Day 6 of 100 Days of AI

Yesterday, I skimmed through k-fold cross-validation. I wanted to return to it today because I still wasn’t sure about what exactly we are assessing with it, and why.

Here‘s the clearest Youtube video I found on the topic. The screenshot below from that video provides an easy summary.

Key takeaways:

  • K-fold cross validation is a method of evaluating the performance of a specific model configuration.
    • Model configuration includes the choice of an algorithm (e.g. linear regressions, decision trees, KNNs, neural networks), feature selection (e.g. the independent variables used to predict the dependent variable), and hyperparamters (e.g. the number of neurons in a neural network.)
  • To perform K-fold validation you,
    • (1) split the data into ‘k’ equal parts (the “folds”);
    • (2) you train k-number of models using your chosen ‘model configuration’ across the k-number of folds; for example in the image above, model 1 is trained on folds 2-5 and tested on fold 1; model 2 is trained on folds 1 and 3-5, and tested on fold 2 and so forth;
    • (3) you then assess the performance of each model across the k-folds and aggregate this performance measure, usually using an average (or median, depending on the data type);
    • (4) the final accuracy tells you how good your model configuration is.
  • If your aggregated accuracy is strong, you can then train your model on the whole data set.
  • K-fold cross validation is useful in situations where you have a small amount of data and a small test set. In those instances, the opportunity to test your model is limited, and k-fold can boost that, allowing you to test a model configuration across a larger number of scenarios.

Day 5 of 100 Days of AI

Today I learnt about cross validation (scikit-learn has Python helper functions for this.) This is where you split your training and testing data into different sets, and then you iteratively train and test against different combinations to assess how well a model performs on unseen data.

Two common methods of cross validation are k-fold validation and leave-one-out cross validation.

K-fold involves splitting the data into equal parts and rotating across them in terms of testing and training.

With leave-one-out cross validation, you select one data point for testing, and use the rest for training. You then move to another data point for testing, and then use the remainder for training, and so forth, until every data point has been used for testing.

I’ll return to these concepts when I write some more code next week.

Ps. One thing I’m realising as I dig into the basics of machine learning is that it’s a mix of art and science in terms of choosing techniques that may produce the best models. There’s a bunch of trial and error, even though it’s a deeply rigorous and mathematical field.

Day 4 of 100 Days of AI

Today, I went through a classifier lab on the intro ML course. There are several bits I didn’t quite understand but GPT helped get me over the basics. For example, I will need to review my notes on the Jaccard Index and F1-score (evaluation metrics for classifier models), and the concept of normalisation, where you transform your data without changing its distribution. This makes it easier to calculate distances between points, a critical bit when trying to make classification predictions.

On the latter point, I’ve included some charting code in the github repo here (see image below), which helped me understand the normalisation concept. The charting code was written by GPT, with some minor tweaks from me.

Key takeaways:

  • Classification is a supervised machine learning approach.
  • It makes a prediction about what discrete class some item should fall into.
  • Classifiers can be used for spam detection, document classification, speech recognition, or even to predict if a certain customer will churn, based on a variety of characteristics.
  • Classification algorithms include k-nearest neighbour (which I’ve put on github here), decision trees, and logistic regression (which instead of putting an item into a class, gives you a probability that it will fit a particular bucket.)
  • The K-nearest neighbours algorithm was fun to learn about, and the intuition for it is simpler than I expected. The basic notion is as follows: for a given item to predict on, look at a select number of neigbhours (the k-number), and predict the outcome based on the most popular category that those neighbours are in (or the neighbours’ mean or median of the values you’re trying to predict for e.g. house price based on location, square foot size etc.)
  • Classification algorithms can be evaluated with a number of accuracy measures, such as the Jaccard Index, a F1-score, or Log Loss. I didn’t cover these in detail but I did enough to get the very basics.

Day 2 of 100 Days of AI

I spent bits of today watching Youtube videos about linear regression. One of the best ones I found was this StatQuest video on the topic. I enjoyed his style so much, I ordered his introductory book on Machine Learning. I’m planning a digital fast over a long weekend in the second quarter of the year, and I’ll take the book with me to continue learning offline.

Today I learnt about a new concept: Overfitting.

Key takeaways today:

  • Overfitting: This is when a machine learning model fits too strictly to the training data that it was trained on. So much so, that the model doesn’t generalize well to data outside of the training set (i.e. new data.)
  • Why does overfitting happen? A few drivers:
    • Small training data set — limited training examples might throw up patterns that don’t account for the true distribution of the data.
    • High model complexity — having too many parameters might throw up noise or patterns that are not useful;
    • Biased data — this could throw up patterns that are not generalizable across wider a population of data.

Day 1 of 100 Days of AI

Today I’ve started a “100 days of AI” challenge. I’ve used LLMs across a few personal projects here, here, and here. But I’d like to understand the basics of AI and machine learning a bit better.

My motive isn’t to retrain as a ML engineer or data scientist. Instead, I want to challenge myself to go beyond a superficial understanding of one of the greatest technology advancements of our time, particularly given my work in technology investing.

Every day, I’ll aim to do anything from 10 to 90 mins of learning, experiments, and review. Here we go!

Day 1: Overview & Learnings

I’m working my way through an introductory machine learning course by IBM, which provides an AI Engineering Professional Certificate at the end.

I’m about 25% of the way through it, and already, we got to build a simple linear regression model that predicts CO2 emissions based on engine size. I’ve put the GitHub code for this here. Most of the action comes from the code snippet below.

Key takeaways today:

  • Artificial intelligence (AI) is a broad definition that covers computer system that performs intelligent human-like functions.
    • Machine learning is a subset of AI that uses statistical techniques to learn from data and infers patterns and makes predictions.
    • Deep learning is a subset of machine learning and it uses neural networks (inspired by the brain structure) to learn from data.
    • There’s a good explainer of all this here.
  • I’m going to be looking at mostly machine learning for now. This can be split across supervised learning (where you provide data that’s ‘labelled’) and unsupervised learning (where the data is ‘unlabelled’).
  • Popular techniques for ML models include regression and classification (in supervised learning), and clustering and association (in supervised learning).
  • There are a number of open source libraries that make it easier to prepare and build machine learning models. The key ones I’ll probably use a lot are NumPy, SciPy, Pandas, and Scikit learn.

Many concepts feel foreign to me at the moment, but as I spin up a couple of projects and work through tutorials, I should start to get the basics down!

What Drives Power Laws in Entrepreneurship and Venture Capital?

Venture capital is a power law business: most of the underlying investments fail¹ while a tiny fraction generate the bulk of returns. This is in contrast to normal distributions where most outcomes cluster around some central value.²

The chart below — which roughly follows a power law distribution — shows that most investments (65% of companies) don’t work out but a few (4% of companies) can generate 10 to 50 times your money or more.

(Data from Correlation Ventures via Seth Levine)

Moreover when you look at fund-level performance, you find that it’s that very small number of investments that drive aggregate fund returns. This can be seen in data from Horsley Bridge — a limited partner in A16Z and other funds — which found that 6% of investments generated 60% of fund returns.

(LP data on VC firms via Benedict Evans/A16Z)

But why exactly do venture capital fund returns (and the underlying entrepreneurial ventures) follow a power law distribution? What leads a tiny minority to significantly outperform the overwhelming majority? And when you understand power law drivers, can you tilt the odds in your favour?

To answer these questions I spent a few evenings and weekends scouring academic literature on power laws. Admittedly some of the maths involved is beyond me, but I did pick up on 3 possible power law drivers that are worth internalising when starting a business or investing in startups.


Power Law Drivers in Entrepreneurship and Venture Capital

(1) Preferential Attachment 

This power law mechanism is commonly expressed in the maxim ‘the rich get richer’. Or if you’re religiously inclined, the Bible says ‘for whoever has, more will be given to them.’

Academics call this phenomenon preferential attachment (or cumulative advantage). It’s when an initial endowment makes subsequent gains more likely. This leads to a privileged position where early wins lead to further benefits that cumulatively add up to an outsized advantage.

I’ve seen this privilege play out in my life in a small way. Getting my first book deal was a low probability event but once that door was open, getting the second, third, and fourth book contracts became more likely.

Preferential attachment isn’t always fair and it contributes to inequalities in the world that I won’t go into here. However, understanding it in the context of entrepreneurship and venture capital can inform strategy.

(Preferential attachment in action. Notice how the larger blobs more easily scoop up the smaller ones. Original video here.)

We see preferential attachment in entrepreneurship when an ex-Google engineer finds it easier to raise money and recruit talent for their startup because of their pedigree. We also see it in network effects, where each user added to a platform increases the likelihood that other users will flock to it. 

Meanwhile in venture capital the best startups preferentially attach to funds which already have success in their portfolio. In other words, success begets more success and as Samir Kaji puts it: “One massive hit is often all it takes to “mint” an investor, regardless of how serendipitous the investment was.”

To benefit from preferential attachment, identify positive feedback loops in your field then look for ways to engineer them to your advantage. 

(2) Self-organized Criticality

Despite the unusual name this power law driver is something you’re already familiar with. We see it in ‘overnight success’, which is actually an inaccurate description of something that’s so common in nature that theoretical physicists gave it a technical label 30+ years ago: ‘self-organized criticality’.

This power law mechanism is expressed in snow avalanches, neural networks in the brain, earth quakes, financial market crashes, and even social upheavals. It’s a process where lots of seemingly benign interactions in a complex system can ‘self-organise’ that system to a ‘critical’ state3 such that even the tiniest subsequent input can unexpectedly lead to dramatic change.

A useful analogy here is a pile of rice. If you build one by adding a few grains at a time, most of the grains don’t have much impact. But after a while, adding just one grain of rice can lead to an avalanche.

(Video sourced and edited from Andrew Hoffmann.)

This is similar to overnight success. Except that even though the description suggests instant success from nowhere, most successful people make it only after years of metaphorically adding small grains of rice to a pile of effort. 

(You can read Sarah’s story here and from this tweet.)

In startups, getting to product-market-fit is a similar affair. Companies have to iterate continuously until something clicks so that the business can start to scale. Even in later years, those same businesses have to persevere before they can benefit from a step-change in growth.

One example I really like here is that of Microsoft when it was still a startup. In 1980 it secured a landmark contract to supply IBM with an operating system — an event that arguably changed the course of tech history

However, this event didn’t pop into existence on its own. Lots of prior events had already placed Microsoft at a point of “criticality”. To name but a few:

  • Bill Gates and Paul Allen had been writing and selling software for almost 10 years already and were well-suited to getting the job done. 
  • Among other reasons IBM had attracted decades of anti-trust investigations against it and to avoid further regulatory scrutiny, it accepted a non-exclusive contract with Microsoft. 

The latter point meant that Microsoft was free to sell its software to other computer manufacturers and as PC hardware become commoditised, the business grew without restriction and saw its revenues balloon almost 10x, from $16m in 1981 to $140m in 1985.

Uneventful steps taken frequently can also lead to outsized outcomes in venture capital. Fred Wilson was writing and thinking about bitcoin for years before he met and invested in Coinbase. In fact he met the company because he was willing to persevere through a lengthy office hours session at Y Combinator with 16 startups across 4 hours of back-to-back pitches.

To benefit from self-organized criticality, play the long game and take heed of Seneca’s observation that “luck is what happens when preparation meets opportunity”. Or putting things more poetically:

“Chance can be on our side if we but stir it up with our energies, stay receptive to the glint of opportunity on even a single hair above the underbrush, and continually provoke it by individuality in our attitudes and approach to life.” — James H. Austin in Chase, Chance, and Creativity.

(3) Multiplicative Processes 

Power laws also emerge when events are multiplicative instead of additive. One example is word-of-mouth. If you have a fantastic restaurant experience you don’t just tell one other person about it (which would be an additive process). You tell lots of friends who then go on to tell many others too.

Other examples of this process include population growth, the spread of viruses, and rapid wealth accumulation (getting rich through investing is a multiplicative process while building wealth through a salary is additive.)

Multiplicative processes are perhaps the simplest power law generator: a value is multiplied by some variable and the result is further multiplied by another variable. Repeat this process and you get exponential growth. 

(Viral growth simulation by Grant Sanderson.)

In startups, hiring a ‘10x employee’ is a multiplicative process. An elite performer can substantially change the trajectory of a business while most hires tend to be additive.⁴ For VCs, blogging is multiplicative networking. Ideas can spread faster online compared to offline 1–2–1 conversations.

To benefit from multiplicative processes, you have to identify and seek out multiplicative factors — much like preferential attachment — then look for ways to engineer them to your advantage.5


Engineering a Power Law Mindset

Outsized success is rare and takes just as much (if not more) in luck as it does effort to hit a home run. However, if we internalise how power laws work, we become better placed to tilt the odds in our favour by working not just harder, but also smarter.

(3 Power Law Drivers and Examples in Entrepreneurial Ventures)

Notes

[1] Businesses fail for a variety of reasons (e.g. team issues, lack of funding, creating products that people don’t need, competition and market timing.) But while building for the future is inherently unpredictable and hard, it’s worth remembering that..

“In any natural system, failure is the engine that causes growth, that causes new birth, that causes anything to happen. One of the truly big differences between growing economies and economies that stagnate is the acceptance of failure. If you don’t let forests burn, if you don’t let the old trees die out and the new trees grow, you don’t get a healthy forest. The ability to manage failure so that enterprises fail but people can still succeed becomes one of the tricks of how you build a society that can reinvent itself as the world changes.” — Shikhar Ghosh in the Harvard Business Review.

[2] In normal distributions most outcomes cluster around a central value. Human height is one example of this. Over two thirds of the population are close to average height while a minority are really short or really tall.

A height normal distribution example (chart by Michael Minn)

[3] A ‘critical state’ is typically a hypersensitive state between order and disorder (or vice-versa).

[4] Data from 5 studies covering 600,000+ professionals shows that performance in many domains isn’t normally distributed. A small portion of talent outperforms the majority by a significant margin. See examples below from this paper

[5] eBay leveraged the Beanie Baby collectibles mania that run from 1996 to 1999 in order to grow faster. Brian McCullough provides a good historical account of this in his book.

The Antidote to Getting a VC Job: 10 Prescriptions for Aspiring Investors

Photo by Adeolu Eletu on Unsplash

Photo by Adeolu Eletu

Preface

Venture capital (“VC”) found me through entrepreneurship. In fact I knew nothing of it until I came across the blogs of early-stage investors Fred Wilson and Paul Graham.

At that time, I was reading everything I could find on startups and technology. I had a small online business I’d founded while at university and was increasingly fascinated by the art of building things (including companies), technology, and the future.

So when I discovered that there was a job where you supported and backed a portfolio of ambitious technology startups, I knew I had to find a way in on the action.

However, I didn’t fit the profile: I wasn’t an ex-operator or engineer. I didn’t work at a top-tier investment bank or consulting firm. I had no MBA. I lived and worked outside of London and my network was lacking.

It would take 5 years of stepping stones, career experiments, and what at times looked like unfocussed meandering to others, before I eventually landed a role as an Associate at Downing Ventures.

I now get asked almost every week for advice on how to get a VC role. I often hesitate in sharing much, mainly because I’ve only done this once and what works for one person at a certain time won’t necessarily work for others.

That said, I get the VC career question so often and since I’m keen to help more people but can’t always meet individuals for coffee or take calls with them, I decided to weave together a comprehensive blog post on the topic that can provide guidance at scale.

To be more helpful, this article takes account of experiences beyond my own and includes anecdotes of both prominent and lesser known investors. But be warned: there is no sure-fire path to a VC job. Fortuitous circumstances and luck play a substantial role and this is mostly beyond our control.

However, if you reflect on your aptitudes and experiences, complete this Guy Kawasaki test yet remain driven, and know for sure that that even if it takes 3 to 5+ years to get there you are willing to pursue a VC career, then this blog post is here to help you get started.

Introduction

steve

Steve Jobs receiving a seed cheque for $250,000 in 1977 from investor Mike Markkula

A job in VC is highly coveted. It’s one of the most impactful professions (Apple, Google, Tesla, Amazon, and Fedex were all VC-backed) and it can be highly rewarding but it’s also one of the most competitive to crack.

Today there are countless guides on how to break into the industry and with global venture investment at record highs (over $254bn deployed in 2018 alone — the highest since the dotcom bubble!) hiring in the sector has picked up pace.

Make no mistake though, the venture industry is still tiny and employment is sparse. In the UK, for example, there are 170 active firms (up from less than 50 a decade ago) with only 1,400 investment professionals in the country (the USA is around 6x that amount.)

At any one time there are a few dozen vacancies advertised in the UK in addition to roles that are barely public and are filled through referrals. Contrast that to the legal profession, which has close to 10,000 firms in the UK, employs some 100,000 lawyers, and at any one time has thousands of job vacancies.

Getting a VC job certainly takes some luck and if you’re an outsider, you’ll have to work doubly hard. Nonetheless, it may be possible over the long term to cumulatively stack the odds in your favour for a snowballing advantage, but even then, I wouldn’t recommend you entirely stake your career happiness on getting a role in the sector.

More realistically, and as you will see with Prescription 10, you are better off being open to a wider range of career possibilities because if you have what it takes to work in VC, you can still find fulfilling work in other domains.

The Antidote

So how does one to get a job in VC? In some ways you have to do exactly the opposite of what you would instinctively want to do. Most VC job guides provide tactical steps (join a startup, attend events, network with entrepreneurs/investors etc), which is no doubt useful — and you’ll get some of that here too — but considering tactics in absence of higher-level strategic principles inspires a short-term mindset and plays to our craving for quick fixes and easy wins.

This post provides an antidote to our instinctive desire to seek ‘the easy path’, which doesn’t really exist. My hope is that this guide will alleviate false assumptions on how to get a VC job, all the while leaving you better off by inspiring long-term ideas on what you can do to pragmatically improve your chances.


The 10 Prescriptions

What follows is a list of 10 prescriptions of what I’ve observed as effective from the VC career paths of investors in the UK and USA. It’s impossible to excel at all 10 things (I struggle with a few myself) but if you can excel at a few and do the rest of them reasonably well, you’ll stand out and have a better shot at making it.

As you read the prescriptions, be sure to keep in mind what a VC job entails— i.e. (1) finding attractive investment opportunities, (2) earning credibility and trust with the best founders so that they invite you to invest in them (3) helping those founders succeed—and consider what you can do to develop the skills necessary for the job even before you have it.

In no particular order, here are the 10 prescriptions:

1. Don’t ask for help. Be the helper. Don’t be extractive and transactional. The venture community is small and a venture career is long. Find ways of helping others do well. “It’s easier to win if everyone wants you to win,” says Randy Komisar of Kleiner Perkins. And helpers tend to have this effect on people.

So help founders, help investors, and do what you can to support the tech and entrepreneurship community. Least of all you will feel good doing it, but do help even when you expect nothing in return.

A word of caution though: don’t revert to a VC trope and end every meeting with a disingenuous “let me know how I can help” when you don’t really mean it. Find things you can actually help with, while giving consideration to your existing commitments.

Case Note: Prior to joining Backstage Capital London, Andy Davis was already going above and beyond for founders. He built an intimate community of black tech entrepreneurs based on regular meet-ups he hosts and it has grown through word-of-mouth on the value it brings to members.

Andy was helping founders with business models, pitch decks, introductions, and fundraising all before he was recruited to join Backstage Capital. His contribution to the tech community was so evident that whichever fund hired him, it would be partnering with a true value-add individual.

Further Reading: Give and Take.

2. Don’t cold contact. Build relationships. Some people spray-and-pray templated LinkedIn messages, emails, and generically try to cold contact people they don’t have any connection with. Sure, some of these messages will get a response but many get ignored given the volume of inbound messages that investors receive.

So avoid cold contacting where possible. If another trusted individual can introduce you instead, that will get more attention.

How do you get “warm” intros? By building long-term relationships. Someone you’ve met once is unlikely to introduce you with any credibility to their network. But someone you’ve passionately discussed ideas with at an event, or perhaps someone you’ve worked with on a project or socialised with — that person will be more willing to credibly connect you to interesting contacts.

Case Note: There’s a philosophy in tech entrepreneurship about starting a company by doing things that don’t scale. It takes more work but if at a small scale you can “recruit users manually and give them an overwhelmingly good experience,” writes Paul Graham, you’ll find that “it’s like keeping a fire contained at first to get it really hot before adding more logs.”

The same can be said of a personal network. Starting small but with depth is more impactful than networking wide but superficially.

Here’s Ana Díaz Hernández recalling how she landed a role at Kapor Capital:

“I fostered a personal connection with partners at the firm. I met Mitch Kapor and Freada Kapor Klein through mutual interests in advancing diversity in tech and began to learn about the great work they did through Kapor Capital and the Kapor Center for Social Impact.

As a Latina in the startup world, the diversity work of the Kapor Center was very resonant. I had been interested in venture capital for a while, but it was our relationship, our values alignment, and the desire to work together on advancing social impact in technology startups that got me to join the team.”

Further Reading: Never Eat Alone. Also see Friends as Ends in Themselves.

3. Don’t be interested. Be interesting. Everyone that applies for VC jobs says they are interested in technology and entrepreneurship. However, that’s table stakes and it doesn’t say much about you. To stand out, you have to evidence your interest and passion by doing things that are interesting.

Instead of just reading TechCrunch, following VCs on Twitter, and listening to tech podcasts, invest time in doing things that without a doubt evidence the depth of your passion and interest in technology and entrepreneurship.

I can’t prescribe exactly what you need to do here but there are some rules of thumb that can help you identify opportunities that will serve you well. The chosen activity or venture should ideally:

  • quench a personal curiosity;
  • require significant time investment;
  • contribute to a discussion or topic entrepreneurs and investors care about;
  • be relatively original.

Case Note: An inspiring example here is Jenny Gyllander. At university, she conducted 18 interviews with VCs for her 110-page thesis titled:

‘Dear VC, now it’s your turn to pitch’ — an exploratory study on Venture Capital firms’ brand and reputation.

Clearly Jenny’s interest went beyond what most people who are interested in VC’ do, and it didn’t stop there.

After university, Jenny worked at a design agency before joining Slush, one of Europe’s largest tech conference organiser. She was quickly promoted to CMO thanks to her impact and some years later she was recruited by the team at Backed VC.

Even though Jenny had now joined a VC fund, she continued to explore her interests with depth. This eventually manifested through a side project called Thingtesting, a dedicated Instagram channel to discover and showcase emerging direct-to-consumer brands. With over 25k followers on the gram, Jenny has since turned Thingtesting into a full-time job.

Further Reading: Pick the Idea That’s Craziest.

4. Don’t “pick brains”. Present theses. Remember Prescription 1? Many people violate it by cold contacting investors and asking to pick their brains over a coffee. This is a bland way to connect with someone if you are looking for career advice and help. Most investors ignore these one-sided requests.

A better way is to reach out (ideally via a warm intro) with a specific topic that is relevant to the person you would like to connect with. For instance, has the person recently announced a deal in an area you’ve been tracking and you have ideas to share on how it will develop?

Consider reaching out to people with a view to share knowledge and engage in discussions. That’s more compelling than ‘can I pick your brains?’

Case Note: Before getting into VC, Andrew Chen shared his startup knowledge by authoring hundreds of essays online. This got the attention of one the founders of A16Z, which eventually recruited him to join the fund. As he remembers:

“I moved to the Bay Area in 2007, as a first time founder with a lot of energy and a lot of questions. I spent the first year meeting everyone I could, reading everything about tech, and writing down all that I was learning. A few months in, I was shocked to get a cold email from Marc introducing himself. Who knew that sort of thing happened? My blog was pretty much anonymous and I could be anyone — but he reached out to talk ideas, which made a big impression.”

Further Listening: Inventing the Future with Josh Wolfe.

5. Don’t apply online. Get referred. Job postings in VC get many hundreds of applicants within a short time of being posted. And these applicants are far from average. They are smart, ambitious, and often have compelling work experience. Standing out in a swarm of cold but truly exceptional CVs is challenging. And even if your CV does get attention, the document rarely conveys your story in its truest and best light.

While I wouldn’t completely rule out applying online, if you are following Prescriptions 1 and 2, you will be in a better position to first seek help from people who know you well and who might be able to connect you to a particular fund that is hiring.

These people can help surface your CV by referring you, thereby making a recruiter’s job easier. After all, if a trusted party can vouch for you, it saves hiring managers time from filtering through hundreds of other CVs.

To be worthy of a strong referral, see Prescriptions 1 through to 4, and invest in demonstrating that you have the potential to do well in VC.

Case Note: Legendary investor Bill Gurley of Benchmark got his break when a newsletter he wrote on the tech industry caught the attention of a well-connected investment banker. He recalls:

“Frank Quattrone [the investment banker] called me out of the blue and said, “We’re leaving Morgan Stanley, we’ve heard a lot of things about you, we want you to join us.”

Frank and I had a long talk, and he said, “What do you want to be long term?” I said, “I’d love to be a VC.” He said, “Come to work for me, I’ll move you to Silicon Valley and introduce you to every venture capitalist that I know.””

Bill subsequently went off to work as a tech analyst for Frank Quattrone. And sure enough, his work efforts and network paid off with referrals that launched his VC career: Bill Gates referred him to the VC fund Hummer Winblad, which he joined briefly. Then Frank Quattrone introduced him to Benchmark, which he joined in 1999.

Further Reading: How to Be Great at Your Job.

6. Don’t be an expert. Be a generalist. If you are early in your career and break into VC, you will likely start out as an Analyst or Associate, in which case a fund wouldn’t expect you to join guns-blazing as an expert in a specific area. You will still have much to learn and be expected to work across a wide range of tasks.

So at least initially, you’re better off starting with a broader set of experiences. Not only will you better relate to founders across several business functions, you’ll get to know yourself better, with a wider range of experiences about what you could be exceptional at. Once you’re a more experienced investor, you can use this knowledge to help further your specialism.

Is there a generalist skills palette that can serve you well coming into the VC industry? Once again, there is no formula but a foundational understanding of all the following is a good start:

  • Sales
  • Finance
  • Marketing
  • Product
  • Communication
  • Psychology

Case Note: Mary Meeker, author of the popular annual Internet Trends reports, is arguably one of the greatest technology analysts of our time and yet her path did not evidently start with any core specialism:

She studied Psychology for her undergraduate degree and then expanded her repertoire by doing an MBA — a mostly generalist business degree — with some specialism in finance. She then went on to work at a number of investment banks before joining the VC fund Kleiner Perkins (she now runs her own $1.25bn fund).

Today, Mary Meeker is seen as an expert on the Internet but I have no doubt that her exposure to psychology, the MBA, and her analyst roles at investment banks made her a better investor. This is why, as a Fortune magazine reporter once put it,

“…she is absolutely first rate when it comes to spotting big-picture trends before they come into focus. She gathers massive amounts of data and assembles it into voluminous reports that, while sometimes rambling and overambitious, are stuffed with a million jumping-off points.”

Further Reading: How to Fail at Almost Everything and Still Win Big.

7. Don’t learn startups. Learn to learn. It’s important that you don’t just spend all your time reading startup advice. Go a step further and first learn how to learn. Sarah Tavel of Benchmark nicely sums up why this is important on her blog:

“In VC, you’re constantly ramping up in a new area. Each company you evaluate brings with it its own ecosystem that you need to understand. Similarly, trends in the tech ecosystem turnover so quickly, that if you ever stop adapting and learning, you’ll quickly become a dinosaur and won’t know a Snapchat when you meet one. That drive to constantly learn will help you adapt to new environments and challenges.”

So work on strengthening your learning capacity. This will help you pick up new ideas and concepts with speed. Some of the best investors have this down pat. Like investigative journalists, they know how to traverse a new area, speak to several experts, and distil the essence of what was previously unknown in order to make an intelligent investment decision.

Case Note: A brief personal story is worth sharing here. By the time I became a venture capitalist, I’d exercised my learning muscles to a point where I was comfortable (and in fact quite relished) the challenge of having to learn new things often and with efficiency.

I did this first through hobbies: learning to skateboard, learning to play basketball, learning to make music. Then academically: learning how to research, learning how to write. Then in business: starting an online business, learning and reading everything I could find on entrepreneurship, and doing an MBA.

With each learning journey, I picked up key ‘learning’ lessons that I continue to use to this day. No doubt I still have someway to go in getting better at this, but by learning lots of new things prior to my VC career, I found the transition to a fast-moving industry less jarring.

Further Reading: Ten Simple Rules for Lifelong Learning, According to Hamming.

8. Don’t be a critic. Cheerlead and build. Being a critic is easy. In fact it’s more tempting to be a critic when you’ve never built anything of substance yourself. You may also find it harder to relate to entrepreneurs who are building hard things. In addition, being more of a critic than a builder makes it difficult to engage your creativity, which you especially need when assessing novel ideas.

To get around this, spend more time with entrepreneurs and look for ways to support and act as cheerleaders for them (see Prescription 1 again.)

Secondly, work on building something new yourself. It doesn’t have to be a business by the way (I build books). But it should be something new, and with the potential to be of value to others.

Naturally, this thing will also be vulnerable to criticism and rejection, something entrepreneurs have to face daily. But by building something new, you’ll flex your creative muscles, which will make you stand out from all other aspiring VC candidates.

Case Note: June Angelides is an investor at Samos VC but before that, she built something of significance outside of her day job at Silicon Valley Bank.

While on maternity leave, June founded “Mums in Technology”, a child-friendly coding school that taught over 250 new mothers how to code and in an environment where they could bring their children along.

June built “Mums in Technology” out of a genuine need to fill a gap she spotted and felt a deep care for. She didn’t build it to get a VC job but incidentally, her experience of building things no doubt contributed to her being recruited by one of the UK’s leading VC funds.

Further Reading: Originals: How Non-Conformists Move the World.

9. Don’t covet venture capital. Learn its downsides. Like any other job, there are things that investors don’t like about their work. Yet, the best candidates for the profession—and probably those who will have a longer-term career in the industry — take time to understand and accept these challenges.

What makes being a VC hard?

For one, it’s very difficult to make money as an early-stage investor. And then there’s a psychological toll to pay: saying “no” to most of the founders you meet each week can be dispiriting; working through several company failures is anything but simple; and, as Max Levchin (co-founder of Paypal and a serial entrepreneur) puts it, “you have to work relentlessly for ten years just to determine if you’re any good at the job.” (Note: there are some ways of getting shorter feedback cycles.)

Furthermore, starting out as Associate can be unstructured (unlike other professional careers) and all-consuming. However, if you can be cognisant of the downsides and your enthusiasm for a career in the sector remains firm, VC could be the right profession for you.

Case Note: Max Levchin is a well-known serial entrepreneur. He co-founded Paypal, Slide (sold to Google), and at the time of writing is the CEO of Affirm.

Like many successful entrepreneurs before him, Levchin considered a career in a VC after founding and exiting his businesses. But it wasn’t long before he realised that VC wasn’t for him. Here’s an excerpt from an interview the journalist Gary Rivlin conducted with him:

“…Levchin recognized he wasn’t venture capital material. He had practically birthed Yelp in his offices in the mid-2000s and provided the online rating service with its early seed money — and then felt he was the “helicopter parent” that no entrepreneurial venture needs early in its life.

“I remember very distinctly a moment when I thought, ‘Why don’t you get out the way and let me drive this car now?’ ” he confessed. “I was becoming the investor no entrepreneur wants: the guy everyone is wishing would stop with advice and unwanted help.”

Levchin had a similar experience after providing seed money to the founders of Pinterest, the digital scrapbook site. “I’m self-aware enough to know I’m clearly too tactile to be a venture capitalist,” Levchin said. “I really want to be the one doing the building and not offering advice from the sideline.””

Further Reading: Becoming a Venture Capitalist.

10. Don’t pursue a VC job. Plant seeds of possibilities. Victor Frankl put it best in his timeless classic, “Mans Search for Meaning”. In writing about happiness and doing well in life, he notes:

“Don’t aim at success — the more you aim at it and make it a target, the more you are going to miss it. For success, like happiness, cannot be pursued; it must ensue, and it only does so as the unintended side-effect of one’s personal dedication to a cause greater than oneself.”

I appreciate this quote has whiffs of impractical idealism. Yet, there’s something in it for aspiring VCs too. Indeed, those who have the best prospects of getting a job in the sector often aren’t the ones directly applying for VC roles. They are busy doing things—many of which I’ve shared in this blog post—that make it more likely that an opportunity in VC will ensue.

So don’t try to pursue VC jobs directly. Instead, help people in the tech ecosystem, build genuine relationships, do interesting things that pique your curiosity, share your ideas widely, build foundational competences in key business functions, learn how to learn, and build something with some significance.

Doing these things might not necessarily get you a VC job, but they will most certainly plant seeds of possibilities that will open doors to a wide range of fulfilling careers.

Some Parting Words: VC attracts people who are curious about business and technology; people who thrive in varied and dynamic environments where the learning never stops; and people who, fundamentally, enjoy working with other ambitious people.

But VC isn’t the only career choice available to such people. These same themes — business, technology, learning, ambition — can be found in other jobs too. So as you cultivate what it takes to be a VC, I would encourage you to keep an open mind to non-VC opportunities where your talents could be just as fruitful.

Further Watching: Why Greatness Cannot Be Planned: The Myth of the Objective

Special thanks my venture friends, Check, Jayanth, Kathy, and Matt, who read an early draft of this post. And shout out to Albert Wenger, who helped guide my early thinking on a venture capital career and purpose.

How Computers Work – Part 1: “Bits”

Source: Safar Safarov (@codestorm / Unsplash)

Source: Safar Safarov (@codestorm / Unsplash)

Have you ever wondered how computers work? What’s really going on in your smartphone, tablet, or personal computer? What are the high-level concepts and ideas that power one of humanity’s greatest invention?

As a venture capitalist I often spend time thinking about technology trends and entrepreneurship but I’ve always felt like a bit of an imposter not knowing the basics of how a computer actually works.

This series is my attempt to grasp the basic principles of modern computation. Through these blog posts, I will share notes that may someday be turned into a short PDF for non-technical people. But for now, consider this a work-in-progress that will continue to be revised for some time.

The primary sources I will use in the series include this excellent book, this Youtube series, Wikipedia,  Google, and conversations with friends and colleagues.

I hope you find the series of some use and look forward to collecting feedback and discussing the topics as I go along.

Part 1 starts with the most basic unit of computation: a binary digit or in short, a bit.


Bits

Everything a computer does is based on two basic ingredients. The ingredients come down to just two elementary states: “on” or “off”. We commonly represent these states with 1 for “on” and 0 for “off”. Each unit of 1 or 0 is called a bit.

With enough ones and zeroes, you can have several bits of information–hence the name “bit”1–that in aggregate are able to express substantial complexity. This simple binary system is the foundation upon which all modern digital computation is done.

Using a binary system to represent more complicated information predates modern computers. For example in the 1800s, the navy used a signal lamp for flashes of light and intervals of darkness (the modern equivalent of bits of 1s and 0s) to communicate with other ships.2 Then of course there’s also morse code, which uses dashes and dots for messaging.

In today’s computers, instead of using flash lights to convey and manipulate information, we use the flow of electricity (or lack of it) at a microscopic level to represent 1s and 0s. If electricity is flowing in area on a computer chip, that is a 1. If there is no electricity flow in an area, that is a 0.

Electricity flow on computer chips is controlled by millions of tiny switches called transistors (more on this in the footnotes).3 These switches are a bit like a light bulb switch, except that in computers, the switches turn on and off really fast and they operate at a scale much thinner than a piece of hair.

My iPhone XS, for example, has 6.9 billion transistors on its A12 bionic chip and many modern computer chips pack more than 100 million of these microscopic switches per square millimetre.

Illustrative scale of the A12 chip in the iPhone XS Max, which has 6.9 billion transistors

Illustrative scale of the A12 chip in the iPhone XS Max, which has 6.9 billion transistors

Thanks to innovations in transistor technology, we have machines with the capacity to manipulate several billion bits. And given enough bits, you can compute almost anything. The table below highlights some of the common terminology you might know from the memory capacity of computers.

Computing Terminology No. of Bits
Byte 8 bits
Kilobyte (KB) 8,192 bits
Megabyte (MB) 8.4 million bits
Gigabyte (GB) 8.6 billion bits
Terabyte (TB) 8.8 trillion bits

How do we use bits exactly? We can encode information with 1s and 0s thanks to a number of international standards. For instance the American Standard Code for Information Interchange (shortened to ASCII) stipulates what numbers, and ultimately what series of bits, correspond to English language characters.

ASCII uses 8 bits (1 byte) for each character. Examples include the lowercase character ‘a’, represented by the number 65 and encoded in binary as ‘0110 0001’. Another example is the uppercase character ‘W’, represented by the number 87 and encoded in binary as ‘0101 0111’.4

caption

Example of binary code to text conversion

Encoding methods were established for other types of data (e.g. picturesaudio and video) and today, we can compute many things with just 1s and 0s. And what’s just as incredible, is that you never have to deal with the unwieldiness of working with billions of 1s and 0s, since that work is abstracted away with special hardware, encoding standards, and smart software that can turn things in the world into bits.

This ‘abstraction’ truly is the beauty of computer science. Once a low-level problem like turning an ‘a’ into ‘0110 001’ is solved–thanks to transistors, specialised electronic circuits and software–you and I never have to worry about that process again.

That complexity is essentially abstracted away and you can write other types of higher level software on top of it, such as a word processor, which in turn can empower someone to type up a best-selling novel with a computer without ever having to understand how bits work.

Example of higher level work, which is the result of lower level complexity being abstracted away

Example of higher level work, which is the result of lower level complexity being abstracted away

This series of “how a computer works” will move up other related ladders of abstraction in future parts.

This week we have started with bits. But in the next part, we will move up to what engineers call logic gates. They are a combination of transistors that collectively take multiple input bits of 1s and 0s and output a different bit. This is an important feature for arithmetic computation.


Notes

[1]

The term “bit” was reportedly first used in a computational context in 1936 by the American engineer Vannevar Bush and in a paper titled “Instrumental Analysis“, published by the Bulletin of the American Mathematical Society.

[2]

One such system was invented by Philip Howard Colomb, a Royal Navy officer and inventor. He patented the Flashing Light Signal system in 1862, which you can read about in this newspaper from the times.

[3]

Transistors used to be bulky, discrete components that you would have to laboriously wire up to make more complicated circuits and computers. This website has a good summary of innovation in this area. But in short, control of electricity was once achieved with vacuum tube technology (one bit required the space the size of a thumb), then transistors came along (one bit could fit a fingernail); then integrated circuits (thousands of bits in the space of a hand); and more recently, silicon-based computer chips (many millions of bits in the space of a finger nail.) This video explains the a modern approach to transistors.

[4]

You can experiment with binary to text code here. A full table of ASCII codes is available here.

Why Success Formulas Fail

This is a sample of a monthly book recommendation newsletter I send to almost 200 readers each month. You can sign up for it here.

formula

In February I spent time thinking about “hits”. Only a tiny fraction of films, out of hundreds released each year, generate blockbuster returns. Only a fraction of books, out of hundreds of thousands published each year, turn into bestsellers. And in the world of venture capital, where I spend most of my time, a very small percentage of startups generate returns in excess of 5 to 10 times the invested capital.

Creating “hits” is elusive and no matter what you read or hear, there will never be a guaranteed formula for success. Why? A passage from a book I read earlier this year hints at the answer:

Formulas are useful when the underlying variables don’t change. For instance, the chemical formula for salt is a one-to-one ratio of sodium and chloride ions, whose elemental properties are fixed. But let’s say that you believe—as I do—that the fundamental variables of hits include familiarity, surprise, emotional resonance, and distribution strategy. These are not fixed variables. The property of familiarity, for example, changes every year. The sound of guitar-led rock ’n’ roll in the United States was considered weird in 1950, mainstream by 1960, and old-fashioned by 1970. In our desperate search for simplicity, people want success to work like a garage door opener, where a four-number code springs the lock. But culture is not a keypad, and people are not doors. Our codes are ever-changing in reaction to our environment. If you can imagine a keypad that gets bored of the overuse of certain numbers—or a garage that changes its code subtly to mimic the numbers from neighboring households—then you are beginning to think like the hit makers I spoke to in this book. I can’t offer a universal formula in these pages. But I can offer subtler and truer lessons about why we like what we like, and how culture works to give people a better chance at making something that moves others.” – Derek Thompson (Hit Makers: How Things Become Popular)

For a while now I’ve been trying to piece together my thinking on this, reflecting on how we can tilt outcomes to our favour in dynamic and complicated systems. Thompson’s book is a great place to start, especially if you’ve ever wondered how things become popular. That said, the book is also a fantastic reminder of a blunt fact: the pursuit of a universal formula for outsized success (especially in a probabilistic world) is a wild-goose chase. As Thompson rightly points out:

There is no complete and perfect formula for building a popular product. If there were, everybody would know and follow it, and the world would be awash in similarly successful cultural products, which technically means nothing would be very successful.

I highly recommend Thompson’s book and hope you will enjoy it as much as I did.

 

Privacy isn’t Dead but Facebook Might Be

pexels-photo-270514

Some say privacy is dead. For a while, I thought so too. In an age where more than 90% of all the world’s data was created in the last 2 years and where we are increasingly and more easily sharing lots about ourselves and friends, the easy thing to do would be to completely give up on the idea of having strong controls over our data — particularly if giving up control means that you can access lots of free products like Facebook, Gmail, and Snapchat.

However, if it is the case that the world’s most valuable resource is no longer oil, but data, and if it is also the case that the volume of data we are all creating is growing at an accelerating pace (for example connected cars may produce up to 25 gigabytes per hour), then a case must be made for being more responsible about how this data is stored, used, and shared.

Facebook is having a difficult time right now because information it held on 50 million users was compromised through overly permissive and negligent controls. Earlier this week I wrote that everyone loves an underdog ‘til they make it. Well, Facebook is no longer a dorm-room startup. It made it big time. But at this stage it’s hard to love the business because it has failed to become more responsible despite the gargantuan amount of personal data it holds on 2 billion people.

Personally I am long on privacy. It’s why I find startups like Hazy compelling and why I use Telegram as much as Whats App. Indeed, if you are founder working on a startup that takes privacy seriously feel free to contact me.

We are creating more data than ever yet the pace at which we are becoming more responsible with it is far behind. This is why we are seeing so many data breaches. In my eyes privacy isn’t dead. But companies like Facebook just might be.