Interests (Data, AI, Crypto, EV, Startups, Africa, Asia)
4982 words

How to run a vague biz dev meeting

THREAD: How to run a vague biz dev meeting. Some tips I've learned and observed from some of the best (including one favorite from Sheryl Sandberg) over the years.


If you're in any medium/big company (or even a small one), you'll eventually be pulled into a vague partner meeting. Where some partner company will bring their execs/CEO over to visit (or you go visit them). These can be confusing if you work in product/eng but they needn't be!

What makes them confusing: there are multiple stakeholders on all sides. Sales, BD, your CEO perhaps, multiple teams. Same on "their side". And it isn't clear what needs to be accomplished by either side.

  1. Take the meeting! Especially if you're young in your career, say yes to these. You learn a ton about other industries and more importantly, the ppl who run them.

This is part of why I recommend all PMs spend some time in ad products - you get exposed to multiple industries.

  1. How do you get into one? Easy, ask! Every sales person LOVES having someone from tech/product in a meeting - it shows you're serious. If you work in ad products, ask to get in some sales meetings today (I assure you, you'll soon be very popular with your sales team)

  2. Next step: prepare. Google who they are, the people who are attending. Know their competition, their latest numbers, read their interviews, their career trajectories. You never know when this will save you or change a meeting's tone (it has frequently for me).

  3. The pre-meeting: Get together with the "home team" (your company's folks attending the meeting). Find out who's running the meeting (usually the most senior person). This is key - this person kicks it off, keeps things on track and brings it home.

  4. Go over what you want to accomplish from this meeting. Maybe get a test kicked off? Get their CEO to nudge his/her CMO to take you folks more seriously? Get a lay of the land of the people and the politics (a good home team prep goes deep on this).

  5. Game-plan who talks and covers what. For example: "home team" exec kicks off, throws to lead account manager to pitch what your company's big message is, throws to you as product manager to talk about some cool launch, lead into some test/partnership opportunities.

  6. Maybe have a few slides to help anchor and guide the discussion. Don't go overboard, I have sat in way too many of these meetings which have drowned in slideware. They're to act as a reference point, not the core of the meeting.

  7. Always start with the big picture and the basics of what your company do. Assume they know nothing about you. Even today any Twitter meeting will start with assuming they nothing about Twitter.. even though that is always hilariously not true.

  8. Customize the deck/narrative for the company visiting. Showcase past wins together, tie it in with what you know about their current challenges. A good account manager should be pre-gaming this with his/her counterpart on the other side.

  9. Have scripts/talking points prepared for the "obvious" questions that will come up (e.g. have you been in the press recently?). If you dont know what these are for your company, you aren't doing your homework.

  10. Onto the meeting itself: location is important if you can control it. If it's at home, make sure you have tested AV, have a notepad at the table, entered their names in the visitor system, etc. Get some swag/merch. The details matter.

pro-tip: customize the swag/merch. Does one of the execs have an upcoming wedding? A new baby? A baby oneside with your company logo will go a long way.

Again, details matter.

  1. The meeting starts. Always start with introductions. Always assume they know nothing so contextualize. Don't say "I'm a PM of the internal FooBar team". Say "I work on the product team that helps recommend content to users" (say).

  2. Meeting kicks off. Usually home team exec kicks off with a little big picture. Now here comes the first BIG tip.

ALWAYS ask them (usually the most senior person) what they care about, what they hope to get out of it. Get them to explain their biz to you.

I've seen so many meetings fail because this wasn't done. By just letting them talk (and not interrupting - pay close attention), you'll learn so much. What is worrying them, what are their priorities, what are their preconceived notions, etc.

  1. Use this to frame everything else. Don't be afraid to change up the agenda on the fly if it is clear the partner CEO is going to only care about the last bit of your agenda. This is their meeting. Reflect their priorities.

  2. You're now in the thick of the meeting. This is very much an art. You are part pitching what you have but also discussing, trying to understand how your pitch is landing on them.

Very key here: keep quiet! it is ok to let the room breathe and to let them talk.

  1. Especially awesome if they talk among themselves for you really get a sense of their internal thinking.

  2. Leave a lot of time for open discussion. ideally your deck should be only half the meeting (or less). It's ok to go very short, I've never seen any of these end because someone had too little slides.

  3. Now, you as young PM are called upon to speak. Newbie mistake here is to just dive into "cool feature X". Always start with big picture. Why does the world care about this? What is the industry situation on this? Why should they care about this?Contextualize. Make it relevant

This is where your homework really helps you. Meetings can go amazingly well when you can say "oh and btw, when this launches, your next movie launch X will be perfect because X fits this model perfectly" It shows you paid attention.

  1. Finally, you are about to wrap up. Here are two big tips. One from @nedsegal. Always get a commitment from them for a next step in the room. Show them a menu of options and get their CEO/SVP/whoever to say "Hey why don't we try this out?". Get a point of contact/commitment

  2. And one tip I learned from Sheryl who does this in every partner meeting. Say at the end "What are you going to say about this meeting when we leave the room?". This TOTALLY changes the tone of the meeting and acts as a meta conversation. Do this!

  3. Finally, you're done. Shake hands. Get everyone's emails or cards. And then - follow up! Add them on Linkedin, send them thank you notes.

  4. Most important: the follow up. Make sure you (or your team) follow up with them later and actually do the things you said you were going to do. Meetings are momentum creators - seize the momentum, don't let it fizzle out.

And that's it! I used to hate these but now see them as fun ways to learn other businesses. Go attend a few and try these out and I bet you'll get a mini-MBA and have fun along the way. Fin!

By Sriram Krishnan

Product @Twitter. Previously FB/Snap/MSFT.



People are the foundation of any company’s success.

The primary job of each manager is to help people be more effective in their job and to grow and develop.

We have great people who want to do well, are capable of doing great things, and come to work fired up to do them. Great people flourish in an environment that liberates and amplifies that energy. Managers create this environment through support, respect, and trust.

Support means giving people the tools, information, training, and coaching they need to succeed. It means continuous effort to develop people’s skills.

Great managers help people excel and grow.
Respect means understanding people’s unique career goals and being sensitive to their life choices. It means helping people achieve these career goals in a way that’s consistent with the needs of the company.

Trust means freeing people to do their jobs and to make decisions. It means knowing people want to do well and believing that they will”

From: Trillion Dollar Coach. by Eric Schmidt

Objective: Not everything is possible

Not everything is possible.

You can’t jump twelve feet in the air. And while many children dream of becoming an astronaut, only a few will actually pilot a spacecraft. Impossibility isn’t a popular topic in our culture, where we say that you can be anything that you want to be, do anything if you put your mind to it. But in this chapter we’ll stare impossibility straight in the eye—though we won’t sacrifice our optimism in exchange. Instead, we’ll discover a source of optimism that embraces the uncertainty of the far future rather than fearing or denying it. This journey begins by considering both the power and limitations of searching for novelty.
It might seem like the lesson of novelty search is that finding the objective is often easier when not looking for it. You might even solve more problems by not worrying about them instead of actually trying to solve them. So if you look at it that way, novelty search might seem to be just a new tool that can be added to the existing toolbox for achieving objectives.

And it’s true that some of the computer experiments with novelty search do actually produce this kind of outcome: The maze-navigating robot learns to solve mazes best when it’s not trying to solve them; the biped learns to walk farthest when it’s not trying to walk.

But we need to be careful how we interpret these results—what they seem to say on the surface is possibly misleading. We should be especially cautious when scientific results introduce something strange and new. Just as a car is not merely a new kind of horse, novelty search is not just a new or better way of reaching an objective. While the evidence clearly shows that sometimes you can do better without a specific objective, a deeper point is that of course novelty search will not always find what you want. Surely we can concoct problems where wandering without a care in the space of all possibilities fails to stumble upon a particular objective.

And even if your own personal journey doesn’t end where you had hoped, the idea of the solitary inventor striving relentlessly towards her inevitable objective was always a myth. Rather, it’s the combination of many minds with many different interests that ultimately plunders the search space in the long run, not any individual objective or person. We can be confident that the Butterflies and Cars of our future will be found not because someone is looking for them, but because everyone is looking for everything. The future will arrive off schedule, but it will arrive nonetheless.

This insight may seem sad, that we’re left with no sure compass, that all our efforts to create certainty and to search with purpose may be futile. But our disappointment may be misplaced. Perhaps search isn’t really about objectives but about something much bigger. In that case, abandoning the false compass can be liberating, opening up a new frontier. Novelty search shows that it’s possible to capture the process of open-ended innovation and divergent thinking even within a computer.

So it can’t be a mystical form of voodoo but rather a principled and logical process that we can understand and even capture. If discovery without explicit objectives is the guiding light of natural evolution, of human innovation and novelty search, then we might harness it for our own purposes. Instead of something to fear it can be something to embrace.

From: Why Greatness Cannot Be Planned, Kenneth by Stanley and Joel Lehman

Business is a never-ending quest to deliver the same result in an easier fashion.

When the first voice-activated speakers were released—products like Google Home, Amazon Echo, and Apple HomePod—I asked a friend what he liked about the product he had purchased. He said it was just easier to say “Play some country music” than to pull out his phone, open the music app, and pick a playlist. Of course, just a few years earlier, having unlimited access to music in your pocket was a remarkably frictionless behavior compared to driving to the store and buying a CD. Business is a never-ending quest to deliver the same result in an easier fashion.

Similar strategies have been used effectively by governments. When the British government wanted to increase tax collection rates, they switched from sending citizens to a web page where the tax form could be downloaded to linking directly to the form. Reducing that one step in the process increased the response rate from 19.2 percent to 23.4 percent. For a country like the United Kingdom, those percentage points represent millions in tax revenue.

The central idea is to create an environment where doing the right thing is as easy as possible. Much of the battle of building better habits comes down to finding ways to reduce the friction associated with our good habits and increase the friction associated with our bad ones.

James Clear: Atomic Habits: Tiny Changes, Remarkable Results


This is a PyTorch implementation of the DrQA system. DrQA is a system for reading comprehension applied to open-domain question answering. DrQA is targeted at the task of machine reading at scale (MRS). In this setting, we are searching for an answer to a question in a potentially very large corpus of unstructured documents, that may not be redundant. Thus, the system has to combine the challenges of document retrieval, i.e. finding the relevant documents, with that of machine comprehension of text (identifying the answers from those documents).

The Byzantine Generals’ Problem

After verification, transactions are relayed to other nodes in the peer-to-peer network. The other nodes will repeat the verification and relay the transaction to more nodes. Within seconds, the transaction should reach most of the nodes on the network. Transactions are then held in pools on the nodes awaiting insertion into the block chain, which is a public record of all transactions that have ever occurred in the Bitcoin network. The block chain is not just a simple list of transaction receipts though. It is specially designed to solve the double-spending problem for a peer-to-peer network of untrusted nodes. As discussed in the Double-Spending section, the goal is to determine the chronological ordering of transactions so that the first payment can be accepted and the second payment can be rejected. So the peer-to-peer network has to have a method to agree on an ordering of transactions even though some peers might be trying to sabotage the system.

This problem of coordinating a group of peers with possible saboteurs is known as the Byzantine Generals’ Problem. The name comes from a 1980 paper[20] that described a situation in which the generals of the Byzantine Empire’s Army had to agree on whether to launch their attack on the enemy. The generals could communicate by messenger, but any of the generals could be a traitor trying to sabotage the attack. If a treasonous general or group of generals could confuse enough honest generals about the outcome of the decision, the army would be fragmented and meet with defeat. The challenge is to develop a protocol that ensures that the majority’s decision is heard by all generals even when not every general can communicate directly with every other general.

This is just like how all the nodes in the Bitcoin network have to know the majority’s decision about the chronological ordering of transactions, even though not all nodes can communicate directly and some nodes might be malicious attackers.

The Solution

The block chain solves the Byzantine Generals’ Problem by using computational power as a voting system. First, nodes group transactions into blocks that are linked to form the block chain. Nodes broadcast blocks to the entire peer-to-peer network upon creation. Each block contains a hash of the previous block in the chain. Therefore, at the time a block was created, the previous block must have already existed or else the hash would be invalid and the block would be rejected. This provides verifiable proof of the chronological ordering of the blocks in the chain. But if blocks could easily be added by anyone, an attacker could just make an alternate chain that reassigns ownership arbitrarily.


The block chain is not really an easily scalable solution. All nodes have to see every transaction in the world and the block chain is stored in full on every node. In other words, Bitcoin is not a distributed system, it’s just massively replicated. Future versions of the Bitcoin software may make it possible to reduce the storage requirements for the block chain, but collecting all unprocessed transactions on every node is not a requirement that can be easily removed. It works fine currently, but if Bitcoin becomes more mainstream, scalability could be an issue. It may require such an expensive computer to operate a node that individuals wouldn’t be able to do it. At that point, Bitcoin loses some of its major benefits. For example, it would be easier for governments to regulate. Consider a situation in which a small number of companies control the majority of computational power in the Bitcoin network. The government could easily regulate these companies and force them to install software updates that give the government backdoors to manipulate the currency.”

Source: Clark, Chris - Bitcoin Internals

Merkle Trees

Cryptographic hash functions are often used to verify the integrity of a list of items, such as the broken-up chunks of a large download. In such cases, one option is to merge all the chunks and take the hash of the complete download. The problem with this is that if one chunk is corrupted, the user won’t find out until the entire download is complete, and even then they won’t know which chunk is corrupt. A better solution is to take the hash of each chunk individually so that each chunk can be verified as it comes in. However, if there are a large number of chunks, then there is a greater chance that some of the hash values will become corrupted. Furthermore, this is a lot of data for the trusted source to store. Ideally, a trusted source would only have to provide one hash, and the rest of the hashes could be downloaded from untrusted sources, such as peers in a peer-to-peer network. This can be accomplished using a top hash generated by hashing all of the hashes of the chunks. The resulting structure is called a hash list.

If the number of chunks is very large, the list of hashes of all the chunks might also be quite large. In order to verify just one chunk against the trusted top hash, one would need to obtain all of the hashes in the hash list. Ralph Merkle proposed the idea of a hash tree in 1979, which allows a chunk to be verified with only a logarithmic number of hashes.[6] In a hash tree, or Merkle tree, hashes are taken of all chunks as in a hash list, but then these hashes are paired and the hash of each pair is taken, and these hashes are then paired again, and so on until there is only “one hash at the top of this tree of hashes.

To verify the integrity of just one chunk, it is only necessary to obtain a small subset of the hashes in the hash tree. For any hash in the tree, if the desired chunk is not in the branch below it, then that branch can be stubbed out by dropping it and keeping only the hash at the top of that branch. For example, if you wanted to verify data block 1 in the diagram, you need Hash 0-0, Hash 0-1, Hash 0, Hash 1 and the Top hash. The branch rooted by Hash 1 can be stubbed out, removing Hash 1-0 and Hash 1-1, keeping just Hash 1 to represent the whole branch. For large trees, the number of hashes needed to verify one chunk can be much smaller than the number of chunks.

We Are Data

[Book recommendation: We Are Data by John Cheney-Lippold]

This book has two intersecting purposes. First is to understand how algorithms transcode concepts like gender, race, class, and even citizenship into quantitative, measurable-type forms. Second is to recognize how these measurable types reconfigure our conceptions of control and power in a digitally networked world. The political battles that surround structures of patriarchy, white supremacy, and capitalism writ large must necessarily attend to the terms of algorithm.

In other words, while HP’s facial-recognition cameras might promote asymmetrical usages across racial lines, an algorithmic ‘race’ complicates this asymmetry. HP’s response is indicative of this complication, in that HSV contrast, not skin color, shared history, or even DNA, stands in for the concept of race. We’re talking about HP’s construction of a ‘white’ ‘face,’ not White Wanda’s white body. But we’re talking about whiteness nonetheless. Race is incessantly being made and remade, and we must focus on how digital technology also makes and remakes ‘race’ on algorithmic terms.

Explicating the terms that underpin this making/remaking is the ultimate goal of the book. In the following pages, we’ll talk about jazz, terrorists, HP being racist (again), marketing, the NSA, citizenship, and even Santa Claus. The shift to the data/algorithm ontology of the computer conceptually moves identity past explicit, policed boundaries that require negation and exclusivity (either male or female, at risk or not, black or white).

This move lays the foundations for a plane of smoothness, an open set of possibilities where we play on the limits of established truth. Algorithmic identity doesn’t declare that you are just ‘male’ or ‘female.’ Statistical confidence and probability, even the chance that this book will spontaneously combust, can never be 100 percent anything. Rather, you’re likely to be 92 percent confidently ‘male’ and 32 percent confidently ‘female.’ Algorithmic ‘race’ and ‘gender’ isn’t about being a white man. It’s about being a ‘Caucasian’ ‘man’ with a confidence measure of 87 percent. In algorithmic identity, we confirm the inorganic realities of Donna Haraway’s cyborg, one who is “not afraid of permanently partial identities and contradictory standpoints.

Chapter 1. Categorization: Making Data Useful

In order to compute something like ‘woman’ or ‘smiling,’ one needs to first make data useful. In chapter 1, I describe the how-to of algorithmic knowledge production. This how-to centers on how computers create categories through patterns in data, which then construct algorithmically transcoded ideas about the world that I call measurable types. Algorithms are neither magical nor mysterious. Instead, they make data useful through a very intricate but, I promise, also very interesting constellation of different technologies (like metadata or marimbas) that then create different algorithmic identifications (like ‘terrorist’ or ‘John Coltrane’).

Chapter 2. Control: Algorithm Is Gonna Get You

Measurable types are much more than descriptive containers for algorithmic meaning. They also play formative roles in how life and knowledge is controlled. With the aid of Gilles Deleuze’s concept of modulation, I theorize how the deluges of data we produce online help enact a form of control. This type of control substitutes the strongly worded, hard-coded prohibitory “no!” of traditional modes of power in exchange for what some scholars have called “control without control”—and that I call soft biopolitics. These soft biopolitics describe how our algorithmic identities can regulate life without our direct participation or realization.

Chapter 3. Subjectivity: Who Do They Think You Are?

Soft-biopolitical measurable types structure our lives’ conditions of possibilities every time we buy a plane ticket, cross a border, or translate a document on Google Translate. While we are ceaselessly made subject to different arrangements of algorithmic knowledges, these datafied subject relations are foreign to our most immediate experiences. We are not individuals online; we are dividuals. And without the philosophical anchor of the individual to think alongside, we are often at a loss in how we interpret ourselves as users. This chapter explores how algorithms make us subject in ways unique to online, algorithmic life.

Chapter 4. Privacy: Wanted Dead or Alive”

How does one practice privacy in a world where not only is almost everything surveilled but that surveillance is rarely, if ever, felt? I evaluate privacy’s legacy and outline its origins in the nineteenth-century phrase “right to be let alone,” in order to bring that history into conversation with the exigencies of our contemporary era. I argue that privacy cannot just be about whether you have a password on your email or whether there are doors on a bathroom stall. Privacy must be a practical response to the lived restriction and control implicit in ubiquitous surveillance. In this way, I theorize a dividual privacy that focuses especially on how the freedom in being “let alone” might translate to a datafied, algorithmic world.
Conclusion: Ghosts in the Machine

At the end of the book, I return to my central arguments: online we are made, read, interpreted, and intelligible according to data. Our world, and the knowledge that gives it its meaning, is increasingly a datafied world. We are subsequently understood in the datafied terms of dynamic, soft-coded, and modulating measurable types. The contemporary encounters we have with ubiquitous surveillance suggest a new relationship to power that I term soft biopolitics. And the resulting ubiquity and emergent configurations of these different types of knowledge force us to rethink how subjectivity functions and what it is that privacy can practically defend.”

Source: John Cheney-Lippold. “We Are Data: Algorithms and The Making of Our Digital Selves.”


11 CHARLIE MUNGER quotes I like

“It’s remarkable how much long-term advantage people like us have gotten by trying to be consistently not stupid, instead of trying to be very intelligent. There must be some wisdom in the folk saying, “It’s the strong swimmers who drown.”
What’s the flip side, what can go wrong that I haven’t seen?

“The idea that everyone can have wonderful results from stocks is inherently crazy. Nobody expects everyone to succeed at poker.
If [investing] weren’t a little difficult, everybody would be rich.
For a security to be mispriced, someone else must be a damn fool. It may be bad for the world, but not bad for Berkshire.

“Most people who try [investing] don’t do well at it. But the trouble is that if even 90 percent are no good, everyone looks around and says, “I’m the 10 percent.”

It’s a very simple set of ideas and the reason that our ideas have not spread faster is they’re too simple. The professional classes can’t justify their existence if that’s all they have to say.

I don’t let others do projections for me, because I don’t like throwing up on the desk.

What is elementary, worldly wisdom? Well, the first rule is that you can’t really know anything if you just remember isolated facts and try and bang ’em back. If the facts don’t hang together on a latticework of theory, you don’t have them in a usable form.
You must know the big ideas in the big disciplines, and use them routinely—all of them, not just a few. Most people are trained in one model—economics, for example—and try to solve all problems in one way. You know the old saying: to the man with a hammer, the world looks like a nail. This is a dumb way of handling problems.
All the wisdom of the world is not to be found in one little academic department. That’s why poetry professors, by and large, are so unwise in a worldly sense. They don’t have enough models in their heads.

Source: Tren Griffin “Charlie Munger: The Complete Investor (Columbia Business School Publishing)


5 Quotes I was thinking about today

I never, indeed, wavered in the conviction that happiness is the test of all rules of conduct, and the end of life. But I now thought that this end was only to be attained by not making it the direct end. Those only are happy (I thought) who have their minds fixed on some object other than their own happiness; on the happiness of others, on the improvement of mankind, even on some art or pursuit, followed not as a means, but as itself an ideal end. Aiming thus at something else, they find happiness by the way.

—John Stuart Mill

“Visionary companies pursue a cluster of objectives, of which making money is only one—and not necessarily the primary one. Yes, they seek profits, but they’re equally guided by a core ideology—core values and sense of purpose beyond just making money. Yet paradoxically, the visionary companies make more money than the purely profit driven companies.
—Jim Collins and Jerry I. Porras
He is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention. By pursuing his own interest he frequently promotes that of society more effectually than when he really intends to promote it.
—Adam Smith
Tell all the truth, but tell it slant. Success in circuit lies.
—Emily Dickinson

What is the highest good in all matters of action? As to the name, there is almost complete agreement, for uneducated and educated alike call it flourishing, and make flourishing identical with the good life and successful living. They disagree, however, about the meaning of flourishing.