For me, that'd be a multiplayer war sandbox. Take a big map, plop down resources, population centers and factories, spawn a lot of players who then have to organize to fight a war.

• To add espionage, you'd not hard-code a lot about the way teams organize(think removing the guild mechanics from WoW), so one could sign up with one side, then run to a different side (who do not know you already signed up with the first) and do ye olde double agent.
• Research could be modelled as being done by NPC civilians, which produces documents detailing the results. These documents must be on-site whenever the research is being used (i.e. present at factories), so it could be stolen.
• No need to go factorio on the production chains. Their purpose is mostly to enforce cooperation to reach better hardware for the guys in the field.

So, what's your crazy dream game that's never going to be built?

I'm curious if we have any game devs on Tildes, either professional or amateur. If so, share your games, experiences, or advice for any aspiring developers.

I briefly dabbled with game development in the past, which amounted to a goofy helicopter combat game made from the Ogre framework. I've been trying to get it running again, and it's inspiring me to get back into hobbyist game development.

Alex Bruce gave a pair of talks at GDC 2014 on the development process of Antichamber (trailer)

In the first talk he talked about the journey from Unreal mod to full game, presenting at game festivals, networking, the grueling work of getting the game to market, and the lessons he learned along the way:

Antichamber: An Overnight Success, Seven Years In The Making

In the second talk he went over the iterative design process, and how he tweaked things based on how people interacted with the game (spoilers):

Antichamber: Three Years of Hardcore Iteration