> You build it, you run it For any set of non-trivial apps the cognitive load is simply too big. Assuming the engineer can handle it, eventually more time will

> You build it, you run it For any set of non-trivial apps the cognitive load is simply too big. Assuming the engineer can handle it, eventually more time will be spent in keeping up with the current system(s) than writing code/promoting the business logic. A company could chose to use ready-made tools but all them get expensive pretty quickly and make their own assumptions about workflows, view-points, etc. The assumptions might fit the use case or not. Lock-in then becomes another problem. There's no easy way out, it's trade offs all the way :-) Example: The Haskell lead engineer for App1 has all the necessary knowledge to understand how the base-metal NixOs fleet is configured because there's deep familiarity with Nix. At some point the infra team will start making decisions and assumptions to automate _this_ or _that_. The engineer now will have to spent time understand what changed since last time, making sure nothing is broken while performing changes.