There have been a bunch of discussions about language choices and impact on productivity. Wanted to touch upon this here based on my experience. TLDR: Depends

There have been a bunch of discussions about language choices and impact on productivity. Wanted to touch upon this here based on my experience. TLDR: Depends a lot on individual situation. Pick the language you are fastest and most comfortable with. Background: SMS Gupshup - 6 years - Java and some C++ - Built distributed filesystems. Crawled 1B pages with early 2000 hardware. Built map-reduce framework before Hadoop came. Built a search engine on top of it. Built an infra which would send 1B+ text messages with prioritization and monitoring. Social network with 50M+ users. LinkedIn main stack - ~ 1 year - Java LinkedIn mobile stack - ~ 1 year - Javascript - Was one of the early engineers. Founder Startup - 1.5 years - Clojure Dropbox - ~ 1 year - Python Head Technology Tech incubator - 1.5 years - Node.js As you can see my experience has been all over the spectrum. While primarily using one language, I would keep experimenting with other languages like Scala, Go. Following are my learnings and the reasonings: I love Clojure. Love it so much that I feel like quitting everything go to a mountain cabin and just code Clojure. However I would most likely not use Clojure for my startup. The learning curve is really-really steep. It would make it very difficult to me to build a team. It's really fun if you understand it. But first few weeks for a new person might be a nightmare and very emasculating. So even though I really love it, my current strategy is to bring Clojure patterns to other languages. User facing logic I would not use a statically typed Object Oriented language for any user facing logic. User facing logic and behavior tends to be very fluid and I feel that it gets strangled by the constraints of statically typed languages. One you start building things with Class, Inheritance and Polymorphism etc - soon these patterns start driving business logic rather than the other way around. Because of this reason you will see most arcane and slow moving popular sites today are Java based - (LinkedIn, Yahoo, Amazon, Ebay) vs (Facebook, Twitter, Instagram, Pinterest). For this purpose I like using Dynamically typed languages building code in as functional way as possible. Pick Node.js, Python, Ruby - whatever you are comfortable with. In some cases people recommend Statically types languages for catching errors because of type safety etc. However I would solve that problem completely with test suite rather than solving it part of the way using language features while being forced to pick a statically types language. Another big benefit of dynamically types languages is ability to just create a dictionary and start using it. Most CRUD apps work with JSON request and response and I would rather just dynamically pass objects, rather than build a class for the request and response objects for each and every route. This was a nightmare to create and maintain in Scala and Java. Fast developing Infrastructure E.g. Hadoop, Hive, Distributed FS or Queue. 5 years back I would have built the in Java. However I would use Go for this today. However you couldn't go wrong either way. Pick what you are most comfortable with. There is decent amount of talent available. These languages are easy to learn. And you get close to C level performance if you built it right. Development is fast paced. Mission critical performance - File System or Database Most likely C, because you want to squeeze out last of the performance. However be ready for really slow development cycles. And most smart kids out of college wouldn't know C. So you will need to set aside some time till they are really productive.