memamsa

Bitcoin Mining Economics

Sun, 25 Jan 2015 00:00:00 -0800

Two weeks ago, Bitcoin mining service CEX.io suspended its cloud mining service. The continuing drop in bitcoin price makes mining unprofitable against the maintenance costs accrued in USD.

Mining is the process by which a constant amount of Bitcoins are added into circulation, in return for verifying and recording payments in the public ledger. All transactions are publicly announced, timestamped and hashed together in a chain, with each additional timestamp reinforcing the ones before it. This blockchain is essential to prevent double spending when not relying on a trusted central authority such as a mint or a bank.

Mining Difficulty

The blockchain is distributed and anyone can access or add to it. However, it is necessary to maintain a single authoritative blockchain to ensure that a coin has not been spent earlier. Even if we decide to accept just the longest blockchain, there are several issues of accuracy and fairness. Deciding the end of the chain can be subject to race conditions. Nodes with higher transaction volumes would also get an undue advantage.

For these reasons, the Bitcoin protocol requires a proof-of-work scheme to make the chaining more difficult and the blockchain growth rate predictable. The hash must begin with a certain number of zero bits. In computing the hash, the block is combined with a nonce and the nonce is incremented until the hashed value meets the required number of zero bits. This work is exponential in the number of zero bits and requires extensive CPU usage to compute.

Mining is thus an incentive to supply CPU power to the Bitcoin network. Proof-of-work is essentially one-CPU-one-vote. The more CPU power you contribute, the higher your probability of mining Bitcoins.

The Bitcoin protocol increases the hashing difficulty every 2016 blocks, to ensure a rate of one block every ten minutes. The reward for finding the hash is halved every 210,000 blocks.

Computational Arms Race

In the beginning, miners were just enthusiasts playing with the Bitcoin client on their desktops and laptops. Very soon, the evangelists jumped in and sprouted exchanges. Once Bitcoin was pegged to a USD value, there was a gold rush to acquire Bitcoin before all 21 million bitcoins are mined. This in turn triggered a computational arms race.

From using CPUs to GPUs to FPGAs to custom designed ASIC chips, to purpose-built mining rigs with arrays of such ASICs, miners have deployed compute nodes with hash rates of upto trillions of hashes per second. Increased computation automatically increases the difficulty. Like the Red Queen’s Race, it takes all the running you can do, to keep in the same place. At current difficulty, the average time to find a block with 1 TH/s is six years.

That’s a long time to get a few dozen bitcoins. The next step then is to join a mining pool, where a network of miners combine their compute power and share the rewards. GHash, Slush and BTCGuild are a few such pools.

Mining now requires ASIC hardware with an upfront investment anywhere from $2,000 to $20,000. These rigs consume from 300W to 2000W or more. That kind of power consumption promises serious ongoing electricity and cooling expenses. Continual desire for more powerful hardware and an expanding Bitcoin network conspire to make mining rigs unprofitable and obsolete within six months.

Mining in the Cloud

CEX.io provided cloud mining capability to start mining Bitcoins without having to invest in hardware. Mining contracts allowed anyone to rent capacity with an upfront Bitcoin payment and ongoing BTC or USD payments for operating expenses. Mining contracts themselves could be traded. CEX.io thus functioned as an exchange to hedge on mining capacity. They even started futures trading with a small discount to buy mining capacity which would be installed at a future date.

Six months ago, a 1TH/s contract would cost 5 BTC while expecting to yield 2.05 BTC over six months. After that, the contract would earn 0.01 BTC month while costing 0.04 BTC in maintenance fees per month. The contracts could only be profitable by a combination of mining rewards while holding them and increased contract prices while selling them in the short term. Leading Bitcoin proponents had begun suspecting that the contracts might turn out to be Ponzi schemes.

Electricity Economics

Naturally the economic model faced increasing pressure as the BTC weakened against the dollar. According to this mining calculator, with BTC at 250 USD, a no-cost, zero-pool-fee 1TH/s 1200W rig is unprofitable on electricity costs alone, that too at a low price of 0.10 USD/kWh. Even if you use BTC to buy the rig and pay pool fees, you would still be paying for electricity and data center facilities in USD or equivalent currency. That, in a nutshell, is the squeeze.

Right now, many of those who got into Bitcoin early have ploughed their profits into faster mining rigs. Others, lacking resources, can no longer profitably mine Bitcoins. However, the economics now favors those with access to low cost energy. Oregon, for example, has relatively cheap hydro-electricity. In other places, electricity is subsidized, costs much less and can even be “free” for those with the right connections.

Miners colocated in such geos can buy older generation hardware dirt cheap and run more of it to yield profits. They can then afford to invest in the next generation of ASIC hardware utilizing 28nm or 14nm fabrication processes. Having 51% of the mining capacity located in China or Russia might trigger concerns within the US bitcoin community.

True Costs

Bitcoin mining is estimated to consume 5,000 MW of electricity, comparable to the energy demand of Ireland. That level of power can serve 4.5 million air-conditioned residences in Texas for a year.

Paradoxically, the Bitcoin network would work just as well even with a small fraction of the computational power. Overall mining rate would be unchanged even at a few GH/s. This is yet another example of how supposedly rational economic behavior of individuals can, in the aggregate, create systemic inefficiencies.

Regardless of the CPU power allocated to the Bitcoin network, the payment processing rate is artifically limited to seven transactions per second. For comparison, last year, the VisaNet retail payments network stress tested its systems for 47,000 transactions per second.

Add in the environmental impact of manufacturing hardware that can only do SHA-256 hashing and gets thrown away within six months, and you get a sense for the true costs of the Bitcoin network.

Mining Seigniorage

The right of the lord to mint money, now refers to the difference between the value of money and the cost to produce and distribute it. For example, it costs the U.S. Mint five cents for each 25-cent quarter it produces, thus earning more than $6.3 billion in seignorage from the “50 State” series of quarters.

The cost of creating a new Bitcoin is the capital depreciation, electricity, network bandwidth and data center leases. However, as we have seen, this cost is not fixed. Rather, it tracks the market exchange value of Bitcoin.

The rise in BTC price attracted more miners and more computation, making it cost more to mine bitcoin. As BTC falls, miners may continue to operate at a loss in the hope of future BTC appreciation. The CEX.io shutdown illustrates that miners will likely shut down their hashing systems at just below break-even. There will be no ongoing seignorage, only arbitrage.

Lock-In Effect

In the near-term, mining profits will accrue to those willing to amortize computational capacity and resource use over long time periods. As the mining reward halves, those with access to cheaper energy will increase their share of the hashing capacity. This will lead to consolidation and centralization.

When rewards asymptote to zero, transaction costs will have to be recouped from the parties spending and receiving bitcoins (if any). One motivation for staying in the game is that the few remaining miners can control the market and profit as intermediaries.

Those with significant accumulated bitcoins have no choice but to stay in the game. Centralization can lead to a less secure network, and a 51% attack can result in bitcoin theft and fraud through double-spending. They are thus locked into mining as a form of demurrage.

Evolution

In the face of price volatility, proponents have championed the blockchain as the real value of Bitcoin network. However, these economic factors will generally affect any use case that attempts to build on blockchains.

Other crypto-currencies have been designed to mitigate these economic effects.

Litecoin was intended to avoid giving an advantage to GPU, FPGA or ASIC miners.
Peercoin uses a proof-of-stake hybrid to address the high energy consumption and associated centralization.
Stellar proposes trusted gateways and a simpler consensus mechanism to maintain a distributed ledger.

Bitcoin pioneered the peer-to-peer network for transactions without relying on trust. The community self-organized and innovated in operating the network, even as speculators have raised the stakes. Bitcoins and blockchain show signs of increased adoption.

Understanding the underlying economics is critical to ensure we do not tie the financial system to an energy hogging monster.

So What If It's Work?

Sat, 17 Jan 2015 00:00:00 -0800

While it didn’t make the VCTalk links for this week, Paul Graham’s essay on work generated the expected discussion on Hacker News. To distill from his short parable:

If something that seems like work to other people doesn’t seem like work to you, that’s something you are well suited for.

By objective definition, work is exertion or effort directed to produce or accomplish something. Subjectively, if given the choice, work implies something people would rather not do. The assumption is that they would instead favor some other activities, considered as either leisure or play. If what you do doesn’t seem like work, yet other people find it valuable, then you can regard “work as play”.

Naturally, this resonates with programmers, mathematicians and others who solve problems in and of mental abstraction or engage in artistic creativity.

When free thinkers, religious figures or proponents of self-improvement talk about work, they also imply toil and suffering. Work is imbued with deliverance from evils, development of character and discovery of happiness.

It seems worthwhile to work through the various perspectives on work. To not do so would be a willful and self-serving blindness on the part of programmers and investors, when our work, in the form of software, is impacting every industry and every person and their work.

Work as Play

It seems that Paul posits work-as-play as a definitive end-state. Not just one we can all reach, but can also foresee, and thus use to guide what we should do. Historically, this was seen more precisely as an ideal mind state to strive towards. To “blur the line” or “make no distinction” between work and play, was the supreme accomplishment, an attribute of a master.

Work-as-play can be a useful test for choosing certain professions over others, provided you do not forget that anything worthwhile requires schlep. Many activities, including those considered “play”, such as skiing and piano, become enjoyable only after tedious work.

Work as Servitude

In reality, many kinds of work require surrendering autonomy and control. Even outwardly fun work can become dreary due to external forces and imposed requirements. Playing video games might be entertaining, but not if that is your job.

The past several hundred years have seen the deskilling of labor along with increased automation. Compared to artisans they replaced, assembly line workers found their fragmented work to be less satisfying. Their industrial workplace was more demanding even as it turned them into easily replaceable commodities. Nor has this transformation been limited to physical labor. Whether in the warehouse or at the bank, systems circumscribe vast swaths of work.

The trend seems inexorable, with Uber and others making work even more fungible, where the even the chance to perform work is auctioned off second by second, task by task.

Work as Identity

For many people, work not only takes up the majority of their waking time, it also shapes their identity. Whether in occupations as doctors and journalists or through titles such as assistant manager and vice president, people adapt themselves to the norms of the work and identify themselves with their professional status.

In fact, aspiration towards an identity is itself quite powerful in how and how much people work. Being a CEO, holding an elected office or running a non-profit is seen as desirable by many who work towards it, despite the long hours, exhausting schedules and other sacrifices involved.

Work as Worship

Quite a few people have been motivated by sanctifying their work, ignoring titles and eschewing economic rewards. Driven by theological doctrines or other ideologies, some work on grand schemes to open new possibilities, while others find new meaning in everyday work.

Beyond working for a goal - however selfless - is to act without attachment to the work and without desire for a particular result. Not everyone may have such a calling, but those acting upon it may see it as a way to attain self-actualization.

It’s All Work

Not a moment goes by without engaging in what someone would call work. A vacation can be a lot of work before, during and after. A good meal may require cooking and cleaning at home, or driving to and from the restaurant.

It seems better then to consider all these aspects of work as simultaneous and even necessary. They may be multiplexed based on our state of mind and external concerns. Some may act to propel while others help steer the course of our work.

What begins as play, might become servitude even as it shapes identity and provides resources to work on larger goals.

But as we traverse the hierarchy of needs, whether over decades or on a daily basis, let’s hope that we find in ourselves the ability to work with detachment.

VCs Talking Bitcoin

Sat, 10 Jan 2015 00:00:00 -0800

Over the past few months, VCTalk has been gathering top links shared by venture capital insiders. As 2015 gets into gear, we see investors outlining trends and predictions. In addition to devices, wearables and cloud services, Bitcoin has gained currency, so to speak, as an investment thesis.

Last week, Marc Andreessen reiterated his bullish stance with a longer than usual tweetstorm in response to recent downbeat coverage regarding volatility and thefts on Bitcoin exchanges.

Not surprising, since Andressen Horowitz has invested close to $30 million in Bitcoin related startups. Partners Chris Dixon and Balaji Srinivasan are renowned for their well-reasoned arguments for Bitcoin. Chris has outlined some ideas for native bitcoin apps, while Balaji leads the Stanford Bitcoin Group researching bitcoin theory and applications. A16Z is unquestionably the thought leader and the prime backer for Bitcoin startups.

As a result or as a measure of their success, they are no longer alone. A simple search of the vctalk links archive for “bitcoin” brings up a large list of other VCs. These VCs are sharing links about exchanges, wallets, regulation, funding, startups (e.g. Coinbase), payments, adoption and acceptance (e.g. Microsoft). Blockchain related items show up a lot, reflecting VC emphasis on the underlying technology.

Here is a small list, in descending order by number of bitcoin links shared.

@jdrive - RRE Ventures
@BoostVC - Adam Draper’s Bitcoin accelerator.
@pointsnfigures - Jeffrey Carter, West Loop Ventures
@jeremysliew - Lightspeed Venture Partners
@adaugelli - True Ventures
@davemcclure - 500startups
@fdestin - Accel Partners
@vkhosla - Khosla Ventures
@ssubo - Emergence Capital
@bernardmoon - Spark Labs Global

Other organizations include @TransMediaCap, @benchmark, @NEAVC, @seedcamp, @FundersClub, @rothventures, @HighlandCapital, @BirchmereVC, @techstarts, @redswanventures, @tribecavp and @canaanpartners.

Keep in mind that we are simply tracking their interest in terms of news and events surrounding Bitcoin. They may not have any investments yet. Also note that @pmarca, @balajis, @cdixon, @withfries2 and @a16z have been excluded so as to free up the top slots and highlight the rest.

There are likely many notable promoters and investors, such as Joel Monegro and Keith Rabois, who don’t show up on this list, because they don’t share any or enough links. Our goal here is to simply highlight those who are active. Hopefully, it gives some insights into how others are positioning themselves for emerging opportunities.

What other kinds of counts and correlations would you find useful? Let us know.

Uber Underhanded

Mon, 22 Dec 2014 00:00:00 -0800

Last week, the City of Portland and Uber reached an unusual compromise. The notoriously controversial transportation service, normally disdainful of regulators, has agreed to suspend pick-ups in Portland city limits for three months. During this time, the Portland Bureau of Transportation will work on a temporary plan to suspend or streamline the existing taxi regulations to “see how the market operates without those rules”.

Uber’s accomodating attitude may be in response to weeks of bad press and negative coverage. Portland, in turn, acknowledges that the taxi regulations need overhaul while ensuring fairness to other operators. In the past few months, the city has shown itself to be quite flexible towards sharing services.

However, its worth thinking about the larger issues raised by Uber’s practices and expansion.

Uber, Lyft, AirBnB et al have promoted their offerings as leading to a new sharing economy. Many pundits and opinion makers have been cheerleading this as yet another disruptive paradigm that will create jobs or foster “micro-entrepreneurs” or both. But the phrase “sharing economy” is as deceptive as it is oxymoronic.

Lest anyone think otherwise, Uber runs a marketplace for riders looking to hire cabs. Uber owns zero cars and employs zero drivers. For buyers, Uber is a substitute for taxi cabs and each ride is extremely transactional. The drivers are not sharing their car out of goodness, but are instead selling cab-for-hire services within the constraints imposed by Uber.

Unethical Tactics

When Uber muscled their way into traditional taxi services with UberX they ignored almost all taxi regulations around permits, metering, fares and insurance. Uber’s stated position has been that these rules are antiquated and do not apply to their technology and business model.

For the most part, technologists and free market supporters eschew government limits and quotas, so there were few champions outside of the taxi industry for the oligopolistic medallion or licensing system.

Fare metering and anti-gouging regulations are another matter. Passengers used to the benefits of predictable pricing have not been too keen on “surge pricing”. The first backlash against Uber started when party goers got hit by unpredictable doubling, tripling or even septupling of normal fares.

Amongst the technorati, Uber’s tactics against competitor Lyft met with universal condemnation. In targeting Lyft drivers for recruitment, Uber agents used burner phones to book spurious rides, cancelling them when the driver showed no interest in signing up for Uber.

With their billion-dollar war-chest, rallied up riders and politically connected influencers, Uber has built out a nation-wide lobbying machine to bend and bankroll their way around laws and regulation.

Unfair Marketplace

Growth targets and investor expectations require Uber to aggressively expand service areas and rider volumes. It cannot wait for organic influx of providers (i.e. drivers) to sign up to its platform. Unlike Ebay or Amazon sellers, Uber drivers are not signing up to hedge fares or arbitrage capacity. For all practical purposes, they are wage-seeking contractors, even if technically not employees.

In a playbook typical of “sharing” companies, Uber offloaded all legal liability and financial risks to the drivers. UberX drivers are tacitly encouraged to use personal insurance inappropriate for their driving profiles, exposing them to personal losses and perjury.

Uber drivers have just 15 seconds to respond to service calls and accept fares. This is not enough time, especially while driving, to look at the pickup location and make a calculated decision. Drivers who decline or ignore calls while “on app” are temporarily suspended. A string of poor customer interactions can get a driver deactivated, cutting off their income stream.

With no choice in the dispatch, no leeway over payments, no access to the customer relationship and a rating system which treats them as a commodity, Uber drivers have no way to build a brand or control their business.

Umpires, not Empires

In the absence of guiding principles, network effects from choices of convenience quickly lead to lock-in and lack of choice. Privacy concerns around Google and Facebook should make us more cautious of enabling new data empires, especially ones such as Uber, known to fool around with God View.

Uber’s CEO can’t wait for a future where drivers will be replaced with a network of self-driving cars. Well before that happens, it is more likely that Uber’s very success will shrink all other taxi driving opportunities for drivers. Think about drivers building out their own shackles.

Uber is a microcosm of the new Silicon Valley, intent on algorithmically optimizing every human endeavor while inserting itself into every relationship. Software eating the world and egesting automatons at an ever accelerating pace. While this is portrayed as an inevitability, it does not have to be.

Software, more so than other artifacts, embodies the philosophy of its creators. Decades ago, GNU seeded a software commons to enhance choice and freedom for users. It is high time to extend those principles into the cloud. Uber’s underhanded tactics should be a warning to us that fair market algorithms and open reputation systems are absolutely essential.

Hour of Code First Hand

Mon, 15 Dec 2014 00:00:00 -0800

I recently saw this video from Hour of Code.

I have been aware of Year of Code, Code Academy and various other learn-to-code movements. But watching this video with a roomful of high-schoolers as part of a programming workshop was thought-provoking.

The Hour of Code campaign is designed to be inspirational, with appearances by actresses, models, NBA players, rap stars, pop singers, skateboarders, footballers and politicians along with the President of the USA. The message is that everyone should code, anyone can do it, and that it is easy to learn. All it takes is a few lines of code.

Code.org, the organization behind the video, includes a formidable list of backers including Bill Gates and Mark Zuckerberg. Their motives are certainly laudable: to expose as many students to Computer Science as possible.

Nevertheless, this video and the learn-to-code movement in general leaves an unsettling feeling shared by many programmers.

The disquiet has nothing to do with squelching future competition. Unlike medicine, law, plumbing or hair styling, programming is unrestricted and open for anyone to practice. And the majority of programmers prefer it that way. Many are willing to spend untold hours helping newbies and answering questions. The Internet as we know today got started to facilitate such discussions.

Ignoring the choice of role models within the video, the first troubling aspect is about pedagogy. Starting off claiming that a subject is easy sets up students to fail. It almost guarantees that many students are going to feel frustrated and blame themselves or their teachers when they hit the inevitable difficulties.

A related issue is about motivations. Don’t learn programming because the jobs pay well today. Do it if you find it interesting, and can have enough fun for you to put in your ten years.

Seymour Papert, a pioneer in introducing programming to children, had the explicit goal of using programming as a medium, better than pencil-and-paper, through which powerful ideas can be brought within reach. This led to the development of Logo, turtle graphics, Smalltalk and Scratch.

You could say that the tutorials at code.org offer simpler tools and solutions tailored for the browser or tablet. However, they come across as scattershot and unlikely to stick or help develop significant concepts. The feel-good vibe associated with broad and high profile advocacy can make for a shallow experience.

Ultimately, programming is less about coding, and more about conceptual thinking and imaginative problem solving. It would be better to cultivate the analytical mind and hone communication skills without getting bogged down in arbitrary syntax or proprietary frameworks.

This video from the dawn of personal computing should be a useful antidote.

Launch Festival Mentions

Fri, 28 Feb 2014 00:00:00 -0800

Launch Festival 2014 last week Feb 24-26, showcased several startups, in addition to featured keynotes and partner presentations. A streaming search for #launch captured around 7000+ tweets.

The agenda was quite bare-bones, focused on debuting and demoing some 30-odd startups. The audience was large, in the virtual and likely physical space. There were 3000+ unique users tweeting or mentioned with #launch.

The tweets are heavy with name-dropping with an average of 1.5 user mentions per tweet, and many with upto 9 mentions.

So proud of @launch team @LifeAfterCubes @gdelvac @jadetran @EmilyStu @_brandice @steepdecline @lukelightning Andrew Simon Brandon +++ !!!
— jason (@Jason) February 23, 2014

Unlike Strata, the schedule is one-track, and user mentions are a reliable indicator for session of interest. Like much of the social phenomenon, the user mentions have an uninteresting exponential drop-off in the aggregate. As we saw before, keynotes tend to garner the most tweet responses.

Finding the top 25 most tweeted users was trivial. When drilling down to find top 10 mentions on a daily basis, there was an interesting pattern. Some were mentioned almost every day, while many only showed up on a single day. Of course, we expect @Launch and @jason to be regularly talked to. The variance seems to be about anticipation and affirmation (e.g. “looking forward to @mcuban”, or “excited to hear @paulg”).

Curious about how early, and how much later people mention or reference sessions, we mapped tweets to participating sessions and counted frequency by time slots. Here is a levelplot-inspired visualization of the tweet activity during Launch.

The horizontal and the vertical axes are both time slots. Only actual presentation and panel slots are included on the vertical axis, with the left column listing the session participants. The horizontal axis also includes break periods, and the time period between 6pm to 9am next day is one slot. The slots are rendered as a heat-map, the redder the slot, the more the tweet activity.

Hovering over a cell shows the number of tweets during that slot. Slots underlined in blue correspond to the time for the session within the row. Finally, the larger bold font indicates the top 25 users and apps mentioned over the three days.

	24th 25th 26th
	9a 12p 6p 9a 12p 6p 9a 12p 5p
travisk
DavidBookspan bubbam jtriest
coin diamondkracer
OmSignal stephanemarceau
iKnightscope WSantanaLi
VirtuixOmni
robinhoodapp
mchrishawkins signnow barracuda
paulg ycombinator
edwardhieatt gopivotal pivotallabs
Bebo_Official CSTMglobal byjaunt UserbinHQ ramenapp TheTakeApp crediblelabs TinaSharkey hunterwalk jtriest paigecraig
RallyteamApp OHMConnect ImmersiveED1 ReparteeApp getflock MeetingPulse AdapticsHQ Amazi SEOEng mpesce cvwdyn karaswisher erikeccles
heiligerj ZelkovaVC jamesmaiocco joshy88 dfjjosh
jimpatterson Cotap
PerformanceCI tryripple hiloapp vue addmeinapp blushmessenger habarifit dsamuel cyantist msavino daneglasgow
yvesbehar
AtlasWearables iHaloBand ConnectSanFran crowdmics PeerSpaceApp dfjjosh sarahcuda rahulsood tsulliusa
MylesSutherland Esri
HandUp BeCandidCo jamoapp lastrevapp Portfolia1 miiglers Posterro ZelkovaVC adeoressi DonDodge vivekwadhwa
naval
robrhinehart soylent joshtetrick hcfoods
chr1sanderson 3DRobotics csanz skycatch
TimDraper Draper_U
adamse salesforce appexchange
dattchapp PaperfoldApp WehicleApp camiolog oeuvreio brianalvey famolari
thejo
mcuban
milesward
LeylaSeka Desk
useminimal getbettercompany ListenLoop KnoxPayments trendingfm Meshify dappershopping DavidSacks a5steve DonDodge
LudlowVentures

Now, if you are like me, go find out more about @yvesbehar, @Handup, @Portfolia1 and @iKnightScope. Enjoy.

StrataConf Insider Tweets

Sat, 15 Feb 2014 00:00:00 -0800

StrataConf 2014 was from Feb 11-13, with focus on tools, techniques and learnings for Big Data insights and visualizations. We thought it might be interesting (and oh so meta!) to gather the public activity during the conference and decide which sessions generated interest on twitter.

Collection

First off, we are not talking Big Data. Neither are we doing all-out surveillance. Our starting point was a list of 200+ speakers and panelists.

The idea and the plan came together at the last-minute. By the time twitter monitoring was in place, the tweets for the morning of Feb 12 were past the timeline limits. Over roughly 47 hours, we had collected around 2500 tweets from this group of Strata insiders.

The Strata schedule has around nine parallel tracks at most times. We grabbed the .ics file to gather start and end times, and used the full listings to get the ratings information.

Tweets By Time

Here is a histogram of tweets over a two-day period with bin-size of 15-minutes. Caveat: Data before noon is underrepresented.

Clearly, a lot of tweets happen during and just after keynotes. Tweets really pick up steam around 2p and 5p. Surprisingly, the insiders tweet less during breaks and very little during lunch. One can imagine that hunger, thirst, bodily needs and in-person social conversations trump tweeting.

Finding the Most Tweeted Sessions

Obviously, tweets cannot be related to the session based on time. In addition to the multiplicity of simultaneous tracks, a lot of tweets are not about a particular session. Some thoughtful people tweet well after the session.

What we need is a way to classify a given tweet to match a session.

Features and Classification

Features for session include the title, the speaker and organization. The session description is weighted very low (being quite monotonous in terms such as big data, tools, real-time, cutting-edge, cloud, technologies and such mumbo). Tweets include user mentions and
hashtags in addition to text and are filtered for #strataconf. A modified vector-space tf-idf model was used for classification. Sessions within the time-slot corresponding to the tweet time are given a small boost during the matching process.

Methodology and Error

The classification code was hacked up within a day and half. A random 10% of the tweets were used in fine-tuning the model weights with statistical checks. The resulting classes were sorted by membership count and the top and bottom 10 ranked sessions were manually scanned for mismatched tweets and error counts.

Precision as measured in the top 10 sessions is over 95%, while recall is less at an estimated 85%. However, the distribution and the ranking of the sessions will hold to a high degree of confidence.

Classification Results

Of the 151 distinct sessions, at least 120 were associated with one or more tweets. Two sessions proved to be troubling catch-all for tweets (Club Strata and Great Debate) and have been removed from the list with due mention.

Sessions Most Tweeted By Strata Insiders

The list.

A Few Tweets

Embedded here to provide a flavor of the classification problem.

There goes the epistemological neighborhood. -- james burke #strataconf
— Joe Hellerstein (@joe_hellerstein) February 13, 2014

Lispers at #datascience #strataconf : could I do the @beaucronin Church lisp example in #clojure instead? Libs ported there? If not, I might
— Russell Whitaker (@OrthoNormalRuss) February 13, 2014

Lucky people are open to new experiences, easily abandon routines, fail often, shrug off failure, try again. @davidmcraney #strataconf
— Rachel Kalmar (@grapealope) February 14, 2014

Excited to learn about how my data (and lack of sleep) is being used in Data Science. #strataconf @mrogati pic.twitter.com/pTEiEtdBXu
— Noelle Sio (@noellesio) February 13, 2014

And the winner for #strataconf best talk title is... Chicago Bars, Prisoner’s Dilemma, and Practical Models in Search http://t.co/XULQx96Kt7
— Daniel Tunkelang (@dtunkelang) February 13, 2014

More can be done better, automatically. Let us know if you find this useful, or would like to see this kind of analysis in some other context.

Bitcoin Today - 11 Feb 2014

Tue, 11 Feb 2014 00:00:00 -0800

A roundup of links and tweets chronicling today’s gyrations.

memamsa

Bitcoin Mining Economics

Mining Difficulty

Computational Arms Race

Mining in the Cloud

Electricity Economics

True Costs

Mining Seigniorage

Lock-In Effect

Evolution

So What If It's Work?

Work as Play

Work as Servitude

Work as Identity

Work as Worship

It’s All Work

VCs Talking Bitcoin

Uber Underhanded

Unethical Tactics

Unfair Marketplace

Umpires, not Empires

Hour of Code First Hand

More Top Links for More Topics

Top Links in Venture Capital

Launch Festival Mentions

StrataConf Insider Tweets

Collection

Tweets By Time

Finding the Most Tweeted Sessions

Features and Classification

Methodology and Error

Classification Results

Sessions Most Tweeted By Strata Insiders

A Few Tweets

More

Bitcoin Today - 11 Feb 2014

memamsa

Bitcoin Mining Economics

Mining Difficulty

Computational Arms Race

Mining in the Cloud

Electricity Economics

True Costs

Mining Seigniorage

Lock-In Effect

Evolution

So What If It's Work?

Work as Play

Work as Servitude

Work as Identity

Work as Worship

It’s All Work

VCs Talking Bitcoin

Uber Underhanded

The Mythical Sharing Economy

Unethical Tactics

Unfair Marketplace

Umpires, not Empires

Hour of Code First Hand

More Top Links for More Topics

Top Links in Venture Capital

Launch Festival Mentions

StrataConf Insider Tweets

Collection

Tweets By Time

Finding the Most Tweeted Sessions

Features and Classification

Methodology and Error

Classification Results

Sessions Most Tweeted By Strata Insiders

A Few Tweets

More

Bitcoin Today - 11 Feb 2014