Thanx robtillaart.
The Gammon project is very interesting, and is a much more useful way to use 2 ATMéga chip then the amigoJapan project.
I don't know why some people just don't get that yes, it would be more simple to simply use a chip that has more memory, but there would by no challenge or learning benefice. And it would still be limited to max 256kB of instruction. While if it load new command from a SDcard, it would be virtually limited to how much space i have on the card, pretty much like the commodore was limited by length of the audio tape.
My goal is not to make money, not with the hardware has such any way, but it's more to push my student to go over what would be normally possible. I want to make them think outside of the box.
So yeap robtillaart, this reference is what I needed thanx.
but it's more to push my student to go over what would be normally possible.
Unless forbidden by nature's law, anything is possible - OK some things are still quite difficult 
You should learn your students a TRIZ - read the book from Genrich Altshuller, "and suddenly the inventor appeared" -
TRIZ is a Russian creativity method, that moves creativity from art to engineering. It is based upon pattern analysis of the creative sparks in 100.000's of patents (and more). This resulted a.o. in 40 (base) patterns to solve problems. It is not trivial but once you know the patterns you recognize them everywhere.
with respect to the original question: I hope you will post the progress and the final results of your students here.
Well there seem to be lot of interest for the question. I teach in a techno class in high school(III) Computer sc (IV and V). So I'm looking for all sort of projects to bring session after session, when possible never the same ones.
I've decide that Arduino and ATMéga will become my main teaching platform. So I've got to learn as much as possible and very fast. What a challenge, but a great on.
But yeah if we come down to this project on in fall or winter session. I'll bring up progress. For sure.
Thank you by the way for your interest and tolerance, considering that, It's obvious that I'm a french speaking person communicating on a English board.
You know, I prefer, from far, people that are looking for way to do hard things, then people who are looking for reason to excuse why they think it's impossible.
Like you, robtillaart, I strongly believe that if something can be imagine, it can be done. It's just a matter to figure out the way to make it happen.
Everything is easy, if you know how to do it. There was a day less than 100 yrs ago when it was difficult to explain the relativity theories (both) to highly educated professors. Now it is taught in secondary schools
... then people who are looking for reason to excuse why they think it's impossible.
Maybe these people are even more creative
There is this video your students should see - http://www.boxofcrayons.biz/free/movies/eightprinciples/ -
I don't mind doing hard things, but they need a reasonable purpose. 
As for your basic idea, I believe the original Zork games consisted of a small(ish) interpreter with the bulk of the game as data. So if you wanted to have a huge adventure game, certainly you could store it on an SD card in batches, and then the interpreter could load a new batch as you changed "areas".
Just to show this isn't entirely hypothetical, here is one of my early Arduino projects:
At that stage the game was just held in program memory, but knowing what I know now, moving the data onto an SD card would be easy enough.
Well mister Gammon, for all i know this is one of your project http://www.gammon.com.au/forum/?id=11635 - + - Arduino Uno R3 as a true ISP programmer for any AVR - Microcontrollers - Arduino Forum -
So for sure, I was not talking about you. In fact you seem to be the guy who can make anything possible And this is why I was happy you got involved in this post in the first place.
