So you come up with the twist on the fly, there was nothing being set up, it just happens in the situation. Is that correct?
<snip>
If everything has to be initiated by the players, how can you add a twist, wouldn't that also have to come from the players then?
In the play I was describing, in which the GM does not know more than the players about the fiction, yes, things are established in play ("on the fly"). This does not mean that no one thinks about things in advance - just as players may have ideas for things they think would be cool, so I as GM might have ideas for things I think would be cool. Here's an example (from Classic Traveller):
given that Methwit has Wheeled Vehicle skill, and the yacht comes with an ATV, and two of the PCs have vacc suit skill and the party is equipped with five vacc suits, I'm pretty committed to finding some way to get the PCs out in either the corrosive or disease-ridden atmosphere with nothing between them and near-certain death but their ATV and some vacc suits.
Similarly, when I play my PC Thurgon in Burning Wheel, I have ideas about things that I would like to happen.
But these are not
truths in the shared fiction until they come about, via narration and the rules of the game. As it happens, the PCs in Traveller did end up with nothing between them and near-certain death but their vacc suits, but I didn't
know that that would happen until the players declared the requisite actions - to hire a guide, to leave the dome to try and track down their enemies, to leave their ATV to try and approach the enemy installation, and then to storm the installation when their enemies opened fire on them.
As far as adding twists, that is done by having regard to what the players put at stake. When the PCs storm the installation, and one of them tears their vacc suit while wriggling through a narrow space (failed check on Vacc Suit expertise) then
you are exposed to the corrosive atmosphere is a twist that honours what the player has put at stake in playing the PC. (Details of the above are
here.)
The action in
my most recent Torchbearer session unfolded as follows:
*A player had had their Elven Dreamwalker PC try, and fail, to drive the dream spirit of a dead Dwarf out of an Elfstone (failed Abjuration conflict); as a result the PC was haunted by the Dwarven spirit and obsessed by the stone (the GM-established consequence was to apply the rules for a cursed gemstone);
*Later, the PCs confronted some bandits and convinced one of them, Gerda the Dwarf, to leave the bandits and join with the PCs (successful social conflict); the player of the Dwarf PC added Gerda to his list of friends on his PC sheet;
*Later, the PCs were trying to purchase a crowbar from a villager who was reluctant to sell it, and in the end the villager refused to sell (failed Resources test): the twist that I imposed (as per the rules for failure) was that, when the Dreamwalker PC got back to where she was staying (with her mother), she found her Elfstone missing!;
*At that point I made a decision which was secret from the players (this is one example of why I said, upthread, that some - but not all - Torchbearer sessions involve me knowing no more than the players): Gerda had stolen the Elfstone;
*Some sessions later, the PCs learned from the Dreamwalker PC's enemy, Megloss (this enemy had been established by the player as part of PC creation) that Gerda had the Elfstone (successful test on Persuader to persuade Megloss to tell them what he knew about the Elfstone);
*In a subsequent session, the PCs tried to bind an evil spirit (that had been created when the Dreamwalker PC failed to cast a spell (failed Arcanist test; GM-established twist)) into a spellbook that they had found, so that (i) it would stop possessing Megloss's housekeeper Krystal and (ii) would carry the spellbook into the Dreamalker's dreams, so that she could increase her spell knowledge - this Bind conflict failed, and so instead the spirit joined with Megloss, and carried the spellbook into his dreams (this was the GM-authored consequence);
*In our most recent session, the PCs persuaded Megloss to help them confront Gerda (successful social conflict); they stopped Gerda fleeing her apartment (successful Pursue conflict); Gerda then nearly killed the Dreamwalker PC with her spear, but the PC miraculously survived, and was freed from her lust for the Elfstone (conceded Kill conflict, with GM-suggested and player-accepted consequence, and player declaring that the PC "has the will to live"); Megloss then killed Gerda with one of the spells he'd gained from the spellbook (GM-established framing); the Dwarf PC and two other PCs (but not the Dreamwalker) then killed Megloss (successful Kill conflict).
In this sequence of events, there is only one bit of hidden GM knowledege (Gerda's theft of the Elfstone). It is
not used to negate or veto any player action declarations - what it does is establish parameters for both (i) subsequent GM narration of consequences (what does Megloss tell the PCs when they persuade him to tell them what he knows about the Elfstone?) and (ii) framing (the events when the PCs go to Gerda's apartment to confront her (her attempt to flee with the stone; her near-killing of the Dreamwalker PC; Megloss's attempt to take it from her).
And all the stakes and consequences unfold in relation to PC-authored concerns: Elfstones, dreams, Dwarves, Dwarves vs Elves, the befriending of and betrayal by Gerda, the ups and downs of the relationship with Megloss.
This is another illustration of what I regard as relatively high player agency RPGing.
