of a comic. He enjoyed it.

5) I did a twist on the crowd-working one by asking someone to work on 2Catpcha which is a middleware API for defeating captchas by getting lowly paid crowd workers to solve them manually. I got dissed at – imagine earning 75 cents for solving 1000 captchas and a chance to get carpal tunnel syndrome.


Dr. Google will see you now

Like Ryan, I believe that search trends on Google can be useful but we would do well to be careful to curve fit our hypotheses and bias to the data we get. One of the more notable projects at Google Research is the Flu Trends project ( which was recently closed. Because flu is such a common symptom, the results consisntely overpredict the severity of flu in regions – too much can lead to paranoia. That being said, ‘nowcasting’, as this strategy is being called, does have a place in economics and predictions.

The Wisdom of Crowd

I trade for a living, and I see this phenomenon of crowd wisdom every first Friday of the month, when the Non-Farm Payrolls (NFP) comes around. NFP is the biggest news event which shows the employment health of the US economy. Every month, people will post their predictions on NFP on Twitter. One of my favourite things to do is to compile a sample of predictions of Twitter users and get the average of their predictions. They are actually surprisingly quite accurate. I guess this prediction model can be improved if I had chosen only US based Twitter users, who are more likely to be in tune with the ebbs and flows of the US economy. Another interesting example I can think off hand is the use of satellite images to capture the parking lots of various (strip) malls. The more cars parked in front of a business compared to the previous year, the more likely it is doing well. Hence, investment firms can have better conviction in buying their stock, even before earnings results are released. This shows the power of crowd wisdom.

Mistaken Analysis

Ahhh, I didn’t read this before I made my first article post, but yes – analysis needs to adapt to changing technological trends. For example, big data analysis is used in Singapore to predict traffic patterns and determine the best traffic light timing for each junction, depending on the road and time of day. If we were to wholesale use data from 10 years ago today, it will definitely not be as useful.

Should Reddit be Blamed for the Spreading of a Smear?

I believe that as a platform, Reddit isn’t too blamed. The discussion that arose from that subreddit is a form of crowdsourced, citizen journalism, a field of which merits a lot of controversy in itself, if you look at how people takes things in their own hands in China. The thing is, I believe education is the key to making such platforms work. Just like trolling is commonplace online, supposed facts should be taken with a pinch of salt online. People who are influencers need to be more discerning about the things they disseminate on their online social media accounts lest they spread half-truths and sully their own reputation.


I perused topics from this list – and settled on the Historicity of Jesus which is a hotbed for religious controversy as seen on the talk page. A quick glance of the talk page tells me that in a way, meaningful contribution to Wikipedia entails the jargon that comes with editing the page – things like WP:SPS, WP:SELFPUB and WP:DUE are thrown about as though it was the norm. Those, by the way, are acronyms used to refer to making a neutral point of view in writing a Wikipedia page. It seems like the main source of contention on the page is the veracity of certain sources, as well as a certain Christian slant which frankly I do not really detect. Also, there has been huge debate on what is history and what is not factual (and hence not considered history).


Mining Our Reality

This article is pretty outdated in the sense that data sharing between entities have become more secure and standardized. For example, there is a service (I forgot the name, but it is pretty big and used by many businesses) that allows you to get the person’s physical address from his/her email address, without both parties disclosing their email addresses. Both parties run their email addresses through a standard one-way hashing algorithm known to both parties and see if they match up. If there is a hit, they retrieve the other details. In the same vein, much of the internet’s data brokers work the same way. The more egregious ones attempt to link up your real life activities with your online activities. They do this by offering you membership cards and loyalty coupons so they can track you both in-store and online. Data is being sure that’s for sure, but there needs to be a line drawn. The service I mentioned at the start of my post is very creepy.

Inferring friendship network structure by using mobile phone data

This study shows the advatanges of using handphone behavioral data over self-reported data. Increasingly, such features are used to cut down on cognitive biases inherent in self-reports. For example, Dark Sky the weather app uses the pressure sensors in iPhones to crowdsource data on whether a particular place is going to rain. Also, Waze the map app uses your location data to automatically tell other users if there is a traffic jam on a particular stretch of road, based on how fast you are going. In this respect, data insights from smartphones are able to serve as a proxy for other types of data such as relationships.

Reality Mining: Sensing Complex Social Systems

I’d like to point out that Bluetooth technology these days have been revised in view of the privacy concerns of yesteryear. Last time, people could send other people contacts via Bluetooth without the recipient knowing. I created a MIDP app to do such pranks on the trains. Lol. But seriously though, Most bluetooth devices these days are not broadcast publicly anymore – the user has to explicitly say they want to pair their device with other users. Nonetheless, I like this study alot as it took an innocuous piece of technology to identify social patterns. I can see how this is useful in situations like urban planning and crowd control during events. The privacy implications are even more real with the 3m accuracy provided by GPS and GLONASS technology. Moreover, it is easy for apps to get location data on Android phones since the permissions are baked in upon installation.

The Livehoods Project

I found it hard to appreciate the implications of the paper. The clustering algorithm, could cluster checkins of people to distinct groups, but that’s about it. Without further research like interviews with people in those clusters, not much information can be gathered. Moreover, in any city there are bound to be clusters of folks which makes up the identity of an area. I expected that their algorithmic clustering would have highlighted certain distinct features of an area – for example, the flow of people from other regions to Market District to buy groceries, for example. This would be more useful in the sense that we now have the ability to see which neighborhoods are underserved in terms of these amenities.


I used to use 4square before it became Swarm, primarily because I used it to document places that I have been to in a particular region. So when I was cafe hopping around NYC last winter, I checked in at each place I went. When my friend asked me what to eat at NYC, I just gave her the list of check-ins. Pretty nifty. Over the past few days, I checked in to mostly lecture halls and classrooms. Pretty boring to be used on campus actually. But I guess that’s CMU for ya. Anyway, Swarm is so much more gamified than last time. And there are more obnoxious ads forced in your face right now. I really want to like it, so I will give it a spin during winter break. A few of my friends use it, including my dad, so it is fun to see what they are up to (and where they work and live :O ) Again, if you are checking in to place X every 9am and place Y every 6pm, people are going to know where you work and live. Use judiciously.


Cooks or Cobblers? Crowd Creativity through Combination

Cooks or cobblers? Cooks take raw materials and conjure a scrumptious meal out of it. But cobblers take something slightly imperfect and mend it to be somewhat functional. Hence, I would have to say this paper strongly proves the point that crowd creativity is like cobblers. The chair, whilst creative (which I think is already pushing the limits of its definition), might not be ergonomic, functionable or useable. Also, many users have their interpretation of what should go into a product, leading to feature bloat. Say you are designing a stool for a person to sit on. Someone might add seat contours so that it fits your butt cheeks snugly, another might want to add an AC below it to cool your bottom during summer, or a deodorizer underneath so your farts won’t stink up the room while seated. Point is, this study centers on the fact that all coordination can be mediated through designs, which I think is a blasé perspective. If you want to see a good way on how such crowdsourcing can work well, take a look at Quirky. I am a member there for quite some time, and the processes there for such work impresses me. (

Redistributing Leadership in Online Creative Collaboration

The author is right in pointing out that Pipeline bears much similarity to GitHub; he argues that Pipeline aims for breadth rather than depth, but GitHub also has very strong administrative tools like Pipeline made, in addition to its world-class version control system. That said, there has been much movement in software development towards depth aka “niche” areas. You have Foursqare and Swarm as one example; and Facebook’s main app, Messenger, Camera, Groups, Pages, Notify, among many other apps that serve only one purpose. There is an article online on why they decided not to integrate all functionality into one single app – those interested can do a quick search to read it. I would say in terms on task delegation, Pipeline’s tools reduced the friction of admins – the main problem they were trying to solve. Yet, I feel that just like GitHub is known for code collaboration and Wikipedia for knowledge editing, perhaps Pipeline should focus on one niche area – image collaboration perhaps? Then they can develop tools such as realtime online editing of images (plenty of plugins online to do this) as well as displaying comments on the image canvas.

Science by the Masses

This was a very interesting article even though it was dated 7 years ago. When I started reading it, I doubted that Innocentive would be successful, because I thought a technical field like science and technology would be difficult to vet and verify, especially over the web. However, it seems like I was wrong. A key factor on why they were successful is that solutions to science often require a different perspective from a person in a field other than the one being researched on. This is shown by the conclusion that “the further a challenger was from a person’s field of interest, the more likely they were able to solve it”. Also, such specialized knowledge may be possessed by researchers not employed by the company, and such crowdsourcing allows companies to reach out to a wider audience. Moreover, it is cheaper than traditional research methods – why not harness the collective brainpower of the world’s scientists? Perhaps one day InnoCentive can expand into other domains too, not just science and tech.


DMCA: A Balanced Public Policy Brief

11 years ago, the authors of the paper recognized severe shortcomings arising from the DMCA

15 Years After Napster: How the Music Service Changed the Industry

This article talks about how Napster forced industry’s hand in trying to change how music distribution works. Industry was slow in predicting people’s music needs, and thus Napster and later, iTunes then Spotify filled that gap. While I concede that it is copyright infringement, the fact remains that there will stil be a big group that will not pay for music no matter how easy any service makes it to be. This group is content with searching YouTube or Chinese websites to satisfy their music craving, even when popular services are down. We can see this in the recent Popcorn Time takedown – not only did other services take its place, more popped up like a two-headed hydra. These group of users was never a lost revenue stream for companies. In fact, the freemium models of Pandora and Spotify allows artistes and companies to make some money out of this group of people, when previously they could not.

The 2 Teenagers Who Run the Wildly Popular Twitter Feed @HistoryInPics

This is an eye-opening case into the issues of copyright infringement in social media. Even Facebook (the company, not its users) have been accused of piggybacking on the original creators of viral videos, which Facebook users have reuploaded on Facebook. Previously, Facebook embed Youtube videos in news feeds; while they still do now, its algorithm shows videos uploaded on Facebook’s platform much more than on ther platforms – to keep eyeballs to their site. Facebook relies on safe harbor protections – it is not responsible or infringements by user-generated content, unless it blatently ignores DMCA notices from content owners. To this end, I can see that the 2 teenagers running @HistoryInPics doing the same to avoid copyright problems – they can simply ask for submissions by its wide swathes of followers. I am sure they would have no lack of submissions. If they receive a violation notice, they simply take it down promptly.

Just an ordinary guy, living in an extraordinary world