Here, for example, is the implementation of `atom` (I think):

  /(atom /{
    /\[\S\+\]/{
      s/.*/[]/
      bpop_context
    }
    s/.*/t/
    bpop_context
  }

The way you can nest conditions in sed, like /foo/ { /bar/x }, is pretty powerful, and I knew a good bit of sed before I realized you could do that. It's not readily obvious from the manual.

The author has won an IOCCC contest. This would have just pricked him a bit.

If this was (the only available) programming, I'd quit.

Kudos, I think, to the author. His pain tolerance exceeds mine by a vast amount.

> If this was (the only available) programming, I'd quit.

We started with nothing, remember? Just the bare metal. If this were the only available programming, I'd write an interpreter for a better language.

Never imagine that your current suite of tools is all there is.

It will be a challenge to make an interpreter for a better language faster than dead slow.

I would bet on a compiler for a better language instead. That way, only one person would have to wait ages and ages. I doubt the result would be fast, but it wouldn't be insanely, insanely, insanely slow, either.

I would much rather hand assemble machine code than program in sed. I imagine most people have a better understanding of how computer architecture works than how to do arithmetic and logic with sed.