June 29th, 2026 - Neil Hanak
No A.I. was used in the production of this blog post
It's a beautiful summer afternoon and you are building a chicken coop in your backyard. You have a bunch of nails, some wood, a wrench and a hammer. You want to drive the nails through the wood to make the frame, so you pick up the hammer and get to work.
Why didn't you use the wrench to hit the nails? After all, a wrench is perfectly capable of hammering in some nails. However, intuitively you knew there would be some drawbacks to using the wrench as a hammer. A wrench is made for twisting, and a hammer for striking. This difference in intended use leads to all sorts of different design decisions between the two that make them excellent at what they are meant for, but worse at what they are not.
While you can use a wrench as a hammer (or a hammer as a wrench), that is not where either of them shine. Using a hammer in a situation that requires a hammer and a wrench in a situation that requires a wrench works best. The right tool for the right job makes your life easy. It's the same thing with programming languages. Using the right one in the right situation can make the impossible feel trivial. Using the wrong one can make the trivial feel impossible. Programs are a bit more abstract than hammering in nails, so when it comes to programming languages, how do we choose the right one?
As far as I can tell there are three main metrics when judging what language to use: ecosystem, developer experience, and performance.
When we talk about ecosystem, we're talking about the community that surrounds a given language. Some questions you might ask about ecosystem are:
When we talk about developer experience, we are focused on how the engineers working on a program written in this language feel. Some questions you might ask about developer experience are:
When we talk about performance, we are talking about how will a program written in this language perform as it is running. There are many facets to performance, such as:
As with everything else in engineering, there are tradeoffs in programming language selection. All languages have their own strengths and weaknesses. It really depends on whether or not those strengths make up for the weaknesses for the problem you are trying to solve.
There are many programming languages out there and I can't cover them all, but here's a brief overview of some languages and their tradeoffs.
If you are most worried about having the best possible run time performance, systems programming languages like C++ or Rust are your best bet. These languages are like razors, very easy to make mistakes and cut yourself resulting in a comparatively difficult and slow development experience, but also the most performant when it comes to memory usage and processing power. Video games, high frequency trading, and operating systems are all domains where these languages shine as they place the most importance on raw performance.
If you are worried about maturity, stability, having a large talent pool and massive ecosystem, C# or Java can be great choices. Enterprise software is often made with these languages as they have stood the test of time, and are still flexible and performant enough to handle a wide variety of use cases. Starbucks, Walmart, Intel, and Cisco all use C#. Amazon, Netflix, Meta all use Java. The drawbacks of these languages is that they are not as performant as systems programming languages like C++, and since they are so mature they can be bloated and verbose.
If you want to do anything natively in web browsers, JavaScript (and TypeScript that compiles down to JavaScript) is the defacto standard. If you've used any website with complex user interactions, its very likely that JavaScript was used to make it. The main drawback is performance, JavaScript was designed to be used in the browser, and is not as performant outside of that domain (i.e. on the server) compared to other options.
If you are doing AI and Data Science tasks, Python is a great choice. As it is taught in universities, there is a large ecosystem of people using it for scientific purposes. It's also pretty easy to learn. The drawback is that Python runs quite slowly compared to most other languages.
If you are worried about performance and ease of development in a cloud environment, Go is a great choice. It was specifically made by Google to be easy to learn and fast to run. A drawback is because of its simplicity, Go can feel too verbose at times.
If you are most worried about correctness of your program, you would be well served by Haskell. Some domains where correctness is most important are military, finance, and microchips. Standard Chartered (bank), Palantir (military software contractor), MatX (microchip startup) all use Haskell. The drawback of Haskell is its smaller ecosystem, comparitively unfamiliar syntax, and increased difficulty in optimizing performance compared to other languages.
If you are worried about scalability and fault tolerance, any language that runs on the BEAM (Elixir/Erlang/Gleam) is a great choice. Discord and Slack both use Elixir to handle the massive amount of messages between users each day. The drawback of these languages is poorer performance on tasks that don't need to be scaled and comparitively smaller ecosystems.
The above overview is a heuristic, and there are of course "exceptions". For these "exceptions" the language choice can feel surprising, until you consider tradeoffs that were consciously being made.
Figma, a web application for designing user interfaces, has its user interface written in C++ which is compiled into WebAssembly. This was done because for their use case they wanted the best raw performance that JavaScript could not match.
The video game Minecraft was originally written in Java because thats what the creator was most familiar with and it allowed him to rapidly prototype and "chase the fun", something that would have been more difficult in a language he was unfamiliar with. It was eventually re-written in C++ after the game had become a success and performance became a concern.
Haskell is used at Facebook (a social media web application) to prevent spam. Being an expressive functional language made it easier to create a highly concurrent, spam filtering rule engine compared to using other languages.
There is no one perfect language for all situations. Even for languages that excel in a certain domain, there may be valid reasons to choose an alternative that better matches the problem you are trying to solve. Thinking about what your program places the most importance on and what tradeoffs you are willing to make can make your development experience much easier than it would otherwise be.
