What came to me in the past three years (2020–2022).
The unexamined life is not worth living.
— Socrates
Miscellaneous Thoughts
- Three strategies to solve engineering problems:
- Abstraction
- Decomposition
- Iteration - Just as you can’t improve it if you don’t measure it, you are not likely to fix a bug if you don’t reproduce it.
- Don’t rest on how things work in one way. Think of alternatives and reason about the trade-offs.
- How can testing and CI go wrong?
- Tests are not automatic. Every time someone has to manually check whether something still works.
- Tests are brittle due to overly rigid expectations, and thus become diff checkers.
- Tests are flaky due to non-hermeticity. They rely on something beyond the single machine or even something not built from the same source version.
- The system under testing is non-deterministic. This is not the fault of a test, but ends up harming testability.
- The presubmit set consists of too many medium or large tests. They increase the cost of continuous integration and create frictions on developers who submit code.
- The automatic culprit finder is missing in post-submit. The on-caller has to make non-trivial analysis to pinpoint the bad change. - Shift left name resolution to compile time by letting users include your header files. Failures at link time or runtime is way harder to decipher.
- A single repository may hide the fact that code needs to read or write different versions of data or configurations. Any change to data schemas should take compatibility into consideration.
- Tech managers are often faced with a tradeoff between short-term performance and long-term development of their teams.
- Rules and laws influence people by exploiting their rationality, whereas culture and norms rely on social psychology.
- Ideologies and technical decisions are often delusions of the underlying contention for power.
- Political entities neither are rational nor can be personified, as they are composed of irrational individuals with potentially conflicting interests.
- We need friends not because of enemies, but we combat enemies to protect our friends.
- An epistemological way to categorize things:
- Those that can be physically observed.
- Those that must exist in theory based on our observations and logic.
- Those that we are not sure whether exist or not.
- Those that cannot exist otherwise it would contradict our observations and logic. - Determinism is neither sufficient nor necessary for predictability. There are stochastic systems that exhibit crystal patterns and Newtonian systems that are incredibly chaotic.
- What makes you comfortable is not necessarily what is correct to do. In many cases, a pain is a signal that you are on the right track to grow. (Inspired by 《认知觉醒》.)
- Health and emotional stableness are the foundation of diligence and intelligence.
- Giving vent to emotions or desires usually doesn’t calm them down. Exercising self-discipline tends to strengthen willpower.
- When you ponder why something makes sense, you are rational. When you ruminate on how something doesn’t make sense, you are emotional.
- Rationality needs to be balanced with pragmatism. Follow your instinct to wade through uncertainty.
- For example, teleological viewpoints don’t reveal reality, but may provide directions to dive into reality. - The human mind is biased towards
- Increasing odds of survival;
- Enhancing sense of security;
- Eliminating cognitive dissonance;
- Reducing costs of cognition. - Opportunities are like trains. Prepare early and wait with patience, since they won’t await you once they arrive.
- What the market demands often takes precedence over what we wish to supply. The environment doesn’t change for us, but we could adapt ourselves to the environment.
- Be a smart pig who takes every chance to approach the next potential whirlwind. (Inspired by Lei Jun’s quote.)
- Take calculated risks instead of blind gambles. Base your speculation on observable trends or driving forces, not on a hypothetical tendency of regression-to-the-mean.
- Liquidity counts more in the short term. Even if the expected return of a repeated game is positive, it doesn’t mean you are able to “survive” to the point when your accumulated investments pay off.
- Fluctuations are not necessarily a sign of vulnerability, but could be useful signals for helping a system to stay in health. (Inspired by Antifragile: Things That Gain from Disorder.)
- Sacrifice of constituents is the cost for a system to step out of local optimums. (Inspired by natural selection.)
- A good system gets better when each component drives towards its own optimum, just as a good market enjoys prosperity as individuals pursue their self-interests. (Inspired by incentive compatibility.)
- Don’t take a risk that doesn’t give you the enough risk premium, and don’t expect to get paid without taking the necessary risk. (Inspired by MIT 15.401.)
- A person’s risk preference is a functor that maps return distributions to utility values. Diverse risk preferences (along with heterogeneous beliefs) underpin the financial market.
- The market tends to overreact to either exuberance or depression. Chances of making profits come from mis-predictions of other players in the game.
- Mis-predictions are those due to lack of information or suboptimal decision-making, excluding those due to aleatory uncertainty in reality. - Knowledge should be a continually updated belief state that approximates objective reality. (Inspired by the Bayesian filter.)
- The world never goes wrong. It is our mental model that underfits it.
- A scientific theory is like a null hypothesis in statistical inference — it will never be proved perfectly correct, but always stands a chance to be falsified.
- Simplicity is a feature of linearity, and bias is the price we pay for it. Be prepared to see an ongoing trend to accelerate or diminish.
- The power of humanity is based on our sociability. Most of our opinions come from others, and most of our knowledge lives outside our minds. (Inspired by Sapiens: A Brief History of Humankind.)
- Pay more attention to your policy than the state of your environment. What really deserves worrying is not a financial loss itself, but the fact that your strategy doesn’t work well. (Inspired by the Markov decision process.)
- A plan is a sequence of states and actions assuming the state transition is deterministic, while a policy is a mapping from states to actions tolerating uncertainty. Having a plan is better than nothing, but is not as robust as having a policy.
- History is full of survivorship bias and hindsight bias. Retrospection is a useful but inferior substitute for on-policy learning.
- “Set systems rather than goals.” Thrive in the long run, not just follow adrenaline rushes.
- No action is an action on its own. Procrastination takes an opportunity cost.
- It’s more informative to look at a man when he’s in adversity than in success. A crisis is just the start of your showtime.
- Appreciate those who care about others’ feelings even when they are in bad luck themselves. Smile, even if you are sad.
- Embrace imperfection. Enjoy every challenge that you are not well prepared for, at any unexpected place, at any improper time.
- Don’t expect to follow the paths you traveled in the past, since history doesn’t always repeat itself. Enjoy attacking something new.
- Don’t strive to win a race. Instead, strive to create unique value for the society or just for the few people around you.
- “Be the change you want to see in the world.” If you feel the pain, you move the needle.
- People are mostly creatures of their environments, which further justifies efforts to change for better environments.
- Life is bound to be an endless war due to how evolution shaped the human brain. Just prepare for more battles down the road and enjoy the current game.
Thank you for reading this!
