In 2004 Paul Graham coined the Python Paradox. The paradox, in its most generic form, states that if a company chooses an obscure tech stack, it will have better hiring outcomes, and conversely, if it chooses a well known tech stack, it will have worse outcomes.
This result is surprising on the surface. Normally hiring is analogous to sales: fill the funnel, move down it with phone calls, in person meetings and networking, and at the bottom of it there’s a transaction, in this case a job offer.
As such, if a company has a more common tech stack then the top of the funnel should be bigger. If the top of funnel is bigger and the bottom stays the same, those that make it through should be higher quality. This is a linear relationship.
Unfortunately for our hiring model, the technology industry is predicated on exponential relationships, not linear. People have called it a lot of things (10x Programmer, Power Law, 80/20) but a few inputs determine a lot of output.
The python paradox is a specification of that generality where a few inputs (candidates) determine a lot of the output (developer productivity). Similarly, it isn’t shocking to notice that the venture industry operates the same way where a few inputs (portfolio companies) determine a lot of the output (return on investment). However, what is unusual is that there is no equivalent paradox for venture.
——————————————————————————-
I believe the Python Paradox is real for venture and has its own analogous form: if a startup chooses an obscure tech stack, there’s a higher chance they’ll have a differentiated value proposition, and conversely, if it chooses a well known tech stack, it will have a less differentiated value proposition.
Now, obscure tech is not inherently a good thing to invest in (or hire for). I’m not saying invest in every new cryptocurrency, hire anyone who has made their own LISP dialect and disregard anyone who uses PHP.
I am saying follow the trends and be logical. Rust and TypeScript are soaring in popularity. Functional programming is discussed daily on HackerNews. Startups that use Haskell, Clojure or Scala might be more likely to solve modern distributed systems issues in a scalable way … and companies that are building another marketing automation platform can probably get away with Java …
——————————————————————————-
In the coming months, I want to do a deep dive into commercial applications of programming languages, type theory, category theory and functional programming to show how young companies can leverage these tools. I believe in and enjoy working with programming languages and I think they’re the next heuristic for startups.
Dan-Knight 