• Atomic Habits

    A book about building systems of small habits that compound over time. Even small changes add up to big thingsβ€”1% improvement every day result in 37x improvement.

    About Habits

    Makes the distinction between goals, systems, and identity as agents of change. You might have a goal which motivates you to make changes, but once that’s gone it’s easy to regress. Therefore focusing on the systems is a better strategy to consistently improve. Finally, identity (your set of beliefs) has a large impact on behavioral changes–if you see yourself as a person that does X as part of your identity, you are very likely to keep doing X. Actions should be dictated by values not emotions. Habits change your identity–they provide evidence of who you are repeatedly.

    Habits reduce cognitive load, they’re automatic solutions to recurring problems.

    Habits are formed by following a loop–cue, craving, reaponse, reward. As a result, the reward gets associated with the cue. To create a good habit, make it obvious (cue), make it attractive (craving), make it easy (response), make it satisfying (reward). To break a bad habit, invert each step.

    Make it obvious

    Habit stacking–use existing habits as a cue for new ones.

    Implementation intention makes the likelihood of following through much higher by specifying the time and location up front. The level of apecificity is important–be precise and leave no room for thinking about it.

    Environments are powerful cues, in particular visual ones. Highly visible cues are effective at driving behavior as is convenience. Rather than an environment being a collection of objects think of it as a collection of relationships. Your relationship with drinks is different at a bar than at home. It’s easier to form new habits in new environments so you can build new associations and cues without wrestling with existing ones.

    You can break a habit but are unlikely to forget it. Cues are strongly associated.

    Make it attractive

    Supernormal stimuli creates a heightened reality which triggers a stronger response. Examples include junk food, social media. Highly engineered reality aims to make you form habits or addicitions.

    Anticipation of a reward causes us to act due to dopamine (released before the reward). This creates a powerful desire because the ‘want’ is stronger than the reward.

    Temptation bundling is when you connect something you want to do with something you need to do e.g. a stationary bike that plays Netflix only when you ride a t a certain speed. This is Premack’s principle in practice (more probable behaviors will reinforce less probable behaviors).

    Habits are often transferred from those closest to you so it’s best to surround yourself with people who embody what you value. Belonging to a group that shares an identity helps support the changes you are making.

    Behavior that get us approval, respect, and admiration we find attractive and are more likely to do.

    Cravings are an expression of a deeper underlying motivation e.g. obtaining food, social acceptance, reducing uncertainty. Habit forming products tend to latch onto these motivations which is why they seem addictive. However, there are many ways to solve a need–you dont need the taco to satisfy the need to eat food. Good habits find positive associations to relieve an underlying motivation.

    A craving is the desire to change your internal state–you desire to feel different which causes you to act.

    we find evidence for whatever mindset we choose e.g. I have to vs I get to.

    To create a positive association with a difficult habit, create a ritual that is associated with something positive and then link to the difficult habit by always performing the ritual prior to doing it.

    Make it easy

    Motion feels like progress but it’s not. Action is whats needed for a result to happen. Preparation is sometimes procrastination.

    Hebb’s law: neurons that fire together wire together. it takes repetition for habits to form. Habit formation is more related to repetition than time.

    Law of least effort: people tend to take the option that requires the least effort. Convenience is king.

    Reduce friction to new habits by designing your environment, make the cues more obvious, don’t require that you go out or your way to do it. This simlar to product design. Also similar to engineering principle, make it easy to do the right thing.

    Prime the environment by putting things related to the habit in their place or prepared for use e.g. leave the running clothes out the night before. The inverse also works for bad habits e.g. unplug the TV after watching to make it harder next time.

    30-40% of our behaviour is determined by habits. However a small habit can lead to many other things e.g. putting on gym clothes guarantees working out and all the downstream choices. It’s like path dependence.

    When starting a new habit start with something that take two minutes to start. Starting a gateway havit makes it much easter to continue doing it once you’ve started (convenience).

    Committment device is a choice you make to that controls all actions in the future e.g. Victor Hugo’s locked away clothes to make it unsuitable to go outside. Procrastination can be beaten by making it hard to do other habits.

    Automation is a way to make following good habits automatically. e.g. cash register makes it hard to steal. “Civilization advances by extending the number of operations we can perform without thinking about them.” -Alfred North Whitehead

    Make it Satisfying

    A positive emotion towards the result of a habit increases the likelihood it will be repeated e.g. mint flavored tooth paste leaves a clean mouth feeling instantly.

    Time inconsistency means we value the immediate more than the delayed. This explains why we think so short term, the instant reward is (in the moment) valued higher i.e. smoking while you know it causes cancer eventually.

    The cost of good habits is are in the present, the cost of bad habits is in the future. Instant gratification wins.

    To make habits stick you need immediate reinforcement (reward or punishment). a good example is with avoidance habits like no latte’s. To make the reward more immediate do something when you avoid it like transfering the money you would have spent into another account for a future purpose. The reward should be aligned with identity.

    Eventually the long term goal becomes the reward and short term reward is more to kickstart it. Incentives start the habit, but identity sustains it.

    Habit tracking is a way of making progress visible and rewarding after each successful activity. As an example, moving a paperclip from one jar to another after each sales call or marking off a calendar.

    When a habit streak ends, don’t miss twice in a row, start the streak again as soon as possible or it becomes a habit. In general, no zero days. Show up even if it is not your best.

    Consequences need to be immediate and painful enough to cause behavior change. Laws are an example e.g. seat belt laws provide accountability. A habit contract with others that outlines the goal and consequences is another way.

    Peak motivation comes from working on tasks that are not too difficult and nit too easy (goldilocks rule). To get into the zone, research suggest 4% beyond your current ability. Without variety and progress we get bored and are unlikely to stick with the habit. The greatest threat to success is boredom. You need to fall in love with boredom to keep showing up even when you don’t feel like it.

    Habits are not sufficient for mastery. To continuously improve takes deliberate practice otherwise you merely reinforce the same habit.

    Career best effort, a way of measuring progress that Pat Riley introduced to the lakers asking them to improve by 1% each year. They would reflect and review on performance regularly.

    Reflect periodically on identity and how your actions and habits reflect that.

    Keep your identity small (Paul Graham) means you won’t be resistant to change since any one thing is not tied up in your identity. This provides flexibility to grow and change.


  • Actions Should Be Dictated by Values, Not Emotions

    It’s healthier to determine the actions you take based on your values rather than your emotional state. When your actions conflict with your values you tend to feel worse since you’ve also compromised your beliefs. For example, not working out because you feel depressed is likely to make you more depressed if being physically fit is something you value.


  • Event Sourcing

    A method for representing state of a system using immutable events over time that can be aggregated to derive state at a particular moment. This software architectural pattern has the advantage of replay-ability (events can be re-processed as needed), time travel debugging (view of state at a specific moment in time), and decoupling systems (event emitters separate from consumers and aggregations).

    A real world examples is double entry bookkeeping. A software example is a shopping cart checkout flow.


  • ECS

    An entity component system (ECS) model is a software architectural pattern way of organizing a program that is oriented around and is easily extensible often used in games.

    Rather than an object oriented model which couples state and methods that act on the state, an ECS has an entity (a unique ID), components (data that belongs to at most one entity), and systems (functions that operate on components). This has the advantage of being easy to optimize (systems operate on contiguous data that can parallelized) and can model complex objects without common pitfalls of OOP (e.g. diamond problem of inheritance).

    See also:


  • In-Group Favoritism

    The tendency to give preferential treatment to those belonging to the one’s in-group over one’s out-group.

    Attempted explanations for why this occurs include 1) groups conflicting over the same scarce resources and 2) the psychological drive to raise one’s self-esteem through a distinct social identity.


  • The Universe Is Not Infinite

    A theory that the universe is actually finite due to having a nontrivial topology (i.e. like a mobius strip).

    An analogy of how you can know that something is finite, given by Janna Levin, is Pacman. The player exists on a 2D plane and when reaching one end of the screen, wraps around to the other side. We can make sense of that by visualizing in 3D (i.e. a higher dimension). However, the math does not require it, just our perceptions to understand what we see–Pacman need not exist in a higher dimension.


  • Cooperative Cryptonetwork

    Applying crypto/blockchain concepts (e.g. proof of work, aligning incentives) to building large scale co-ops that require cooperation and governance. Famous examples of co-ops include Visa and REI. Visa for example needed a co-op model to incentivize banks to join the network which onlocked more network effects (more consumers could use credit cards in more places).

    In an essay from A16Z on cryptonetworks, they argue that co-op models built on top of the properties of cryptocurrency would allow these kinds of models to scale even further. However, it’s hand-wavy about how exactly this solves the social and economic alignment required by cooperative models.

    See also:


  • Load-Bearing Quirks

    In software development, code describes what the program does, but is not helpful when we need to understand what it is intended to do or why. This leads to important quirks that is load-bearing–changing it or cleaning it up actually breaks it unbeknownst to the engineer.

    To better understand intent we need comments, documentation, or a formal spec. Describing code that has no intent is also important.

    For example, an algorithm implemented using a BTreeSet as a backing data structure, could that be changed to something else? Is there any intent here since BTreeSet preserves insertion ordered and hashable (in Rust)?

    See also:

    • This tweet from mcclure111 that talks about having to uncover authorial intent many years after some code is written and the concept of load-bearing quirks.

  • Emotional Thermometer

    A framework to improve mindfulness of anxiety and emotional distress. You can use this mapping to plan what to do when you find yourself at various points in the scale.

    Anchor points:

    • 100 (extreme distress and panic, maximum anxiety, fear, anger)
    • 75 (high distress and/or bodily distress)
    • 50 (mild to moderate stress, distinct unpleasant sensations)
    • 25 (slightly upset, annoyed, or bothered beginning bodily tension)
    • 0 (complete peace and serenity)

  • Sushi Train

    Visualize your thoughts as a sushi train that constantly revolves around you with different dishes that represent the flavor of thought. Some are pleasant thoughts you take off the belt for awhile and then place back on it. Some are unpleasant thoughts that we hold on to and it’s difficult to put back.

    A useful model for mindfulness, we can observe our thoughts objectively as a sushi train constantly revolving. We can choose to put a thought down and pick up another if we can observe our thoughts.


  • Self-Compassion

    People tend to be more judgmental and harsher on themselves than on others. Applying compassion that you would have for a close friend, but to yourself is a powerful coping and support mechanism.

    Steps to self compassion:

    1. Acknowledging this is a moment of suffering, being mindful and not resisting what you are feeling.
    2. I am not alone, it’s natural to feel this way.
    3. What do I need right now? What would I say to a friend in this situation?

    See also:


  • Mindfulness

    The practice of bringing your attention to the present moment without judgment. This is particularly useful to treat anxiety where your mind is fixated on the future.

    Mindfulness is non-secular, an independently useful practice that can be extracted from spiritual or religious practice. Mixing the two is problematic because religion calls for little evidence whereas the practice of mindfulness as psychological treatment does.

    See also:

    • Sam Harris' essay on Killing the Buddha where he argues that mindfulness should be separated from Buddhism if we want to further the science and the practice.

  • Personal Note Taking Practice

    Knowledge work processes should be accretive (rather than ephemeral or ad-hoc). For note-taking, adding new notes should make other notes more useful and the accumulated knowledge should lead to new connections and thereby new ideas.

    I use a Zettelkasten-like system (org-roam) where notes are short and atomic, densely linked bi-directionally, and reviewed often to add more connections. Over time I hope this becomes a useful thought partner, helping me make new connections and formulate useful ideas.

    Tags provide light organization without imposing a strict hierarchy e.g. index for how I might want to retrieve this note in the future not a taxonomy.

    Certain tags have special meaning:

    • (draft) indicates a note that isn’t complete and not ready to publish
    • (private) should not be included in public notes
    • (journal) my reactions and feelings which can provide more context to how I was thinking about certain notes

    Notes are published (except when tagged as private or draft) to notes.alexkehayias.com. Care is put into making the notes pleasing for me to read and invite exploration of links and backlinks. It also adds a useful friction that forces me to understand what I’m writing better since it’s public.

    As an interesting side-effect of using git for version control, when I publish new notes the git diff gives a good indication of how many new connections I’ve made with each new note added (untracked files to unstaged modified files).


  • Knowledge Work Should Be Accretive

    Most knowledge work is ephemeral–we write documents, emails, code and then it’s done. The ways in which we work don’t tend to compound or accumulate over time. This makes knowledge work lossy. A good example of this is note taking–we tend to never look at notes once they are written.

    Ideally, all work we do builds on and increases our capacity for new ideas. Systems we put in place should allow us to add to and extrapolate from our knowledge.

    See also:

    • Andy Matuschak’s note where this idea came from (actually I heard it at talk he gave)
    • Zettelkasten is a note taking system which attempts to build up a ‘conversation partner’ by accumulating notes with a dense set of connections thus accreting knowledge systematically.