The three-prongs:
1) The players are not given prompts by the GM - players have no clue what to choose to do, and "anything they want" is not a sufficient answer, as one cannot make an informed choice without information.
2) The GM prompts everything in existence in the sandbox - the players are overwhelmed with choices/information.
3) The GM prompts some manageable sublist of everything - the players end up assuming those are the only things available, and the sandbox reduces to "pick a mission I prepare for you" play.
"Q&A" is also not a sufficient answer, as without information, you cannot frame useful questions. The GM is the player's eyes and ears, and so must prime the pump, and that leads to the three prongs, above.
The question is then how to manage to keep focus on player choice, while avoiding the three failure modes above. What are the techniques used?
This trilemma only applies if the referee’s primary role is to prompt, to steer players toward specific options through suggestion. If, instead, the referee reports the world as it stands, like a battlefield scout relaying conditions, the trilemma becomes irrelevant. Players aren’t choosing from a list; they’re navigating a situation.
In the military, scouts are trained to deliver accurate, concise descriptions to give commanders situational awareness. Tabletop referees can do the same for players. The goal isn’t to nudge or hint, but to convey what the characters perceive, allowing players to make informed choices.
Reports are filtered through the characters’ perspective. In a forest clearing, the referee might describe an old tree with a hollow, but not what could be inside. It’s up to the players to decide and act: “I check the hollow.” In complex scenes, hidden elements may lack obvious clues. The players’ knowledge, prior experience, and current focus shape what is described.
This is contextual. If the group knows rebels operate nearby and use tree hollows to pass messages, then a tree with a hollow is worth describing. If not, and they’re simply passing through, it might not be noted at all, not by fiat, but because the characters wouldn’t give it attention.
Referees can use this principle to avoid spotlighting everything. I liken it to walking through Manhattan. On a busy day, I pass hundreds of people, but I only notice what draws my attention. When I was a child, I noticed everything. As an adult on business, I filtered most of it out.
The same applies at the table. Referees describe what would plausibly catch the characters’ attention, given their context and goals. And just as in real life, not everything that happens is relevant. One day in Brooklyn, while I waited outside a shop, a guy sped by on a skateboard, wearing a cape and DJing from a boom box. It had nothing to do with me, but it happened.
Including such random moments, chosen by judgment or random tables, keeps the world from feeling curated. Not everything described is important. That’s how I avoid the third prong: players don’t assume what’s described is the only thing available, because they know not all of it matters.
Finally, players in my Living World sandbox begin with an Initial Context, a situation report that outlines their characters’ circumstances and what they already know at the start of the campaign.