1. Use distanced self-talk.

One way to create distance when you’re

experiencing chatter involves language. When you’re trying to
work through a difficult experience, use your name and the
second-person “you” to refer to yourself. Doing so is linked with
less activation in brain networks associated with rumination and
leads to improved performance under stress, wiser thinking, and
less negative emotion.
2. Imagine advising a friend. Another way to think about your
experience from a distanced perspective is to imagine what you
would say to a friend experiencing the same problem as you.
Think about the advice you’d give that person, and then apply it
to yourself.
3. Broaden your perspective. Chatter involves narrowly focusing on the
problems we’re experiencing. A natural antidote to this involves
broadening our perspective. To do this, think about how the
experience you’re worrying about compares with other adverse
events you (or others) have endured, how it fits into the broader
scheme of your life and the world, and/or how other people you
admire would respond to the same situation.
4. Reframe your experience as a challenge. A theme of this book is that you
possess the ability to change the way you think about your
experiences. Chatter is often triggered when we interpret a
situation as a threat—something we can’t manage. To aid your
inner voice, reinterpret the situation as a challenge that you can
handle, for example, by reminding yourself of how you’ve
succeeded in similar situations in the past, or by using distanced
self-talk.
5. Reinterpret your body’s chatter response. The bodily symptoms of stress
(for example, an upset stomach before, say, a date or
presentation) are often themselves stressful (for instance, chatter
causes your stomach to grumble, which perpetuates your chatter,
which leads your stomach to continue to grumble). When this
happens, remind yourself that your bodily response to stress is
an adaptive evolutionary reaction that improves performance
under high-stress conditions. In other words, tell yourself that
your sudden rapid breathing, pounding heartbeat, and sweaty
palms are there not to sabotage you but to help you respond to a
challenge.
6. Normalize your experience. Knowing that you are not alone in your
experience can be a potent way of quelling chatter. There’s a
linguistic tool for helping people do this: Use the word “you” to
refer to people in general when you think and talk about negative
experiences. Doing so helps people reflect on their experiences
from a healthy distance and makes it clear that what happened is
not unique to them but characteristic of human experience in
general.
7. Engage in mental time travel. Another way to gain distance and
broaden your perspective is to think about how you’ll feel a
month, a year, or even longer from now. Remind yourself that
you’ll look back on whatever is upsetting you in the future and
it’ll seem much less upsetting. Doing so highlights the
impermanence of your current emotional state.
8. Change the view. As you think about a negative experience, visualize
the event in your mind from the perspective of a fly on the wall
peering down on the scene. Try to understand why your “distant
self” is feeling the way it is. Adopting this perspective leads
people to focus less on the emotional features of their experience
and more on reinterpreting the event in ways that promote
insight and closure. You can also gain distance through visual
imagery by imagining moving away from the upsetting scene in
your mind’s eye, like a camera panning out until the scene
shrinks to the size of a postage stamp.
9. Write expressively. Write about your deepest thoughts and feelings
surrounding your negative experience for fifteen to twenty
minutes a day for one to three consecutive days. Really let
yourself go as you jot down your stream of thoughts; don’t worry
about grammar or spelling. Focusing on your experience from
the perspective of a narrator provides you with distance from the
experience, which helps you make sense of what you felt in ways
that improve how you feel over time.
10. Adopt the perspective of a neutral third party. If you find yourself
experiencing chatter over a negative interaction you’ve had with
another person or group of people, assume the perspective of a
neutral, third-party observer who is motivated to find the best
outcome for all parties involved. Doing so reduces negative
emotions, quiets an agitated inner voice, and enhances the
quality of the relationships we share with the people we’ve had
negative interactions with, including our romantic partners.
11. Clutch a lucky charm or embrace a superstition. Simply believing that an
object or superstitious behavior will help relieve your chatter
often has precisely that effect by harnessing the brain’s power of
expectation. Importantly, you don’t have to believe in
supernatural forces to benefit from these actions. Simply
understanding how they harness the power of the brain to heal is
sufficient.
12. Perform a ritual. Performing a ritual—a fixed sequence of behaviors
that is infused with meaning—provides people with a sense of
order and control that can be helpful when they’re experiencing
chatter. Although many of the rituals we engage in (for example,
silent prayer, meditation) are passed down to us from our
families and cultures, performing rituals that you create can
likewise be effective for quieting chatter.
Tools That Involve Other People
When we think about the role that other people in our lives play in
helping us manage our inner voice, there are two issues to consider.
First, how can we provide chatter support for others? And second,
how can we receive chatter support ourselves?
Tools for Providing Chatter Support
1. Address people’s emotional and cognitive needs. When people come to
others for help with their chatter, they generally have two needs
they’re trying to fulfill: They’re searching for care and support,
on the one hand (emotional needs), and concrete advice about
how to move forward and gain closure, on the other (cognitive
needs). Addressing both of these needs is vital to your ability to
calm other people’s chatter. Concretely, this involves not only
empathically validating what people are going through but also
broadening their perspective, providing hope, and normalizing
their experience. This can be done in person, or via texting, social
media, and other forms of digital communication.
2. Provide invisible support. Offering advice about how to reduce
chatter can backfire when people don’t ask for help; it threatens
people’s sense of self-efficacy and autonomy. But that doesn’t
mean there aren’t still ways of helping others when they
experience chatter and don’t ask for assistance. In such
situations providing support invisibly, without people being
aware you’re helping them, is useful. There are many ways to do
this. One approach involves covertly providing practical support,
like cleaning up the house without being asked. Another involves
helping broaden people’s perspectives indirectly by, for example,
talking in general terms about others who have dealt with similar
experiences (for example, “It’s amazing how stressful everyone
finds parenthood”) or by soliciting advice from someone else but
without signaling that the questions are meant to help the person
in need. For example, if my colleague was struggling to connect
with their graduate student and we found ourselves at a function
with other advisers, I might casually ask a group whether they’ve
experienced trouble connecting with their students and, if so,
how they managed the situation.
3. Tell your kids to pretend they’re a superhero. This strategy, popularized in
the media as “the Batman effect,” is a distancing strategy that is
particularly useful for children grappling with intense emotions.
Ask them to pretend they’re a superhero or cartoon character
they admire, and then nudge them to refer to themselves using
that character’s name when they’re confronting a difficult
situation. Doing so helps them distance.
4. Touch affectionately (but respectfully). Feeling the warm embrace of a
person we love, whether that be holding someone’s hand or
sharing a hug, reminds us at the conscious level that we have
supportive people in our lives whom we can lean on—a chatterrelieving
psychological reframe. Affectionate touch also
unconsciously triggers the release of endorphins and other
chemicals in the brain such as oxytocin that reduce stress. Of
course, for affectionate touch to be effective it has to be welcome.
5. Be someone else’s placebo. Other people can powerfully influence our
beliefs, including our expectations about how effectively we can
deal with chatter and how long it will last. You can utilize this
interpersonal healing pathway by providing the people you’re
advising with an optimistic outlook that their conditions will
improve, which changes their expectations for how their chatter
will progress.
Tools for Receiving Chatter Support
1. Build a board of advisers. Finding the right people to talk to, those
who are skilled at satisfying both your emotional and your
cognitive needs, is the first step to leveraging the power of
others. Depending on the domain in which you’re experiencing
chatter, different people will be uniquely equipped to do this.
While a colleague may be skilled at advising you on work
problems, your partner may be better suited to advising you on
interpersonal dilemmas. The more people you have to turn to for
chatter support in any particular domain, the better. So build a
diverse board of chatter advisers, a group of confidants you can
turn to for support in the different areas of your life in which you
are likely to find your inner voice running amok.
2. Seek out physical contact. You don’t have to wait for someone to give
you affectionate touch or supportive physical contact. Knowing
about the benefits they provide, you can seek them out yourself,
by asking trusted people in your life for a hug or a simple hand
squeeze. Moreover, you need not even touch another human
being to reap these benefits. Embracing a comforting inanimate
object, like a teddy bear or security blanket, is helpful too.
3. Look at a photo of a loved one. Thinking about others who care about
us reminds us that there are people we can turn to for support
during times of emotional distress. This is why looking at photos
of loved ones can soothe our inner voice when we find ourselves
consumed with chatter.
4. Perform a ritual with others. Although many rituals can be performed
alone, there is often added benefit that comes from performing a
ritual in the presence of others (for example, communal
meditation or prayer, a team’s pregame routine, or even just
toasting drinks with friends the same way each time by always
saying the same words). Doing so additionally provides people
with a sense of support and self-transcendence that reduces
feelings of loneliness.
5. Minimize passive social media usage. Voyeuristically scrolling through
the curated news feeds of others on Facebook, Instagram, and
other social media platforms can trigger self-defeating, envyinducing
thought spirals. One way to mitigate this outcome is to
curb your passive social media usage. Use these technologies
actively instead to connect with others at opportune times.
6. Use social media to gain support. Although social media can instigate
chatter, it also provides you with an unprecedented opportunity
to broaden the size and reach of your chatter-support network. If
you use this medium to seek support, however, be cautious about
impulsively sharing your negative thoughts. Doing so runs the
risk of sharing things that you may later regret and that may
upset others.
Tools That Involve the Environment
1. Create order in your environment. When we experience chatter, we
often feel as if we are losing control. Our thought spirals control
us rather than the other way around. When this happens, you
can boost your sense of control by imposing order on your
surroundings. Organizing your environment can take many
forms. Tidying up your work or home spaces, making a list, and
arranging the different objects that surround you are all common
examples. Find your own way of organizing your space to help
provide you with a sense of mental order.
2. Increase your exposure to green spaces. Spending time in green spaces
helps replenish the brain’s limited attentional reserves, which are
useful for combating chatter. Go for a walk in a tree-lined street
or park when you’re experiencing chatter. If that’s not possible,
watch a film clip of nature on your computer, stare at a
photograph of a green scene, or even listen to a sound machine
that conveys natural sounds. You can surround the spaces in
which you live and work with greenery to create environments
that are a boon to the inner voice.
3. Seek out awe-inspiring experiences. Feeling awe allows us to transcend
our current concerns in ways that put our problems in
perspective. Of course, the experiences that provide people with
awe vary. For some it is exposure to a breathtaking vista. For
someone else it’s the memory of a child accomplishing an
amazing feat. For others it may be staring at a remarkable piece
of art. Find what instills a sense of awe within you, and then seek
to cultivate that emotion when you find your internal dialogue
spiraling. You can also think about creating spaces around you
that elicit feelings of awe each time you glance at them
Here are the algorithms to live by, with my notes from each section (Warning: This is long. Seriously long)

1. Optimal Stopping - Evaluating choices - Stop at 37% (in no-information games)

• The crucial dilemma is not which option to pick but how many to even consider
• Ordinal numbers and cardinal numbers - Relative ranking between options vs ratings on a general scale
• Two ways of failing - stopping early and not discovering the best option or stopping late and passing by the
best option

• “Look-then-leap” is precisely what it is - Look at 37% of the options (or 37% of total time available) to form
an intuition and gain information and then leap.

• Variants - If there is 50/50 chance of proposal being accepted (option has agency), then stop at 25% and if
immediate proposals are a sure thing but belated ones are 50/50, be noncommittal till 61%.

• If we know exactly what we are looking for - then its a full-information game - stop when you find the first
option that matches

• Corollary - In the case of slim pickings, lower your standards and when there is more fish in the sea, raise
them

• Turning no-information games into full-information ones - Finding where your option stands relative to the
total population at large changes the look-then-leap to threshold rule (Higher odds of finding best option)

• Where to Park - Pass up all vacant spots available before a certain distance to the destination and then take the
first one that’s available after - choose distance based on occupancy rate (Rational brain does this amazingly
intuitively)

2. Explore/Exploit - Latest vs Greatest

• Life is a balance between novelty and tradition

• Explore vs Exploit depends on how much time you have in the game - In the game life, it depends on your age

• We are more likely to try a new restaurant when we move to a city than when we are leaving it
• Explore when you have the time to use the resulting knowledge, exploit when you are ready to cash in. The
interval makes the strategy

• Movie sequels are “exploit” strategies. It is short-termist and maybe studios think they are in the end of the
interval (their imminent demise)

• Win-stay, Lose-shift strategy (Multi-armed bandit problems) - Stay with a winning option long as it pays and
shift when it doesn’t. Though Win-Stay is generally a very optimal strategy for most games, Lost-Shift may not
be - It could be a rash move. Good options shouldn’t be penalized for being imperfect (Bit of Bayes will help?)

• Gittins index for multi-armed bandit problems (Devised for R&D in pharma for drug trials) - Uses geometric
discounting. Play the arm with the highest index.

• An untested rookie is worth more (early in the season) than a veteran of seemingly equal ability, because we
know less about him - Taking future into account drives us more towards novelty

• Regret and Optimism - To try and fail is at least to learn, to fail to try is to suffer the inestimable loss of what
might have been (Chester Bernard)

• Regret will never stop increasing, even if you pick the best possible strategy - because even the best possible
strategy isnt perfect every time. Only rate of growth of regret goes down over time with the best strategy

• Minimum possible regret is regret that increases at a logarithmic rate (As many mistakes in the first year as the
next 9 in the decade which would be the same as what you would make in the next 90)

• Upper confidence bound algo (Optimism in the face of uncertainty) - Doesn’t care about past payoffs as much
and looks for what could perform better in the future (restaurant with a single mediocre review has a higher
potential for greatness than one with 100s of mediocre reviews)

• Assume the best about new people, things or options in the absence of evidence to the contrary - in the long
run, optimism is the best prevention for regret

• Hippocratic oath in medicine - Do no harm (fundamental of medical ethics)

• Adaptive trials as an alternative to current clinical trials - chance of using a given treatment is increased with
each win and decreased with loss (mod of Win-stay, lose-shift)

• In general people tend to over-explore (could be because the world is a restless bandit and things change all
the time - that bad restaurant doesn’t have to remain bad - maybe its better now)

• Having instincts tuned by evolution for a world in constant flux isn’t necessarily helpful in an era of industrial
standardization

• Childhood is designed in such a way that exploration can be done without concern of payoffs

• Our intuition about rationality are often informed by exploitation than exploration - if you treat every decision
as your last - only exploitation makes sense

• Elderly have few social connections by choice focusing on meaningful connections (exploit)

3. Sorting
• The tabulation of the 1880 census took 8 years, barely finishing by the time the 1890 census (US) began

• First stored program ever written was for sorting (IBM)

• Finding largest, smallest, rarest, indexing, finding dupes - they all start with a sort

• Sorting has negative operating leverage (more to sort, worse the cost) Corollary: Do laundry more often if
sorting socks is becoming a pain

• Worst case analysis lets us make guarantees on computing times - (Big-O notation in comp-sci always deals
with worst case times - as size of problem increases, how does the running time change?)

O(1) = constant time - Computing time remains same irrespective of size

O(n) = linear time - Twice the guests in a dinner table, twice the dish takes to come around the dinner table
O(n^2) = quadratic time (If each guest hugged each other)
O(2^n) = exponential time - Where each guest doubles your work
O(n!) = factorial time - Like shuffling a deck and hoping it were sorted (Run through all combinations)

• Bubble sort - Sorting a book shelf pass after pass in quadratic time

• Insertion sort - Pull off all the books and put them back one at a time at the right place (Again quadratic time)

• Even checking if a bookshelf is sorted is in linear time involving a full scan of the shelf (So sorting in constant
time O(1) is out of the question). Ideal solution lies between Linear and Quadratic times (Mergesort)

• Mergesort has a linearithmic time O(n log n) - Sorting two already sorted stacks is a lot easier than one big
stack. Excellent for parallelizing - Call a bunch of your friends over and give them all a pile to sort and then
merge them all in the end (Difference in times in staggering - like 29 passes vs a few million for census level
items)

• Bucket sort - Sorting by categories - This can be done in almost linear time (Though will depend on number of
categories so O(mxn). Libraries do this all the time by genre

• Sorting something you will never search is wasteful. Searching something unsorted is merely inefficient (Okay
if you do it infrequently). Err on the side of messiness (I do, with my bookshelf which is merely bucketed into
read and unread. Besides, I enjoy the searching :-))

• We search with our quick eyes and sort with our slow hands - important thing to consider as well. Sometimes
mess is the optimal choice

• Round-robin and ladder tournaments have quadratic complexity and can be quickly overwhelming when
number of teams increases

• Single-elimination tournament can only decide Gold but not silver or bronze accurately (Still we use it in
Olympics and in smartphone camera shootouts)

• Games with noise like soccer (fluke or luck plays a part) gain from inefficiency of a Bubble sort. (IPL uses a
comparison counting sort)

• One of the important skills as a poker player is to be able to evaluate how good you are

• Blood sort - Pecking Order/Dominance hierarchies - Establishing a pecking order avoids a lot of confrontation
and bloodshed.

• Pecking order is easier to establish in a herd/flock/pack if group size is small. Ethical raising of livestock must
take this into account (keep group size small)

• When every knows their position in a pecking order, no games ensue (be it poker cash games or fight among
monkeys)

• Race is fundamentally different from a fight - Marathon is cardinal than ordinal - naturally orders the set by
finish time and doesn’t need pairwise comparisons

• Things like national GDP establish a dominance hierarchy among nations and avoids conflicts to some extent

4. Caching

• What to do when cache gets full is decided by an eviction policy (or replacement policy). Idea is to minimize
the number of times you can’t find what you are looking for in the cache

• The idea is to evict whichever item it is we will need the longest from now (almost like clairvoyance)

• Random eviction - As the name suggests - nuke whichever and make space, even if its your favorite shirt in
the daily rack

• FIFO - First In, First Out - Get rid of whatever has been sitting on longest

• LRU - Least Recently Used (Gold standard) - Get rid of whichever has not been used recently. Considering
frequency of use can also help improve along with recency

• Nearest thing to clairvoyance is to assume history repeats itself - backward

• People watch movies set close to where they live (So Netflix caches these movies in the closest CDN)
• Brain forgets to reduce cognitive load. Unlike eviction or replacement, brain loses references (Ebbighaus
foregetting curve)

• Human society functions like human beings in forgetting (similar curve for newspaper headlines) -
Availability bias is caused by this

• Cognitive decline from ageing could be caused due to sheer amount of information that must be processed

5. Scheduling - First things first

• Minimizing maximum lateness - Pick the task with the earliest Due date first (or serve customers who arrived
first)

• Minimizing the number of foods that spoil in a fridge - Eat food by earliest expiry date but toss the largest
item if it means consuming more items (Moore’s Algorithm)

• Minimizing length of TO DO list - start with the items with shortest processing time

• Only prioritize a task twice as long only if its twice as important (form some rules of thumb)

• A man with one watch knows what time it is, a man with two is never sure

• Give a system an overwhelming number of trivial things to do and the important things get lost (denial-of-
service)

• Staying focused on getting the weighty important things done - can also be an non-optimal approach (Mars
Pathfinder issue due to priority inversion)

• Precedence constraint (fancy name in scheduling theory for task dependency)

• Problems without an efficient solution (Intractable problems)

• Straightening out a to-do list can become an item on the to-do list when the system doing the scheduling is
same as the one being scheduled

• Price paid for switching tasks - context switch (that’s why 16 hour days are more productive than 8 hour days
sometimes in writing and programming)

• Everyone you interrupt more than a few times in an hour has a danger of not doing anything at all

• Thrashing - when all you do is context switch without doing anything productive

• When a juggler takes one more ball, he doesn’t lose just that ball, he loses all

• Interrupt coalescing - responsiveness (how quickly can you respond) and throughput (how much work can you
get done) are always at odds - better to have someone answer the phone (responsive) while you get work done
(throughput)

• Donald Knuth - patron saint of minimal context switching (mails replied once in 3 months)

6. Bayes Rule - Making decisions from small data

• (w+1)/(n+2) - Laplace’s law for estimating probabilities where w is wins and n is attempts (When you have
one win from one attempt, you have 67% chance, which is reasonable “optimistic” estimate than over-confident
100%)

• It was Laplace who did all the heavy-lifting though its called Bayes rule

• How long will something last - As long as it has already lasted is a useful rule of thumb (Gott’s Copernican
principle). This sounds very much like Lindy effect

• Copernican principle is just Bayes rule with an uninformative prior

• The richer the prior information we bring into Bayes rule, the better our predictions
• Be wary of what distributions your real world priors draw from (power-law vs gaussian). Populations and
incomes follow power-law than normal

• Copernican principle is Bayes rule for priors that are power-law distributed. (Multiplicative based on power-
law exponent - hence 2x for Lindy and 1.4x for movie collections based on collections so far)

• Things that are neither more nor less likely to end because they have gone on for awhile (Erlang distribution) -
Predictions based on additive rule.

• Small data is big data in disguise - The reason why we are able to predict well from small data is because our
priors are so rich

• Good prediction require good priors

• Our judgements betray our expectations and our expectations betray our experience

• ability to resist temptation maybe a matter of expectations than willpower (if you know how long you have to
wait, you develop the will to wait)

• Priors are formed by experience but when a species gains language and the ability to speak, priors are formed
not just by personal experience but by shared experience which may have a skew to special/interesting things
affecting priors

• News reports interesting/special things which may be infrequent - don’t let infrequent things reported
frequently affect your priors (protect your priors for they are what you base your decisions on)

7. Overfitting - When to think less

• How hard to think, how many factors to consider? (There’s wisdom in thinking less)

• Better fit doesn’t mean better prediction

• Overfitting taste - when food taste excellent but nutritionally poor (Taste was a fit for nutrition as required in
the past)

• Company will build whatever it is the CEO decides to measure (Sam Altman)

• Ruthless and clever optimization of the wrong thing (Goodhart’s law, though the book doesn’t mention it by
name)

• Training scars (A cop habitually handed the pistol to the assailant after grabbing it from him as he had done so
hundreds of time in training)

• Cross-validation detects overfitting by seeing how well a class generalizes what it learnt (to see if it was only
taught to the test)

• Regularization - placing constraints that penalize models for their complexity

• Lasso used frequently in ML drives a lot of weights down to zero

• Language forms a natural Lasso - Convey what you intend fast or you lose the audience’s attention

• Less information, computation and time can improve accuracy (Elevator pitches and investment thesis in few
lines)

• Nudge a model towards simplicity by controlling how quickly it adapts to new data

• single most important factor than multi-factors lead to better predictions in some cases

• Early Stopping provides the foundation for a reasoned argument against reasoning (thinking person’s case
against thought)

• If you have high uncertainty and limited data - Stop Early

• Further ahead you are in a brainstorming session, thinner the pen’s stroke size should be (simplification by
stroke size - can’t get into minor details when coming up with broad outlines due to pen size)
• You can also regularize to the page (what doesn’t make the page isn’t important)

8. Relaxation

• Traveling salesman problem is O(n!) problem. Its not that a computer can’t find the shortest route but that as
number of towns increase, the problem has n! solutions and finding the best out of it is computationally hard

• Defining difficulty - Any algo that runs in polynomial time O(n^2) or O(n^3) is considered efficient or in
general O(n^m) even. O(n!) is considered intractable

• Relax the traveling salesman problem by allowing him to retract to the same town or visit same town multiple
times and form - this shortest route produces the minimum spanning tree

• Constraint relaxation like above lets us solve an easier version of a complex problem making the intractable,
tractable and making it a starting point

• If you are willing to accept compromises, even the hairiest problems can be tamed

• Lagrangian relaxation - “Do, it or else!” problems replied with “Do it or else what?” - coloring outside the
lines at a cost to make the intractable tractable. (Converting impossibilities into penalties and teaching the art of
bending/breaking the rules and living with the consequences)

9. Randomness

• Choosing the random option feels like a cop out but its far from it

• You need to know when to rely on chance, in what way and to what extent

• Complex quantities can be estimated by sampling (value of pi can be estimated by dropping a needle on paper)

• Test of first-rate intelligence is the ability to hold two conflicting thoughts and still be able to function (F.
Scott Fitzgerald)

• Monte Carlo simulations - Replacing exhaustive probability calculations with sample simulations

• Sieve of Erastothenes - One of the first algorithms for finding prime numbers in ancient Greece

• It is much easier to multiply two primes than factor them back out (especially for very large ones) - forms the
basis of cryptography

• Testing for prime - primality - Miller’s approach is better than sieve of Erastothenes but it involves probability
than certainty

• Along with Time and Space tradeoff, in recent times, the third dimension of error probability is added (You
can get an answer fast but there is small chance of error)

• Bloom filters works like Miller-Rabin test for primality (Used for detecting malicious sites)

• Greedy/myopic algo - One that takes the best thing available, every step of the way (in the context of Gradient
Descent as in ML or Hill-Climbing as used in the book). To get from a local minima to a global minima, you
have worsen your solution a bit

• Use a little bit of randomness everytime you make a decision (Metropolis algorithm)

• Simulated Annealing (heating and then slowly cooling) from metallurgy is used in ML for optimization

• Even if you are in the habit of acting on bad ideas, you should always act on the good ones (Hill Climbing
algo)

• Your likelihood of following a bad idea should be inversely proportional to how bad the idea is

• Front-load randomness, rapidly cooling out of it, using lesser and lesser randomness as time goes by, lingering
longest as you approach freezing (Temper yourself, literally)

10. Networking
• Circuit switching - constant bandwidth, always on - made sense for human communication. But having a
dedicated connection to something that’s never talking but when it does wants an immediate line gave rise to
packet switching (like postcards moving at the speed of light)

• Packet switching - reliability increases exponentially with network size unlike circuit switching where calls
fail if any link gets disrupted

• Byzantine generals problem (confirmation of receipt of message requires another message causing endless
recursion)

• 10% of peak hour upstream traffic is Netflix ACKs

• Exponential backoff algorithm - The algorithm of forgiveness. Ubiquitous for reconnecting - Max delay length
is exponential (Its a random number of seconds delay under max though to attempt reconnect)

• We must replace three strikes and you are out (finite forgiveness against infractions) with finite patience and
infinite mercy (Simply take longer intervals to try again but never give up)

• Congestion/Flow control - AIMD - Additive Increase/Multiplicate Decrease (Another fantastic algorithm) -

TCP sawtooth pattern is caused by this. I think this is a fantastic algorithm for allocation to bets/trades - keep
adding by 1 until trouble ensues but when it does, cut back by half

• Control without hierarchy - More ants leave the nest, the more successful the foraging but unsuccessful
returnees result in diminishment

• Every employee tends to rise to his level of incompetence (Peter Principle) - Every employee gets promoted
but stagnates in a position where he doesn’t do well. So over time an organization is filled with people doing
their worst in the post they are in. (Demotions help)

• With poor feedback from the listener, the story falls apart

• Photons that miss the retina aren’t queued for later viewing - In real life, packet loss is total

• It used to be that people knocked on your door and went home if there was no answer, now they wait in queue
(on Email)

• Circuit switching to packet switching has happened to society - Instead of dedicated lines, we send messages
and instead of reject, we defer

• Tail drop (Can’t accept any more messages)

• Companies that advertise fast internet connections are actually advertising higher bandwidth than lower
latency

11. Game Theory

• Algorithmic game theory - cross-pollination between game theory and computer science

• Successful investing is anticipating the anticipation of others (Keynes) or what Average opinion expects the
average opinion to be

• Halting problem - when a machine or mind tries to simulate something as complex as itself, it finds its
resources maxed out, by definition

• You really want to play only one level above the opponent - if not you are going to think they have
information they don’t really possess

• Equilibrium - Content with my strategy given yours, and you are content with yours given mine

• Every two player game has at least one equilibrium (John Nash)

• Predictive abilities of Nash equilibria only matter if they can be found

• If your laptop can’t find it, neither can the market

• A low price of anarchy means that the system is about as good on its own than being carefully managed
• Tragedy of the commons - two player prisoner’s dilemma extended to many players - we can easily end up in
a terrible equilibrium with a clean conscience

• If the rules of the game force bad strategies, we shouldn’t change the strategies, we should change the game
(reverse game theory or mechanism design - what rules will give us behavior we want to see?)

• If a tree grows taller to get more sunlight and the rest of the trees do too, to the same level - the canopy finally
gets the same as it did before - except now it supports the trunk at a higher cost (Dawkins)

• Morality is herd instinct in the individual (Nietzsche). Emotion is mechanism design in the species

• Sealed-bid first price auction, Dutch auction (Descending) and English auction (Ascending). In a Dutch
auction, its the absence of bid that reveals information

• Information cascade (or infinite misinformation) - When players take others’ actions for beliefs and act
accordingly and that reinforces someone else’s belief. Consensus unglues from reality (Happens all the time in
the stock market)

• Be wary of cases where public information exceeds private information (to avoid being trapped in information
cascades) and also be wary of situations where you know more about what people are doing than why they are
doing it and hesitant to overrule your own doubts

• Vickrey auction (second-best bid) - winner pays second best bid - participants are incentivized to be honest

• Any game that requires strategically masking the truth can be transformed into one that requires nothing but
simple honesty

• Hell is other people (Sartre) - (Not because of maliciousness but because of the way they affect our beliefs)

• Popularity is complicated, intractable, a recursion in a hall of mirrors, but beauty in the eye of the beholder, is
not. Adopting a strategy that doesnt require anticipating, predicting other’s tactics is one way to cut the Gordian
knot of recursion

• Seek out games where honesty is the dominant strategy. Then just be yourself

If you have made it this far, this book is for you. 11/10