What to Build Next?

Some learnings on how to innovate and build

As engineers, our thoughts are constantly interrupted by this question and although getting to choose our next projects is itself an immense privilege, sometimes it comes with crippling decision paralysis.

It is important to note that what we choose to work on is fundamentally an investment, our next stock of time.

The first thing to do is — follow the market.

It is a great place to start as you will need money to sustain an endeavour, and it is simpler while riding a wave. You could always take the complicated route and at the extreme, build things for yourself initially, by following your passion. But you will work with lesser leverage then. Either you will have fewer people to help with the task or get little revenue from it. Even worse, you will lose mobility, and things are not always rosy in your area of interest.

The sad part about this is, when it comes to money, winner takes all. So ‘follow the money’ means following the winner, which implies there is only one or maybe a few right answers at best, and we are all in a game of musical chairs. Building the next big thing also requires guessing the vertical among many it will occur in beforehand. But if you want to start something, the world is much kinder.

To create something, one needs opportunities. But if you want to make a fortune, you need scalability. It is immensely useful to develop skills that help you accomplish scale — people and money management. But before that, to start getting returns out of something, you need to be the one placing the flag first, and it requires you to exercise your advantages in the form of domain knowledge. So it isn’t just the same stock price for everyone. Something might be the golden ticket just for you.

This ticket is the expertise amassed over a long period that very few possess in a sector and referred to as specific knowledge. Acquiring it is another form of investment of time, and sometimes requires us to work with particular entities that can be hard to stumble on. What we are trying to do is accrue a form of wealth — that lets you buy those stocks in the first place.

Paul Graham also goes deep into this topic regarding what kind of education one must accrue. There are a few sub-optimal investments, particularly those with moral fashions, such as social sciences, where truth changes with time and speaking out against trends can be catastrophic. And even within popular choices like engineering, it is crucial to steer clear from challenging problems approached with inadequate techniques (AI in the olden days was the prime example, many questions in core engineering do not have exact answers).

But there are a lot of hidden roadblocks in many fields that might be too expensive to figure out by trial. Hence, the concept of a dropout graph, which indicates fields of study as nodes and immediate areas to branch out as neighbours. Math features as a high-degree node, as it very much feeds the pipelines to hard sciences, engineering and economics in later career along with sharpening one’s cognitive skills.

However, pursuing something that faces high attrition is not the best use of time. Most of us wouldn’t sign up for a company if we saw people running off from the fire exits. Math can seem like a highly positive-sum human endeavour as the number of theorems one can conjure up is endless. But to sustain everyone’s passion might not be possible with the funding it attracts, leading to high attrition, and it might be better to sign up for something else in the first place.

Majoring in multiple topics that are in demand and doing it well would place you in good stead to capture that specific understanding that no one else has. Framed in terms of our analogy, it will allow you to buy the best stocks and in plenty. But realize that, with time and given how fast things are moving, we all have to learn new skills. And since acquiring skills takes time, it is better to invest in those that provide returns for long.

What is the ideal ‘range’ one must have? There won’t be a simple answer. The compromise between a diversified skillset and specializing is analogous to the exploration versus exploitation trade-off in reinforcement learning. After all, life is very much non-linear and partially observable. There are high leverage skills that will help you cruise in life, but to catch the ‘tail event’ needed to build things, you need to have a portfolio of interests.

It helps massively to realize fields of study are connected and many of the departments we see today, for example, computer science, have emerged from pre-existing ones, such as mathematics. Similarly, many branches of engineering are interlinked, building upon each other and primarily differ in historical timelines. So, acquiring knowledge in new fields need not be very strenuous once there is a strong foundation in the sciences, which are highly connected.

Most of all it helps to abstractly view knowledge based on the rate of innovation, as that is what matters when we want to build, and it usually parallels trends in funding due to it being a feedback loop.

In summary, acquire broad but unique knowledge and understand how things are connected. Stay away from some pursuits because life is too short. Learn to follow and even better, predict trends. Once you determine an idea, start building!

Blogging about data, systems and ML