However, the author justifies that Clojure was necessary because he only had 4 engineers and didn't have time to scale Ruby, he just wanted something fast because he had to run too many servers.
I don't buy it, unless your developers are free (highly unlikely).
The most expensive part of any tech startup is almost always developer time. AWS is pennies compared to how much it costs to learn a new tool/language or iterate slower. Clojure is amazing, and I think times will change, but Rails + all the libraries are hugely faster for me to execute in than anything Clojure can do.
If I had a choice of paying another $10K to AWS or hiring a $150K/year dev in the short term, guess which one I'm going to pick. You can buy a lot of cpu time for that, and hold on to that $140K.
The thing that kills startups is cash flow, not IO or perf.
Again, I love Clojure. But I have a really hard following this guy's logic because 99% of startups don't need their boxes to go 10x faster, they need $$$ for runway.
1. Ruby was only backend/apis, front-end app was PHP going over to Javascript,
so Rails would have had a limited benefit
2. He specifically said he needed to do a lot of concurrency which can get
awkward to do in Ruby, and is natural in Clojure
IMO this is a use case example where Clojure beats Ruby, even in some rewrites.
More to your point, he didn't hire extras to port over to Clojure, in fact he fired most of the previous team and replaced them with 4 developers. So his point was that he gained BOTH developer productivity and server costs by doing the move, which improved the cash flow.
Enjoyed the article. Agree totally on using the empirical evidence: managers understand numbers and will buy them if they're the right ones presented in a straight-forward way. Not always, but often works well enough.
I'd like to see a follow-up on how they used the genetic algorithms. Always neat seeing how they get applied.
