RoamLetter Sample – First Edition

Essays of the week​

On Chesterton Fences, by Luca Dellanna

Writer G. K. Chesterton noted that people react in two ways, when faced with a fence across a road whose use they do not know. Some move to break it down, while others wisely wait until they understand its use. After all, whoever built it must have had a reason. Tearing it down might have unforeseen consequences. The same applies to laws and institutions. We should be wary of calling out the uselessness of what we do not understand.

Our world is full of Chesterton Fences: objects, institutions, and rituals apparently inefficient and yet effective. For example, in his book “Alchemy”, Rory Sutherland explains how London taxi drivers have to spend years memorizing every single street. This seems useless and inefficient, especially in the age of GPS navigation. And yet, it has its use: it is an upfront cost designed to weed out unmotivated candidates who might later provide a degrading service.

​Chesterton Fences in business

One component of the typical stock market cycle, “growth → bubble → crash”, is investors increasingly treating common-sense accounting practices as useless fences bounding the potential of stock prices. Those who do usually discover a few years later that such practices did have a purpose.

Some of the most famous companies of the 21st century, such as Uber and Airbnb, seem to be built upon the destruction of fences. Uber obsoleted taxi licenses and Airbnb the lengthy process of befriending the owner of a house where you’d like to live for a week. However, Uber and Airbnb didn’t merely destroy the fence. They acknowledged its use and built a surrogate. Uber gave drivers something else to lose: a 5-stars rating. Similarly, Airbnb introduced a two-way rating system and provided (limited) insurance to both home-owners and visitors. In fact, in a previous essay, I argued that Uber and Airbnb are Trust-As-A-Service businesses: for a fee, customers can buy trust in transactions with strangers.​

Chesterton Fences and Ergodicity

In my recent book on ergodicity, I noted a difference between what matters when we consider narrow intervals and what does when we consider broader ones. For example, closing a sale using deception is efficient only in the short term and inefficient in the long one. Over the short term, consequences that apply beyond the short-term do not matter. Over the long term, they do.

We often ignore this banal principle in the assessment of Chesterton Fences. We use our assessment of their short-term usefulness as a proxy for the long-term one. This assumption is often unwarranted.

A similar mistake was committed during the early months of the COVID-19 pandemic. Face-masks were deemed useless based on narrow studies. Their effectiveness was measured over short periods of time, over small spaces, and over small populations. For example, during hospital studies. Instead, face-masks are effective over large periods of time, over large spaces, and over large populations. This is because they only offer partial protection that compounds – and compounding phenomena can only be observed at large scales.

The genesis of Chesterton Fences

Dio Mavroyiannis recently noted, “If the Fence is a result of bottom-up processes (say evolution), Chestertonian skepticism is justified. If an academic put the fence there, you can probably get rid of it.”

This argument is similar to the one I made in my “The Power of Adaptation”, in which I argued that bottom-up ideas that spread are more likely to be good than top-down ones. This is because bottom-up ideas tend to spread because they proved to work, whereas top-down ideas tend to spread because they make sense, i.e. before their effectiveness got proven.

Of course, this is not an absolute rule. In particular, it lost power during recent times in which social media allow ideas to spread the bottom-up way before their effectiveness (or dangerousness) has time to emerge.

(Readers familiar with The Lindy Principle might have noticed that the rule “bottom-up ideas that spread tend to be effective” is a proxy for Lindy if and only if bottom-up ideas require time to spread.)​

Maturity

The desire to destroy Chesterton Fences is a characteristic trait of immaturity.

My definition of maturity is “ownership”. One is mature when he owns his mistakes and his future. The only way to do so is through learning – and Chesterton Fences are a great learning opportunity. When you see one, pause and attempt to understand its builders (which is different from putting yourself in their shows – you don’t want to know their point of view, that’s banal; you want to know their experiences and their motivations). Perhaps, you’ll conclude the fence is obsolete, or maybe, you’ll conclude that it still has a purpose. Either way, you’ll have learned something and avoided a mistake.

On the necessary vs the sufficient, by Luca Dellanna

Today, I read the following tweet.

“Making work remote without also making it asynchronous only makes it worse. You get the isolation of working alone without the benefit of control over your schedule.” – Tiago Forte, link​

Companies transitioning to working from home will soon discover that adding Zoom is a necessary but not sufficient step. Other processes must be tweaked. For example, most communication must become asynchronous – at least for technical roles.

Some of the worst mistakes happen when you do something necessary believing it is sufficient.These mistakes are vicious because, if you believe that what you’ve done is sufficient, you won’t discover nor address problems until it’s too late.

This is not unlike cargo cults. They also mimic something necessary believing it is sufficient, then get frustrated for the lack of results.

We’ve seen this during the COVID-19 pandemic. Countries cargo-culted Chinese lockdowns believing it would be sufficient to stop the spread of the virus. This led them to neglect the other measures that the Chinese took, such as mass door-to-door testing. The result: Western lockdowns lasted longer and were less effective. That’s the cost of mistaking the necessary for the sufficient.

If I could go back in time when I was 15, one of the things I would tell my past self would be: “When you fail at something, do not immediately conclude that what you did was wrong. Instead, consider that you might have done something necessary but not sufficient.” This is crucial to ensure that we react correctly to failure – not by losing motivation, but by understanding what ingredient is missing for our success.

On the pandemic: “What to expect from the second wave” by Luca Dellanna (link)

A common question is, should we worry about the second wave, even though there aren’t that many deaths?

I expect deaths to rise soon. Mortality is a function of hospital access. The more cases, the more hospitals clog, the fewer patients can be cured, the more death. This applies both to COVID and non-COVID deaths.

Moreover, if mortality were confirmed to be a function of viral load, it would rise further. The more people get infected, the more people exhale the virus, and the more virus there is to breathe in.

In addition to deaths, I would expect more logistical problems. Stronger ones, as their severity grows exponentially. This winter, our governments will have to solve problems relative to hospitals overcrowding and the procurement and distribution of treatments and, perhaps, vaccines. In March and April, many industries brilliantly solved novel supply chain problems, but I’d be wary of extrapolating the performance of bottom-up entities such as the markets to that of top-down governments.

Knowing what to expect is hard, so let’s say what not to expect this winter. A fixed date for the first public vaccination campaign. A working delivery logistic plan. Herd immunity. A silver bullet. Maybe one or more will happen, but I wouldn’t make plans that assume their occurrence.​

On legibility and centralization: “Big Tech Sees Like a State”, by Byrne Hobart (link)

Note: the below is not a summary of Hobart’s essay; the linked article is complex, well-written, and touches on many points. Instead, it is the abstract of one of its points, and a mini-essay of mine built upon it.

What I learned

In “Seeing Like a State”, historian James C. Scott explains how governments increase legibility to tax their citizens and prevent them from rebelling. They assign them names, addresses, and IDs. They standardize occupations and outputs (e.g., what a company produced). With all of this, they know how much tax they can collect from each business and citizen.

Byrne Hobart notes that the biggest legibility imposers today are big tech companies.Think about Facebook assigning an ID to billions of people, or Amazon having a front seat to global supply and demand.

My comments

Centralization and legibility strengthen each other. On the one hand, the central management has a necessity to improve the legibility of its dominion. On the other hand, legibility favors winners-take-all mechanics.

This week, the EU Council Of Ministers has been reported working on banning end-to-end encryption in apps such as WhatsApp. Legibility uber alles.

Similarly, it has been recently revealed that the Swiss government, known for its impartiality, helped the CIA control a Swiss company that secretly sold rigged encryption systems to foreign governments. (link). If you want to tempt an “incorruptible” government, do it with legibility.

I previously wrote that centralization is only efficient to the central observer. Someone who is geographically, socially, or culturally close to the center of an organization is more likely to have a positive opinion of it. Similarly, a centralized organization is more likely to choose to pursue objectives that are beneficial to those that share the location, class, or culture of its top managers.

This is not necessarily due to tribalism or self-interest, but to data compression. The central management can only measure what it understands. It can only solve problems it feels, and can only do so using the tools it has. A reason why centralization is limiting and inefficient.

Centralization has three sources of efficiency: economies of scale, coordination at scale, and standardization of legibility. However, these are only beneficial up to a point and become inefficient afterward.

Economies of scale implies standardized outputs. This is good in the measure that the needs for such outputs are standardized, and bad otherwise.

Coordination at scale means larger successes and larger mistakes. In an ideal ergodic world, the former offset the latter. However, in the real world, some mistakes are so large that no success can offset its irreversible consequences.

Standardized legibility is good in the measure that economies of scale and coordination at scale are good, and bad otherwise.

We overestimate the benefits of centralization in the medium term, and underestimate its long-term costs.

In the short-term, centralization and legibility are inefficient due to the high up-front costs. It is hard to figure out how to centralize the right way and to gather the resources to do so.

In the medium-term, centralization and legibility are the most efficient.

In the long-term, centralization and legibility collapse. Either their coordination capabilities make them commit a mistake too large to recover from, or they collapse under the weight of their inefficiency, as perceived by the population that is geographically, socially, and culturally distant from the central management.

Quotes of the week

“Life is easier when you know who you are, simplest when you know what game you’re playing, and best when you play only your game.” – Theresia Tanzil, [link]

​

“A surprisingly repeatable source of insight about the world: seek deep knowledge of a transaction that your peer set does not realize exists.” – Patrick McKenzie, link​

There are many applications. For example, as a software engineer working for a bank, you might specialize: in transactions for the restaurant industry, or in “status transactions” between stakeholders in a project (i.e., how to move a project forward), or in collecting feedback from the customer – not about mere bugs, but about what tasks of theirs are irrelevant now but might become a business in five years.

The best advice makes implicit “rules of the game” legible.

​

“Behavior is contagious because we catch it from other people. […] As in virology, people’s susceptibility varies. If you do not own a microwave now, it is unlikely that you will ever buy one — you are effectively immune. One reason it is useful to use virology as a model for behavioral adoption is that […] what matters is often not the value [of ideas] but their transmissibility.” – Rory Sutherland, link​

The same applies not just to ideas but to leadership and innovation. A good leader not only has good ideas but ensures that they spread. An inventor’s job does not end with the stroke of genius, but with having facilitated its spread across a population. Inventors who don’t acknowledge this second half of their hole end up frustrated and forgotten. (Frustration comes from believing that the necessary is sufficient.)

“The basics may be boring, but they lead to a non-boring life.” – Orange Book, link​

The best investment you can make at the beginning of your career is to spend the time to understand what its foundational skills are. If you don’t, two bad things will happen. One, because you’ll lack required skills, your development will stall and you’ll grow frustrated. Two, you’ll overcompensate with other actions, and the development of foundational skills will suffer. This is the case of the employee who overcompensates skill with overtime or of the manager who overcompensates clarity with micromanagement.

It is relatively easy to understand what the fundamental skills are in fields with skin in the game and no “promotion”, such as sports or craftsmanship. Imitation tends to work great, though you might require coaching.

Instead, problems arise in fields with no skin in the game. Because of that, charlatans can stay in the game, and you are at risk of imitating them and learning the wrong lessons. Choose your models wisely.

Similarly, problems arise in fields with “promotions” or “career paths”. Too often, people focus on skills that make them thrive at the current position but are irrelevant to get at the next one. They mistake the necessary (being good at their current tasks) with the sufficient (to get to the next level).

So, spend more time than it seems appropriate on understanding what skills you should develop. Often, the answer is not apparent, or you might not like it (and therefore need extra evidence to work on them). Either way, do not skip this step.

​

“Only scalars can be ranked. Single, one-dimensional numbers. People are not well described by scalars. You can’t rank them.” – Ole Peters, link

Ole Peters refers here to the fact that any ranking implies that (1) we can measure success (2), we all share a single definition of success, and (3) what’s required to succeed today is also what will be required tomorrow. In other words, rankings require either an omniscient hive-mind or self-centered arrogance.

The only situations in which all three points above apply are competitions with well-known previously-established rules. Success at them is poorly correlated with success in life.

People who disagree with the last statement must be careful not to have arbitrarily created a set of rules for an imaginary competition for whom the rest of the world might not have signed up.

​

News​

Pfizer announced that it developed its apparently successful vaccine trial without taking government subsidies because of the added bureaucracy they would come with. When, earlier, I wrote that “centralization is only efficient to the central observer”, I also meant that “bureaucracy is a cost external to the center, therefore not counted in the self-assessment the center makes of its efficiency”.

​

Amazon hired 400k people this year so far. It makes you wonder what information about the job market and/or HR are legible to Amazon and a few other companies.

​

Italy’s “click day” to access micromobility subsidies causes a nation-wide breakdown, as 400k people wait in an “online queue” requiring a click every 20 minutes to avoid having to start all over, and the authentication system breaks down on all websites of the public administration.

Centralization ensures that mistakes hit the maximum number of people. If the 20 regions of Italy each had their own click day, on different days, it would have been clear that it was a terrible idea after the failure of the first region. The other 19 would have learned their lesson and tried something different. But because centralization mandated experiments to be performed nation-wide, nation-wide breakdowns ensued.

​

“A partner at McKinsey told me that their profits are expected to be up 10-12% this year just by virtue of not having to travel to clients” – Sheel Mohnot, link. “Amazon saved $1B in travel expenses this year because of COVID-19. A chunk of it will come back, but something tells me it won’t be close to 100%” – Tanay Jaipuria, link​

​

“The average Uber Eats customer orders from 3.6 restaurants. Really begs the question of how important marketplaces are or if the winning model will be restaurants going direct to consumer (D2C)” – Damir Becirovic, (link)

​

Fun & satire​

“Pollsters Claim Their Polls Were 100% Correct, But Everyone Voted Wrong” – @TheBabylonBee

“Tour De France: cyclist tested positive to everything except COVID” – Lercio

​

Prompts

How will your life be in 10 years, if I keep doing what you’re doing every day?

​

This was the first edition of Luca Dellanna’s RoamLetter​, displayed as a sample. Click on the button below to subscribe and receive more, in your favorite format.