Well, Brian is firmly in the LISP-style camp of programming which eschews static typing, promotes heavy use of macros, one-man-army style development, excessively interesting code. He also doesn't like automated testing, or any kind of tests. "Why would you plan for failure?"

It is not a scalable attitude to have in the modern software world.

Yeah, there were aspects of it that were utterly genius, and aspects of it that kept me busy for a while.

Yes, at a prior job I tried to follow a bug in his metahtml interpreter many years ago, and the code bottomed out at an unfinished if condition. Actually an unstarted condition. It was just shipped that way. Very difficult to follow such a smart person's code I situations like that.

It depends on what you mean by scale, PG's Viaweb sold for $49 million and it was built completely around this approach. Lisp has a debugger that allows you to edit code while it is running, and allows you to continue half-finished evaluations with the fixed version, so what it means to "fail" in any given circumstance is altered by that meaning.

Keyword might be modern, Viaweb was sold in 1998

During the crazed bubble 1.0 days, and the word is it got re-written in C++ soon afterwards.