The Directed Life

Notes from Farnam Street’s Great Mental Models Volume 3

Date read: 1/25/2023

Part I: Systems

Feedback Loops

When the outputs of a system affect the system’s very behaviors. Can either be balancing (towards equilibrium) or reinforcing (process amplifying). (24)

The faster you get accurate feedback, the more quickly you can iterate to improve. (25)

Make sure to create correct incentives and correct at the margins. (28)

Unfairness often has a purpose—specific unfairness can lead to most amount of fairness in the macro, justice. (30)

Margins matter because it’s often where feedback loops start. Think: who leaves first, a long-time patron or the person on the verge of buying / one-time purchaser? (31)

Through many iterations, by paying attention to and incorporating feedback, we end up becoming more capable. (41)

Equilibrium

Majority of real-world systems experience dynamic equilibrium, fluctuating within a given range. (45)

Short-term deviations from equilibrium are often what is needed to maintain it in the long term. (47)

Bottlenecks

The limiting factor for production of outputs. Once you fix one bottleneck, there will be a new limiting factor. (63)

Liebig’s law of the minimum: growth of plant will be limited by the least available required nutrient. (65)

Instead of addressing bottlenecks as they appear, your time might be better spent on a root-cause fix that makes a foundational improvement that leads many bottlenecks to disappear indefinitely. (73)

Bottlenecks push us towards new and better ideas to solving the same problems. (75)

Scale

As growth occurs, resilience can be increased by keeping a measure of independence between parts of a system. (80)

Staying small helps with longevity. (82)

Scale is often non-linear.

Margin of Safety

A margin of safety is often necessary to ensure systems can handle stressors and unpredictable circumstances. This means that there is a meaningful gap between what a system is capable of handling and what it is required to handle. (97)

Backups—multiple paths—improve a system’s resiliency. (98)

Ongoing learning is a key aspect of margin of safety. (101)

Don’t let your ego get in the way of knowledge accumulation. (104)

Benoit Mandelbrot on risk: (105)

The professionals plan for “mild randomness” and misunderstand “wild randomness.” They learn from the averages and overlook the outliers. Thus they consistently, predictably, underestimate catastrophic risk.

Broad competence seems very costly compared to specialization, but it is more likely to save us in the outlier situations of life … We are all going to face extreme events where failure is disastrous. (109)

Churn

The process of attrition purging the unnecessary components of a system. Churn is also helpful in assessing product-market fit. Retention rates greatly increase customer size. (113-114)

Algorithms

Method followed for making a calculation; the set of steps that produce a solution given the necessary inputs. (127)

Defining characteristics: (128)

• substrate neutrality
• underlying mindlessness
• guaranteed results

Algorithmic thinking: refining which inputs you’re implementing for deriving results. (136)

By testing various inputs in a repeatable process, you can use the results to refine what you feed into the algorithm.

We can also conceptualize them as if-then processes that are useful because they can help us ignore variables that don’t matter and focus on requirements. (141)

Critical Mass

The final unit of change that produces a disproportionate impact—the tipping point. (147)

The Overton Window: the range of acceptable ideas shifts slowly from radical to popular. (149)

In cities, it’s not the amount of infrastructure that produces interactions, it’s how that infrastructure is laid out. (156)

Emergence

When systems are unpredictable from just looking at their parts; elements combine that produce something unimaginable before the parts combined. (167)

Chess is emergent: new outcomes come from simple, unchanging rules. (170)

If you have a simple starting point on the right trajectory, surprising things can pan out through the power of emergence. (175)

Irreducibility

Reducing a thing to the minimum amount required for it to remain the same thing (187).

Gall’s law: complex systems always evolve from simple systems first (197).

Law of Diminishing Returns

The law of diminishing returns posits that inputs to a system lead to more output, up until a point where each further unit of input will lead to a decreasing amount of output. (201)

Societies collapse because of the ever-increasing complexity that requires more energy. (215)

Part II: Mathematics

Compounding

Compounding stems from a power law distribution—non-linear and outsized returns with time. (227)

Incremental efforts over time allow for enormous, outsized future returns. (229)

Sampling

Data needs to be collected in order to be analyzed, so you need to ask yourself if you have the right mechanisms to collect the data that will give you the fullest picture possible—or at least enough to make a good decision with the potential for good outcomes. (261)

Randomness

Engage with randomness and harness it as a tool for creativity. (271)

Regression to the Mean

Luck is random. So outlier results with a luck component are probably followed by more moderate ones. This is regression to the mean: when data far above or below the mean is more likely to be followed by data close to it. (283)

Repeated iterations provide eventual convergence to most likely outcome. (284)

Multiplying by Zero

It’s always worth investing effort in the weakest part because nothing else can compensate for it. (299)

Find the zero in the system before time and effort in any other part of the system. (302)

Quote made me think of Taleb’s advice on needing to tinker instead of ponder, convexity bias:(304)

The history of invention shows us that smart people fail dozens of times before they succeed. They build on the failures of others by testing their own hypothesis, tinkering and refining, and learning a remarkable amount in the process.

You earn knowledge through failure.

Don’t allow excessive dependency on one thing that could fail. (308)

Equivalence

The work of a researcher is the product of a lifetime spent absorbing the work of others. (320)

Surface Area

Understanding when increasing exposure helps vs. when it hurts—there are different benefits and consequences of being small / large. (333)

Thus, when we communicate, it might be helpful to reduce surface area to provide useful content. (342)

Global and Local Maxima

Often you have to go down to get back up. (347)

Changing the scale at which we are optimizing gives us perspective on where to go and how to get there. (356)

Understanding that sometimes we have to go down in order to climb even higher helps us make short-term sacrifices to play the long game. (360)

Supporting Ideas

Complex Adaptive Systems: When the components of a system are greater than the sum of its parts and the system adapts to environmental change. There’s complex interaction between components as well as system “memory,” where the system is impacted by previous events. (142)

Chaos Dynamics: Modeling outcomes in complex systems necessitates correct starting conditions which will always be unknown to us. Deterministic predictions will always fail and we have to equip ourselves to thinking probabilistically. (181)

Distributions:

Characteristics for identifying distribution type (220):

• Is data continuous or discrete?
• Symmetric or asymmetric?
• Upper and lower limits?
• Likelihood of extreme values?

Don’t forget or underestimate long-tailed events—life is distributed more power law-wise than normally. (221).

Lesson from Epicurus: (222)

The state we should aim to be in is at the top of a normal distribution curve—a life free from pain and also free from the negative consequences of excess pleasure.

Also from Epicurus: (222)

Therefore, being accustomed to simple, not extravagant, ways of life makes one completely healthy, makes man unhesitant in the face of life’s necessary duties, puts us in a better condition for the times of extravagance which occasionally come along, and makes us fearless in the face of chance.

Network Effects

Network effects set off a reinforcing feedback loop wherein the added value attracts new users, who in turn create new users. (247)

Once a network effect occurs it creates a sizable competitive advantage.

Pareto Principle

20% of the inputs result in 80% of the output.

Inputs and outputs are not evenly distributed. Not all inputs lead to the same sort of output. (279)

Order of Magnitude

Measurement of scale to aid human perception of things equally small and large, providing context when comparing numbers. (329)