I think this criticism is unfair in general, and wrong in particular.
The complexity of implementation of a game should be just as complex as
required by the story and characterisation, and no more. Just because
it is possible to write a sliding-block puzzle in Inform or TADS,
doesn't mean that every game should have some similar piece of complex
machinery. Similarly, just because computers are large enough to store
hundreds of thousands of words of prose, doesn't mean that every game
should have pages and pages of irrelevant descriptive text (which is
very hard to write vividly). It's kinder on the player to just say
"that's not important" than to produce a dull description that
nonetheless has to be read carefully for clues.
When I play "Adventure" today, I don't think, "This game would have been
much improved if the lamp had a wick that had to be cut and adjusted
every 100 turns, or if the nasty little dwarves had Eliza-style natural
language parsers so that `dwarf, why do you throw knives at me' would
produce the response `Is the fact that I throw knives at you the reason
why you are unhappy?'".
If a story can be told well using only objects and rooms, then why not
tell it that way?
In any case, I don't think the "AGT" criticism applies to "The One that
Got Away". You'd be hard put in AGT to implement the commands "put worm
on hook", "tie large rock to line" or "ask bob about <blah>", or to
represent the connectedness of the hook, line and pole.
-- Gareth Rees