It’s one thing to create a system that does what you designed it to do. But the most interesting design problems center around unexpected circumstances. How does your system respond to environmental stress (heat, vibration, radiation, resource depletion, shock forces, high pressure, low pressure, corrosive substances, etc.) or deliberate attacks? How does the system as a whole behave if subsystems fail? How will your system be restored to optimal condition if it takes damage of any kind for any reason? Can your system be reconfigured and used in ways you didn’t expect yet still behave in the way others will hope for? These are the questions that separate good engineering from great engineering.

Sometime in 1959, Harriet Pierson was running JOHNIAC at night and would have great difficulties with the punch—an unreasonably high rate of echo-check failures. When the job was rerun during the following day, no such trouble was encountered. After a bit of digging it was determined that the only difference in the operation was that the JOHNIAC machine-room lights were turned out at night when the run was made and no one was in the room. The only conclusion that could be drawn was that JOHNIAC was afraid of the dark, and indeed it was true. Neons had been used as active elements in the I/O lines and some had aged and become de-ionized so that they would not conduct without the aid of some external radiation—namely, light. So it came to pass that JOHNIAC was provided with its own lights independent of the room lights and never had to be in the dark again.
The History of the JOHNIAC, F. J. Gruenberger
RAND Corporation, Santa Monica, California

We don’t need deep theories to tell us that parallel universes exist — single-particle interference phenomena tell us that. We need theories to explain and predict such phenomena: to tell us what the other universes are like, what laws they obey, how they affect one another, and how all this fits in with the theoretical foundations of other subjects. This is what quantum theory does. The quantum theory of parallel universes is not the problem, it is the solution. It is not some troublesome, optional interpretation emerging from arcane theoretical considerations. It is the explanation — the only one that is tenable — of a remarkable and counter-intuitive reality.

— David Deutsch

This is always the case when you design cryptographic systems: the basic trust relationships you build upon are all based on contractual relationships.

— Cryptography Engineering (Ferguson, Schneier, Yoshi)

On brevity

Expressing concepts with nice sentences composed from orthogonal words is surely useful.

Yet it may often be of even greater use to employ more words than necessary, without much sacrificing density of meaning; to tell multiple stories with the same moral; to use the constraints imposed by different approaches to the same idea to increase the chances that the reader will converge to a breadth and depth and specificity of understanding that could not reliably be attained through an extremely succinct version, without sharing a large implicit context: a requirement which would limit the potential impact of writings (despite their potentially crystalline clarity when read back to oneself).

I think I need to practice the latter style more. (Maybe I should start by using more periods…)



Michael Nielsen helped me realize that the core innovation of Bitcoin is the consensus protocol for strict sequencing of globally replicated, cryptographically signed commitments (made by anyone at any time). The actual Bitcoin currency and the “mining” lottery are basically trivial given this broadly applicable core protocol.

I’ve been developing an abstraction of a “transaction” from the notion that under ideal assumptions, a “transaction” is just a reallocation of access to resources (like “rights" or "capabilities”) that leaves at least one party better off and no party worse off (a.k.a. a Pareto improvement). (“Capabilities" are like little programs that you can become authorized to execute opaquely and which moderate your access to a resource, a somewhat similar role to that of Bitcoin scripts.)

Such a flexible notion of access to resources also calls to mind the “complete market" of the Arrow-Debreu model: you can create cryptographic bearer instruments dependent on state, time, and location, e.g. the resource “8 coins if it rains in San Francisco tomorrow”. The transaction costs of these instruments would be dramatically lower without clearing and settlement requiring human intervention, thus bringing Arrow-Debreu much closer to reality (or, rather, reality closer to Arrow-Debreu).

"A ride from Embarcadero to Alamo Square at 1pm tomorrow" is also a resource, the kind which makes Uber’s business tick. "A place to stay in the Upper West Side from Feb 2-10" is also a resource, the kind that Airbnb brokers. The "sharing economy" is actually about completing markets, and these highly successful businesses are yet solving only special cases. What about resources like "$50,000 invested in Lou Hawthorne’s cancer research this year"? Even resources that can’t be committed to or traded, like "my cancer cured within the next five years", can be represented locally within user-agent software combining probability theory, logical planning, and interactive modeling with recursive model sharing (e.g. "I trust Marvin Minsky’s model of cognition between 40% and 90%"), thereby establishing a mathematically consistent reputation network. (Take that, star ratings and up-votes.) User preferences can be elicited, and software agents can engage in trade that - approximately, heuristically - optimizes for human utility, up to Pareto-optimality.

Consensus protocols of which Bitcoin is a pioneering proof of concept embody the potential to reform more than just currency, but the entire economy, from corporations and venture capital to co-housing and barter. This seems pretty important, so I’m starting to spend much of my time developing these ideas.


Speculations on the nature of the university

The timeless purpose of the university is to introduce intellectuals to each other so they can function more effectively as human beings and as contributors to society. The circa-2000 university serves this purpose through residential post-secondary education; expresses it through a tenured research faculty in the form of scholarly publications; and asserts it through the function of warranting (its graduates, generally to industry, and its faculty, generally to government). However, as communications technology continues to improve, we should expect the comparative advantage of residential education eroded to less than zero, the scholarly publication supplanted by lower-overhead, higher-value digital products, and warranting authority subverted by online reputation.

So, how could the university of the future serve, express, and assert its purpose? By investing in low-marginal-cost digital education products, pioneering decentralized models of intellectual and scientific discourse, and by delivering products and services previously thought impossible directly to market.

The Buxton Index of an entity, i.e. person or organization, is defined as the length of the period, measured in years, over which the entity makes its plans. For the little grocery shop around the corner it is about 1/2,for the true Christian it is infinity, and for most other entities it is in between: about 4 for the average politician who aims at his re-election, slightly more for most industries, but much less for the managers who have to write quarterly reports. The Buxton Index is an important concept because close co-operation between entities with very different Buxton Indices invariably fails and leads to moral complaints about the partner. The party with the smaller Buxton Index is accused of being superficial and short-sighted, while the party with the larger Buxton Index is accused of neglect of duty, of backing out of its responsibility, of freewheeling, etc.. In addition, each party accuses the other one of being stupid. The great advantage of the Buxton Index is that, as a simple numerical notion, it is morally neutral and lifts the difference above the plane of moral concerns. The Buxton Index is important to bear in mind when considering academic/industrial co-operation.

The Bitcoin Network is 0 FLOPS

FLOPS, FLoating-point OPerations per Second, is a measure of computational power specific to the scientific/supercomputing community. It measures the number of floating-point operations a computer system is theoretically capable of executing every second. Bitcoin mining does not require any floating-point operations (it’s primarily bitwise operations and lookups, with some integer operations and conditional branches). A related fact is that special-purpose Bitcoin mining machines cannot (even theoretically) be programmed to perform a single floating-point operation. Therefore, reports on the number of PetaFLOPS the Bitcoin mining network has achieved are misguided. They’re basically estimating how many FLOPS a supercomputer built for scientific problems would need to defeat the Bitcoin network at its own game. But even rough order-of-magnitude comparisons are misleading, since Bitcoin mining is a very different kind of computation from those FLOPS are useful for describing – apples and oranges.


A friend recently asked why cell phones are supposed to be off during a flight.

Oh, I know ALL ABOUT this. I have sources and everything. Prepare yourself for the low-down.

First of all, it’s worth noting that there are two federal agencies involved, the FAA (which regulates air travel) and the FCC (which regulates radio). That’s a big part of why this particular issue (personal use of radio while traveling by air) is so legally complex.

Let’s start with the history. The FAA ban on Personal Electronic Devices (PEDs) was instituted in May 1961 because of concerns regarding passenger use of FM radio [1]. At the time, one of the principal means of aircraft navigation was the use of ground-based beacons operating in the VHF Omnidirectional Range (VOR), 108-118 MHz [2]. If you’ve used an FM radio lately(?!), you know that the range of stations is 85-108 MHz. So, if your FM radio is tuned to 108 MHz, and your aircraft has a VOR receiver tuned to 108.05 MHz, it is entirely plausible that the FM radio will rebroadcast the FM music signal with a near-field broad spectrum that drowns out the 108.05 MHz VOR that the aircraft navigation system is trying to listen to. So the FAA said that if you’re flying an aircraft using VOR navigation, you better not allow anyone in the cabin to turn on an FM radio. As people began to carry around more Personal Electronic Devices - Walkmans, Gameboys, etc. - those became subsumed in the FAA’s ban as well, out of general caution [1], although the FAA allows aircraft operators to make their own exceptions to the ban as they see fit [4, paragraph (b)(5)]. Meanwhile, the FCC licenses air-to-ground radios separately from land radios, in large part because signals transmitted from the air have the potential to interfere with a much wider swath of land than would be possible from the ground, so in 1991, the FCC explicitly disallowed the use of (land-licensed) cell phones from onboard aircraft [3], partly in response to the FAA’s nagging and partly in response to land-radio licensees who were asking if they could operate from balloons and such.

Therefore, the current situation is that the operation of cellular phones during flight is forbidden separately by both FAA and FCC regulations. Industry standard rules about taxi, takeoff and landing and the magic 10,000-feet number seem to derive from the FAA’s non-binding advice to aircraft operators [1], which is justified on the basis of both the increased risk to aircraft safety if navigational equipment should fail while the plane is close to the ground, as well as the increased risk to personal safety in case of a botched landing if someone is listening to their Walkman and not paying attention when the crew is asking them to brace for impact.

It is worth noting that there is strong evidence that one particular phone, the Samsung SPH-N300, is capable of jamming an aircraft GPS receiver [7], thanks to sideband interference. The SPH-N300 is in compliance with all relevant FCC regulations limiting harmful interference from land-based radio devices, but severely exceeds the FAA’s safety margins for aviation electronics; and it has been demonstrated that switching on the phone in the cockpit of a specific aircraft effectively disables that aircraft’s GPS. Moving the phone behind a solid metal object (such as, say, the typical cockpit door) removes the interference and restores the GPS to its normal function; but this is still worthy cause for concern. There has been circumstantial evidence of other interference [12].

The FAA commissioned a rule-making committee last year to revisit their regulations on Personal Electronic Devices [8], due to widespread consumer complaints, including a perceived double standard since pilots are now allowed to use iPads even in the cockpit (though all wireless functions must be turned off at all times) [9]. They will not consider the possibility of actually placing a phone call from a commercial flight [8]; cell networks would fight such a change [10], probably for the same sort of reason they routinely produce short films asking moviegoers to switch their phones off. However, the Chairman of the FCC has written a letter to the FAA [11] supporting reconsideration of the ban on electronic devices e.g. during takeoff and landing. (That said, the FCC has shown no interest in allowing actual cellular communications during flight; they simply see no reason not to allow devices in “airplane mode” during takeoff and landing, as currently allowed above 10,000 feet.) This rule-making activity is still underway as of this week [13].

Amusingly, the FAA ban on personal electronic devices specifically exempts portable voice recorders and electric shavers, alongside hearing aids and heart pacemakers [4]. However, noncompliance with crewmember instructions is considered a federal offense in its own right [5] under 14 CFR 91.11 [6], so an airline can essentially ban whatever they want whenever they want and you are legally obligated to comply.

[1] http://ashsd.afacwa.org/docs/AC91-21-1A.pdf
[2] http://en.wikipedia.org/wiki/VHF_omnidirectional_range
[3] http://digital.library.unt.edu/ark:/67531/metadc1883/m1/40/
[4] http://www.ecfr.gov/cgi-bin/text-idx?node=14:
[5] http://www.criminaldefenselawyer.com/resources/interfering-with-a-flight-attendant-or-crewmember.htm
[6] http://www.ecfr.gov/cgi-bin/text-idx?node=14:
[7] http://www.cs.odu.edu/~mln/ltrs-pdfs/NASA-2004-tm213001.pdf
[8] http://thehill.com/blogs/transportation-report/aviation/245703-faa-to-study-allowing-electronic-devices-use-in-flight
[9] http://bits.blogs.nytimes.com/2011/12/14/f-a-a-approves-ipads-in-cockpits-but-not-for-passengers/
[10] http://money.cnn.com/2005/06/16/technology/personaltech/cingular_inflight_calls/
[11] http://bits.blogs.nytimes.com/2012/12/06/fcc-calls-on-faa-to-allow-electronics-on-planes/
[12] http://asrs.arc.nasa.gov/docs/rpsts/ped.pdf

San Francisco changes because the world changes. It was formed in a gold rush and reshaped by every one that followed.