Yeah, in the software business we called it “Rapid Prototyping”. The idea was to cobble together enough code to get the right output, for the most part. The classical method, harking back to the days of overnight turnarounds, was to painstakingly code for every conceivable possibility, and then throw every inconceivable form of bad data at it, before releasing it to production.
The only advantage of RP was that it got you a piece of code that would handle the 90-percentile data it would encounter. You could then foist the problem off on the anal-retentive Test&Validation geeks, who, it was hoped, would identify and mitigate the outlier cases. Unfortunately, in nearly all the shops doing RP, they didn’t actually have anyone doing T&V, so by default the code ended up in production. This meant that when an unforeseen combination of factors aligned, there was a very public crash.
NASA used the classical development model, and even then there were notable tragedies. (Frozen O-rings in Florida, really?). It should take a long time before any vehicle designed to the Musk philosophy is allowed to carry humans. Too many unknown unknowns.