Introduction

Austin, just like several other southwestern American cities, is not built for walking. However, after several years of becoming familiar with its terrain (primarily through driving), I discovered discrete pockets up and down the superfluous city grid that run against this popular notion. I was fortunate enough to find a home in one of these blocks, where I can walk to a grocery store, a coffee shop and a bar within 15 minutes, but my luck has run out. The little co-op grocery store I walk to is closing its doors at the end of 2026. One of the several reasons for the closure given by the co-op board stood out to me because of the deadpan irony — construction for Austin's federally funded 9.8 mile light rail extension will block easy access for pedestrians, cyclists and most importantly trucks making restocks at the grocery store.(1)

Austin Transit Partnership (ATP), the local government corporation created to build the urban rail extension, estimates the cost to be $7 billion and the first ride to happen by 2033. After you factor in the long delays faced by such large infrastructure projects, my chances of riding one of these trains is quite slim. Of course, I am neither alone, nor the first person to be in this predicament.(2)

The 100-mile Washington DC metro which was signed into law in 1965, broke ground in 1969, and was estimated to cost $1.8 billion. After several delays, the metro was declared “complete” in 1985 — even though construction for the green line (part of the original plan) would begin only in 1991. When the green line opened in 2001, the project was estimated to have cost a total of $10 billion.

In the 20 years from 1965–1985, one generation of DC residents lived through the construction of the metro without seeing its benefits, while bearing the moderate inconvenience of persistent construction activity and traffic snarls. During these years, the bureaucratic machinery, even though backed by the supposed strong arm of the federal government in DC, ran into several delays and demands by interest groups. These are documented in The Great Society Subway: A History of the Washington Metro:

The peak years of Metro construction — 1973 to 1976 — highlighted the difficulties, and expense, of macro engineering in a crowded city. Every park, every historic structure, every sidewalk could become an obstacle. In one case, engineers had to shift the route to spare a single, historic oak tree. Amid fights over individual segments, broader debates highlighted the tensions inherent in any public transit authority.(3)

In addition to macro engineering challenges, there are social engineering obstacles to big infrastructure projects. In Florida for example, Brightline, the privately run high speed rail that connects Miami to Orlando has been plagued by pedestrian deaths and accidents.(4) A high proportion of these are caused by drivers or pedestrians underestimating the speed of the train while negligently cutting across railway crossings (freight trains that the Florida man is accustomed to runs at 40mph compared to 100 mph average speed of the high speed rail run by Brightline).

While the pain of large infrastructure projects can be put into statistics — number of years delayed and billions spent — the cost of such projects is ultimately personal to the people working on them and the citizens frustrated by the lack of payoff to their tax dollars and mundane suffering. I want to make sense of the lifecycle of such large infrastructure projects — how are they planned? What factors determine if they will be successful or not?

Infrastructure challenges are not equally distributed. Since 2008 China has put into operation over 25,000 kilometres of dedicated high-speed railway (HSR) lines, far more than the total high-speed lines operating in the rest of the world.

The Chinese HSR network has been built at an average cost of $17 million to $21 million per km — about two thirds of the cost in other countries — even though many Chinese lines have a high proportion of their route on viaducts or in tunnels. The Medium and Long Term Tail Plan (MLTRP), developed in 2004 by several state owned stakeholders and modified in 2008 and 2016, along with complementary five year plans, played a key role in the success of the Chinese HSR. The plans, once approved, were never changed, thus providing economic and physical certainty about the future state of the HSR construction. A World Bank paper claims that setting a plan in stone for five years is an idiosyncrasy that is possible only in China.(5) They don't expound on this important claim but I will try to answer the question — why do some countries build rail infrastructure much faster than others?

Increasing complexity through the ages

Modern paved roads in Britain have their origins in the alienating, desolate landscape of the Scottish Highlands. In the 1720s, England was policing the Scottish clans which had threatened to install Charles Edward Stuart as their sovereign. The Scottish mountains were inhabited by free people who resisted English governance. Mountains were traditionally hard to govern and this is why people who did not want to be governed chose to live there.(6) The only way for the English to govern the highlands was to change the landscape in their favour. In his report, commanding office General George Wade wrote: “Roads and bridges alone could render Scotland subject to English power.” In Roads to Power: Britain invents the Infrastructure State, Jo Guildi writes about the first centrally coordinated large scale road building project, designed to rein in the Scots:

They totalled about 250 miles of road, averaging sixteen feet wide. After the second rebellion in 1745, another round of roads connected Dumbarton Castle, Stirling Castle, and Loch Lomond with the Bridge of Fruin and Duchlage. All the construction work was carried out by military labor, and soldiers continued to maintain the roads until 1815, when maintenance was transferred to a parliamentary committee. By this point almost 900 miles of roads had been built, created, and maintained by parliamentary grants of £4,000 to £7,000 a year.(7)

In the process of constructing these roads, the military synthesized the best methods and protocols for road construction. For example, the staggering mountain landscapes of the highlands made sighting and route-planning hard. One of the reports describes the experience of looking out and seeing mangled roads disappear behind hills as “... a kind of whimsical disorder.” This led to many advances in the fields of routing and sight planning, introducing concepts such as vanishing point and educating surveyors in contemporary theories around perspective and vision. The practice of using large boulders as landmarks along roads has its origins here. It seems that this practice is still used in British road construction — I remember coming across such landmarks in the British-constructed roads in Kerala, India, where I grew up.

By 1830, Parliamentary road construction dominated over military road construction in Britain. The Parliamentary roads were funded by the state and led to the formation of the bureaucratic state that was responsible for planning and maintaining the infrastructure. Prior to Parliamentary roads, the advances made by the military in road planning had led to construction of parish roads and turnpikes. These were roads that were maintained by lords in smaller towns, often collected tolls, and were plagued by corruption. They were often disconnected from each other, used different construction methods and varied in width. The Parliamentary roads, on the other hand, were centrally planned.

A critical role here was played by individuals who planned roads and presented them to Parliament for funding approval. This involved advances in data visualization, and standardization of processes and materials used. The synthesis of these skills eventually became the field known as civil engineering today. One of the most important of these standardizations was called Macadamization, after John Loudon Macadam, a proponent of centralized planning of roads.(8) Macadam proposed that all parliamentary roads should be built using the same materials and following the same specifications . This proved to be much more expensive than the prevailing method which emphasized using materials that were local to the area where the road was constructed. The parliament adopted Macadamization nevertheless, because it provided much more legibility to plans and costs than other methods. The high capital cost of parliamentary roads can be considered the cost of legibility.

There was growing opposition in Britain to centrally-planned parliamentary highways in the late 1800s — the strain of high costs on the parliamentary budgets being the primary reason. This led to a comeback for turnpikes and privately maintained toll roads which, once again, proved to be inefficient and expensive for citizens. Parliamentary planning was eventually resurgent, and despite mixed success in the late 20th and 21st century, has remained the preferred modality for planning projects.

One of the bigger innovations in large infrastructure projects in the 20th century has been the utilization of project finance.(9) Starting in the late 1980s, a significant proportion of infrastructure projects (primarily road and rail) follow a project finance model in which the government spins off a private company, with the government as its largest stakeholder. Project finance offsets the risks of defaulting on loans and also enables the private company to enter into contracts with several contractors — providing more financial and incentive legibility to the project. The project finance model creates a “single wringable neck” for the delivery of the project. It also enables the project to raise funding from several sources such as international banks or sovereign wealth funds.

Even though the technology available has increased exponentially, the planning protocols for large infrastructure projects still remain recognizable to someone from the 1800s — project finance being an exception. The fate of Britain's 21st century high speed rail project shows how these protocols are not suitable for the hyper complex environment of today.

Planning and Distrust of Planning

The British high speed rail project was announced in 2009 as a Y-shaped network from London to Birmingham with spurs to Manchester and Leeds. The budget was set to $20 billion. In 2023, after several revisions and re-scoping, the project was reduced to a single line between London and Birmingham.(10) Initially planned to open in 2026, the single West Midlands leg of HS2 is now delayed to between 2029 and 2033. A budget of $55.7 billion was set in 2015 for the whole Y-shaped line, but the initial leg alone is now estimated to cost between $53 and $61 billion.

Critics of the project have pointed out that several aspects of HS2 lack any coherent planning. HS2 was launched as a charismatic political project by Gordon Brown’s Labour government. The stated goal of the project was to connect the economic center of London with the deprived north of the country. The planning process appears to have worked backward — instead of understanding the problems and then scoping the project, the project was scoped first. The “planning” phase essentially involved hand-picking data that fit the assumptions of the project.

For example, HS2 was marketed by the government as a much cleaner and environmentally-conscious means of transport compared to car travel. This was later found to be based on hand-picking a single study about a French rail line.(11) Studies commissioned by another government body, the Department of Transport, suggest that when construction is taken into account, high speed rail journeys from London to Manchester will produce 60% more carbon than conventional rail and 35% more carbon than car journeys.

It appears that even route planning for the project was haphazard. After the route for HS2’s now-canceled eastern leg was published, it emerged that the line cut right through a brand new housing estate. It turned out that engineers had worked from old maps that were made before 2011, prior to the estate’s beginning construction in 2012.(12)

The business case made by the government for the West Midlands leg which is under construction is that the project produces benefits of $32.3 billion with a cost to benefit ratio of 2.7. Most of the benefits produced appear to come from estimated amounts of time saved by passengers, the calculation of which appears “frightfully complicated”, according to government sources.

In their paper Governance Regimes for Large Complex Projects, Roger Miller and Brian Hobbs studied 60 large infrastructure projects and found that the front end of these projects, which includes planning and discussions with all stakeholders, took close to seven years on average and were often expensive (up to 33% of the total budget).(13) They also found that this phase was critical and showed more impact on project performance than management of the engineering, procurement and construction phases.

Under the present conditions, unless you live in a monarchy or under an authoritarian regime, seven years is longer than the length of an average political term. This means that if a project is primarily politically-motivated, such as is the case with HS2, it is likely to be dead in the water before planning is done, or its planning is done in such haste that endless mistakes are discovered once the budget is “approved” and machinations for the project are underway. In such cases, the budget is in fact just an imaginary number that sounds about right to someone. The death by thousand budget cuts of the HS2 project was complete when the conservative government of Rishi Sunak canceled the northern leg of HS2, which had been the cornerstone of the project as envisioned by the Labour government which initiated it.

Ill-planned, charismatic projects such as HS2 also compliment and fuel inefficient bureaucracies. In the paper From Weber to Kafka: Political Instability and the Overproduction of Laws, Gabriele Gratton et al. argue that political instability leads to production of legislation that is not properly planned, which the bureaucracy struggles to implement, which then leads to Kafkaesque bureaucratic nightmares.(14) The result of this, as the paper notes, is that “bad bureaucracy drives out good politicians”, and the cycle of bad legislation continues. Gratton et al call this cycle Gresham's Law.

The transition from Weberian efficiency to Kafkaesque nightmare can be due to three reasons that the paper points out: shorter parliamentary terms, large number of reform opportunities and — ironically — a technocratic government (too many competent people producing too many laws). In the case of HS2, the subsequent parliaments since Gordon Brown’s have gradually lost interest in the project. It is also worth noting that Britain has been going through high political instability, rotating through four prime ministers in the last six years, in addition to Brexit and euro bloc instability since around 2015. High political instability is combined with a large number of reform opportunities as a result of a rotating cast of prime ministers whose terms of office are too short to properly implement policies.

Chinese Exception

One of my fond adolescent memories, growing up in a country which struggled to build new infrastructure, is my father telling me about his visit to China, and the time he got to ride a high speed rail line. He said,  “You don't realize how fast the train is moving, then the tour guide tells you that a train is going to pass in the opposite direction and that the relative velocity at that point would be greater than the speed of sound — it happens so sudden, blink and you miss.”

China has built 25,000km of high speed rail line since the year 2008.

A key role in the development of China's HSR is played by the MLTRP. This plan, first approved in 2004, with revisions in 2008 and 2016, looks up to 15 years ahead and is complimented by a series of five-year plans. These plans are rarely changed once approved. The initial long term plan envisioned 12,000km of high speed rail by 2020. The 2016 revision aims for a network of 30,000km by 2020, 38,000km by 2025 and 45,000km by 2030.

The planning for these high speed rail projects is a mix of highly aligned centralization and incentivizing stakeholders at all levels. For example, both central and local governments have equity in the high speed rail projects — the former through capital and the latter in the form of land provided for the project.

Private participation in the projects was low at the beginning because the relatively low fares of the HSR made cost recovery hard. But since the high speed rail project began, the per capita disposable income of Chinese urban residents has risen from Y14,000 to around Y34,000, thus allowing for fare increases in some routes, leading to more private participation in running these routes.

In addition to keeping the prices low, the Chinese government also implemented legislation that prevents airlines from changing fares for routes less than 800km — the distance range within which high speed rail competes with airlines. These regulations have also been relaxed in the last couple of years — showing signs that the government is loosening its grip on transportation and letting free markets play more of a role.

All of the above shows the high pedigree of synchronized execution of legislation and highly detailed planning that is needed to make large infrastructure projects possible.(15a/15b)

The Chinese state also does not risk becoming a Kafkaesque bureaucracy yet — the legislature term is long, the successions are planned, the state tries to quash or manage instability early on. Perhaps the biggest risk China runs into is the third one: becoming a highly centralized technocracy. In short, the nature of infrastructure projects in China appears to be the Weberian ideal of bureaucracy combined with an Orwellian dystopia of control.

Is it possible for large scale infrastructure projects to be planned and executed in the absence of a highly centralized technocracy?

Plans as Autonomous Worlds

As Ludens writes in Autonomous Worlds, “A World is a container for entities and a coherent-enough internal ruleset about how they behave. When a system of entities and rules comes to life, it becomes a World.”(16) A large-scale infrastructure project is a world of its own. The entities of this world are bound together by legally binding contracts which define the rules of their behavior and roles within the world. An entity is considered to be part of the world if it follows the introduction rule of the world. Introduction rules define the boundary of the world that you are dealing with. For example as Ludens writes, “In the World of Harry Potter, the introduction rule is very simple: if an entity is included in a story written by JK Rowling and published under the Harry Potter series, it is diegetic. Otherwise, it isn’t.”

In the case of the Chinese high speed rail, the introduction rule is the MLTRP. An entity is part of the world of Chinese HSR if it follows the rules defined by the MLTRP.

Projects with well defined introduction rules fare better than ones with more loosely defined introduction rules. A clear introduction rule creates hard boundaries that are easier to enforce for a competent bureaucracy. In their paper comparing the governance of two nuclear reactor construction projects Olkiluoto 3 (Finland) and Flammanville 3 (France), Inkeri Ruuska et al look at the efficiency created by well defined contracts and roles.(17) They find that even though both projects run into troubles, Flammanville 3 fares better because of the well defined relationship between the entities. Another factor that the paper points out is the existence of a previously defined relationship between the entities working on Falmmanville. Entities such as construction companies and governance agencies are worlds of their own, so it makes sense that the existence of a previous relationship makes the entities more aware of each other's internal rulesets and protocols.

Clearly defined introduction rules are also part of the reason why efforts to quash infrastructure projects find more success than building them — the only introduction rule for stopping HS2 or passing zoning regulation in a large city is “are you opposed to the plan or not?”

Coordinated action within blockchain has been largely successful in achieving goals in projects where the introduction rule is simple and well defined (such as constitutionDAO: buy the constitution), or in the case of large multiplayer games (such as memecoins: number go up and have fun with friends) which do not require complex coordinated action other than “should number go up or down?” Perhaps the concept and tools of autonomous worlds can enable more complex coordination on high stakes projects.

Infrastructure Planning Game

A couple of weeks ago, at Edge Esmeralda, Arb from 0xparc brought up Twitch Plays Pokemon in a sidebar conversation. It is a game played on Twitch where the actions of the main character in the Pokemon game are crowdsourced from inputs received in a chat room. Initially the game was designed in such a way that all inputs were tallied and reflected in gameplay. After receiving several complaints about slow progress and confusion, the broadcaster introduced democracy mode and anarchy mode for gameplay. In democracy mode, the inputs from the previous 30 seconds are tallied and the input with a supermajority is played; while in anarchy mode, the inputs are tallied and the winning input is played without any requirement for supermajority. This got me thinking about an equivalent of Twitch Plays Pokemon for planning infrastructure.

The objective of the game could be to build a railway line between any two points within a region (the larger the region, the higher the budget and larger the uncertainty). The players have to come to a rough consensus on the routes for the trains. The budgets are calculated based on the route and there is a budget cap that is arbitrary or decided early on. They have two modes of playing — The Benevolent Dictator For X mode (BDFX) or Democracy mode.

BDFX mode is a play on BDFL, a term that is commonly used in open source communities when referring to people like Linus Torvalds, creator of Linux — leaders who are burdened with making the final decision when everyone else has come to an impasse. They generally tend to occupy this role for life or until they rage quit. In BDFX mode, everyone has to lead some of the time but not all the time. Here’s Venkatesh Rao writing about BDFX:

BDFxing is a pattern that has emerged (not surprisingly) in the crypto world, and in adjacent spaces that use the ideas, if not all the technologies, like Discord-based groups. In these spaces, many worthwhile activities and initiatives often fall into, and languish in, the no-man’s-land between theatrical or real BDFLs on the verge of failure, and collective action mechanisms that are incapable of acting. And since bleeding-edge protocol technologies literally don’t work if you can’t make the leadership model distributed too (see: Conway’s Law; many people have to run all these “nodes” these architectures rely on), BDFxing emerged as a way to semi-ironically announce that you’re unilaterally taking the initiative to lead an activity through a challenging regime where it would fall apart without leadership, and actually doing it sincerely, and then quitting when the need has passed.

Just enough leadership injected, when it is needed, where it is needed.(18)

For the purpose of the game, in BDFX mode, a single user can capture the flag of BDFX and for the next thirty seconds — or until the flag is captured by someone else — make decisions about the routing of the train.

Democracy mode would work similar to its namesake in Twitch Plays Pokemon — all players input their decisions on what route the train should take or what the budget allocated should be, the inputs from the previous 30 seconds are tallied, and the input with the supermajority is taken.

Decisions for routing the train should be based on:

1. Expected returns in the next 10 years: This includes revenue from fares and other realized gains such as taxes on land from increase in land value along the route

2. Environmental assessments: A highly stochastic variable in the game, such as locals protesting the removal of a tree or a park

3. Expected political value: Unsure how this should be calculated but I imagine it would be something along the lines of, more constituents who favour urbanism being served by rail line more the increase in your political strength.

The game ends when the budget cap has been reached.

It would be interesting to compare the routes that are created in democracy mode and BDFX mode. My bet is that the BDFX mode will be the right mix of entertainment and efficiency.

Conclusion

Planning and completing large infrastructure projects has gotten harder over time. In the early years, planning for infrastructure projects was militarized, with the objective of providing more legibility and control for an empire.(19) When parliamentary planning took off, it borrowed heavily from the standardized practices of the military and technological advancements made in route-sighting and planning. The push for a more standardized method of building out large infrastructure projects led to the formation of the field of civil engineering. Standardization, such as Macadamization in the case of the British highways, allowed infrastructure that is used by the public across provincial borders to be much easier to manage, while also providing a better experience for the public.

The project finance model, popularized in the 1980s, was an attempt to bring the efficiency and incentive alignment of private enterprise into building infrastructure. This has had mixed success — because a government still remains the largest shareholder in a private company spun off to build an infrastructure project. The project finance model enables a government to have limited financial liability and allows it to enter into contracts with private contractors, which would have been cumbersome in the previous system.

The ossification of western democracies has led to what Jo Freeman called a tyranny of structurelessness when it comes to large infrastructure projects.(20) In her paper, Freeman studied women’s movements in the 70s. The anarchist, anti-structure movements worked well to bring people together but failed when it came to actually taking action. While the context is different, the same cycle seems to happen in the case of politically motivated infrastructure projects: first, informal coalitions of politicians and bureaucrats emerge, with unclear norms and unwritten rules. The authority within the group is then concentrated to a select few elites. Status games will ensue as individuals in the group jockey for position. Once this happens, the group loses focus on its original goals, making progress next to impossible. High political instability (such as four prime ministers in six years) means that neither politicians nor bureaucrats are able to provide the stability required to finish complex projects. The exceptional performance of China here is the proverbial exception that proves the rule.

While high authoritarianism is offered as the primary critique of the Chinese model, it seems to me that 25,000km of high speed rail would be hard to construct without some BDFXing at various levels. High authoritarianism is over-theorized and BDFXing is under-theorized. The Chinese exception requires further study and perhaps more conversation with people who BDFX such projects successfully.

The Infrastructure planning game could possibly be an experiment to familiarize people with the concept of BDFXing and getting an intuition for it.

In the time that I’ve written this essay, Britain had another general election, Austin’s urban light rail plans have run into legal hurdles,(21) and China’s bullet trains have raised prices due to debt concerns.(22) Planning for large complex projects keeps getting harder and weirder in a world where information and decision cycles are faster and more unpredictable. Responding to this probably requires a new model of planning that is as influential as Macadamization, project finance and China’s MLTRP. Studying emerging models and analyzing their efficacy could be a fruitful area of study.

***

Thanks to Josh Stark and Tim Beiko for their feedback on this essay. Thanks to the Ethereum Foundation for supporting my work with a small grant.