| IME, here are some signals that a company actually values correctness. This is not all-inclusive, nor is any one of them a guarantee. * Their codebase is written in something relatively obscure, like Elixir or Haskell. * They're an infrastructure [0] or monitoring provider. * They're running their code on VMs, and have a sane instantiation and deployment process. * They use Foreign Key Constraints in their RDBMS, and can explain and defend their chosen normalization level. * They're running their own servers in a colo or self-owned datacenter. And here are some anti-signals. Same disclaimers apply. * Their backend is written in JS / TS (and to a somewhat lesser extent, Python [1]). * They're running on K8s with a bunch of CRDs. * They've posted blog articles about how they solved problems that the industry solved 20 years ago. * They exclusively or nearly exclusively use NoSQL [2]. 0: This is hit or miss; reference the steady decline in uptime from AWS, GitHub, et al. 1: I love Python dearly, and while it can be made excellent, it's a lot easier to make it bad. 2: Modulo places that have a clear need for something like Scylla - use the the right tool for the job, but the right tool is almost never a DocumentDB. |