It is worse. The code changes are mostly random, only surviving the tests of fitness nature is applying (on various levels though; immediately catastrophic changes on level of cell biology are sorted out). And at least the high-level tests are also random and unreliable.

So basically it's a codebase mostly composed of bugs, and the features mysteriously work because they're based on bugs that happen to be mitigated by other bugs. :)

A billion years of kludges.