...Oh.

It's hard to overstate how terrifying that is. It's not that wyverns or their riders are invincible, and in fact they often take significant casualties in internicine conflicts where there enemies know full well their strengths and weaknesses. But a blow this fast, and this total? It'd be surprising from an apostle, even if perhaps not quite unprecedented. Almost as one, all the riders lucky enough to have a position in the flights further back are going to turn around.

It's not going to be that easy. The first strike was their best effort, at the closest range from the best firing conditions with the most work done to prepare, but they were hardly engaging at the limits of their range either. They can cut a bloody swathe in the retreating riders as well, and where it's not lethal they can injure them enough to make any further fighting very ill advised. Round one against these creatures was bad, and now that round two is going their way they're going to ensure there isn't a round three.

It doesn't get all of them. Lesser or not, they're still dragons, and a clean sweep like the vanguard would take a similar volume of missiles or a lot of collateral damage. Fortunately for the Sadarans, they have neither, and the dath ilani barrage is not enough to kill most of them.

Unfortunately for the Saderans, it doesn't need to. A dragon rider is not a soft target, but almost all of that comes from their aerial maneuverability and the danger inherent in engaging their mount. Their armor is incredibly expensive, but mostly because of how light it is, and the threats it's supposed to war from them are lucky arrows and a near misses from enemy fliers. When pitted against point blank high explosives, it might as well be wrapping paper, and mortal flesh and bone are vastly insufficient to pick up the slack. The wyvern corps entered the battle a wounded beast, and left it a hollow shell of itself, with the now riderless vestiges following their luckiest counterparts back to what is hopefully some measure of safety. 

(Other wyverns scatter across the city, or are too injured to keep flying, or are so driven with rage that they spend enough time hunting whoever hurt them that a second round of missiles can take them down, but wyverns are very exactingly trained and these fill the minority of cases even now).

The person in command when news of the disaster hits the lines is not the Legate. Rank hath it's privileges, and among them in this case is the privilege of getting to sleep during the night in a real bed safely away from the barbarians. That role falls to Tribune Floros, one of the more senior staff officers, and as much as he normally appreciates the trust and honors that come from being a client of one of the Legate's closest allies, he is not prepared to handle this by himself. The wrong word to one of the surviving, battle-shocked aristocrats in the aftermath could see this turn from a barbarian atrocity to an unmitigated blunder, or worse yet an intentional plan by him or the legate, and the outrage that would evoke doesn't bear thinking of. The legate might survive a blow like that, but the same can't be said for him. His family is equestrians, for Deldort's sake - rich and well connected equestrians, certainly, but he doesn't have half a dozen senators and governors as cousins, and his name isn't on any of the earlier successes of this invasion either. So given that he has neither his commander's standing or charisma to help him avert it, he needs help from the man who does.

A runner to wake the legate, double time like you're being chased by hounds, and tell him it's an emergency. The tribune will be making himself busy in a way that doesn't risk contradicting whatever plan Cattaneo decides on. Starting with standing down troops - they weren't going to deploy the army even if things went perfectly, not with everyone settled into their sleep and how hard it would be to coral them back afterwards, but they did have some night-fighting specialist detachments standing by in case the raid was successful enough to make it worthwhile. That's definitely not happening now.

If you're waking him up in the middle of the night for this it had better be important, or so help him-

Well, that'll teach him to tempt fate. 

The legate allows himself a minute to boggle, in the privacy of his quarters. And then a second minute after. None of his first, second, third, or fourth thoughts are the sort of thing he can afford to project right now, and given the situation that trumps even the need for haste. Then he dons the mask of an unflappable general, and gets to work. They're to move the night fighters into the watch rotation, dispersed to prevent any cluster attacks, and when the sun comes up instead of encamping in place they're to head back for the gate. He's moving up the schedule so the legion won't be spending another night on this side, making them resting close to the front lines a needless risk, and he needs the gateway operating at full capacity come tomorrow to get everyone and everything back through quickly. As for the wyverns and their riders, no expense is to be spared on their treatment, including the intercession of his own personal surgeon. Each one that survives is another family he can sew up into helping cast the affair as a heroic struggle rather than a bitter enemy out for a pound of his flesh, at least if he spins it right, and despite the exhaustion his future self will thank him for putting in the effort to meet with everyone in condition for it. 

And someone get a wizard or a priest or a philosopher on figuring out what the fuck just happened to them, and if it's going to wreak similar havoc on the rest of his army tomorrow.

With the amount of surveilance they're doing on the imperials, it's not actually difficult to piece together what's happening, even through the cultural and language barrier. That's not to say that it's completely clear on the details - especially at first, it's not obvious if they plan to withdraw through the portal entirely or just further reduce their radius - but even the deployment of additional troops to the perimeter doesn't dissuade it once they can see where those are being reassigned from. Prediction markets on how much territotory they will control at various future timestamps start trading down quickly, and the estimates of the casualties to accomplish it trend down; the impact on more speculative future markets is smaller, but of course there's movement there too. Their logistics may be subject to bizarre constraints*, but it's not difficult to miss what the changing tempo and movement of resources to and through the gate means for their future, and from what they've seen it would be an expensive bluff to make.

*As with many parts of the invasion organization, the invaders logistics situation is incredibly baffling. It's not merely a question of being primitive, or even of being inconsistently advanced with dath ilani historical techniques, but indeed missing optimization that any eight year old out of civilization ought to be able to manage from a cold start. It's not impossible they're just not that smart, but it seems incompatible with their other achievements, since heritage optimization just isn't that complicated a project compared to the other humanoid variations they've demonstrated.

That leaves them with the question of what to do going forwards. The null action is of course to continue as they were; the plan if it hadn't been interrupted was a night attack of their own, to test if their (hopefully) superior low-light sensory options and ability to shut out the local light sources at will would allow them to get a decisive engagement with few casualties on either side. It's not a bad plan even now, with the updates to account for the changes in invader strategic posture, and on the margins moderately more likely to succeed with enemy air support as a smaller factor. And aggression has a lot going for it as well - if the invaders retreat safely through the gate and close it behind them, that leaves their only real source of information on how this happened coming from whatever prisoners they've managed to take, most of whom are presumably not specialized in genetic engineering or portal construction and can't be relied upon to have more than a layman's understanding of the subject. If that and whatever corpses and material they capture isn't enough to duplicate the work, in the best case that risks them leaving an unimaginably* enormous amount of value on the table, forever, and in the worst case gets Civilization is destroyed by another attempt having given up their only opportunity to prevent it.

On the other hand, weighing against those uncertain losses and gains are the very real and immediate issues of a lot of people dying, including true deaths. They're not going to underweight the possibilities here, dath ilan works very hard to be about to properly compare counterfactuals and only do time discounting when it's actually warranted, but it remains the case that - presuming that the withdrawal doesn't come to a halt if they reduce the pressure too much - the operation is no longer in service of the objective it was planned to fulfill, and at the very least that ought to adjust the risk-reward calculations here. There's a strong push to scrap the whole affair.

*Used here both in the sense that it's an enormous sum with too many moving parts along too many axes of unknown information to readily visualize, and also in the sense that saying this phrase aloud in dath ilan has the cached response "skill issue."

Are you really going to deontology this, now of all times? Sometimes you have to shut up and multiply, and even with all the uncertainty we both know what the markets are saying here about our expected outcomes if we leave them unimpeded.

Are you really going to give up that easily? I thought this was supposed to be an assembly of the greatest minds in civilization, but the best plan we can come up with for this has dozens of deaths priced in outside of the tail risks? 

I'm not saying it's easy. I'm saying it's our job to shut up and do the impossible.

...The odds of success are only marginally worse if we start it two hours later, accounting for the refinements we can make to the preparation. That's your deadline, then, to come up with a better plan or convince enough people the outcomes are better your way.

Dath ilan makes a science of doing the impossible. Not the logically impossible, which is an distinct word in baseline and wouldn't risk confusing a native speaker, but the apparently impossible, that which seems upon examination as though it cannot be done.

The thing about doing the impossible, and the part that makes achieving it impressive, is that you often can't do it. The fact that something seems impossible to accomplish isn't proof that it is, but it's quite good bayesian evidence that any solution is going to be hard or difficult to see. The potential rewards from some seemingly impossible things is sufficient to still accrue a large outlay of civilization's resources towards solving them, of course, but for every attempt that succeeds there are half a dozen more that turned out on closer inspection to have been accurately assessed even when pitted against a team of genius experts with copious incentives for success - it takes a lot for a dath ilani to declare something impossible in the first place, even in casual conversation. That doesn't mean there aren't any tricks to it.

The ideal method is to cheat. You can pick your battles, take on tasks where you have secret information or advantages, come prepared with more prep than anyone expected you to have done, and end up with a task that may well be very difficult to you and none other. In this as most cases, dath ilan considers cheating to be technique, and if it's sometimes a bit less impressive it makes up for it in ease and effort, but betting on a market that gives incredibly long odds against something happening it and then resolving it by their own actions is one of the archetypical impressive ways for a member of civilization to get rich.

There's probably a number of ways to do that here. The events of the past day have shown that many things they thought they knew about the state of the universe, and possibly the laws of physics that govern it, are wildly wrong, and where your map doesn't match the territory there are usually routes between points you cannot see. Many people are devoting their time to figuring those out, and if they can figure out how to make their own portals or whatever the trick is to this kind of bioengineering or whatever dealt with those helicopters, there's plausibly something applicable here. The problem here is that all of those are already being done by some of the brightest minds in civilization with enormous amounts of resources behind them, and coming in from a relatively cold start it's very unlikely that they'll be able to both pass them by and get it done fast enough to matter. The linguists have a rather simpler problem, in that there's is one that has been extensively prepared for with tens of millions of hours of labor in advance, but whatever simple structures underlay the emergent complexity of alien communication have proven difficult to derive and their example corpus leaves a lot to be desired. Absent a low probability development there, this method won't do them much good in the timeframe in question.

What, then, to do? 

It's a well understood phenomenon in dath ilan that, once something has been accomplished once,accomplishing it again is typically simpler. Not just because they were able to learn from the previous attempts, either - knowing that a problem can be solved, and under what constraints this was done, is itself large amounts of data about what the solution has to look like. If you learn it is possible to build something in 10 hours, you can discard offhand all the search space with longer duration; from these and other details, it is often possible to figure out in broad strokes what the solution must have looked like for it to have been solved. If attempting to solve a problem normally involves first determining what you have, and then what you want, and how to use the former to achieve the latter, this involves looking at what success will look like, and determining what must have happened to achieve it. If something possible seems impossible, then that is not a contradiction, but a sign that somewhere important your map does not match the underlying territory. You need merely assume that a way to win exists, and if necessary reframe success and the material conditions until one pairing matches a way forwards. It can be a dangerous epistemic state to try and stay in for too long, even for a trained inhabitant of civilization. Sometimes there isn't anything wrong with your map, and there isn't actually any limit to how much you can throw away in pursuit of such a strategy if you're wrong enough about how the world works. And it's far from the only method, with dozens of other commonly used reframing schemas existing. It's just good enough to often see use regardless, and this is one of those times.

So what's their goal, then? To remove the invaders from the city without more true deaths?

No, that's both over and under-specified; it demands both too much and too little here. And also missing the point of the deadline. They don't have to solve the problem in a handful of hours, they have to make it seem likely enough it's solvable their way that the timespan is extended.

That's not exactly helpful, though. Saying your goal is to convince someone you can accomplish a task usually still involves you being able to articulate what you're planning to accomplish.

Which is why they're going to hold off on proposing solutions, even for what the goal looks like, until they've actually thought about it for 15 minutes on the clock, and then write down the ideas they have and share them to avoid initial cross contamination. Ideally you'd spend longer than that, but the timeline is short enough to rush things, and better to halfass everything than do this part properly and need to quarterass the rest of it from time overruns.

After the brainstorming and ensuing discussions, a few potential outlined goals take shape across the different subgroups they shuffle between, which eventually synthesizes into one with minority dissent.

"We will know the operation succeeded when neither Civilization or any constituents thereof are under a threat that could be meaningfully ameliorated by attacking the invading troops, and the standard for a flawless success is that no one truly dies accomplishing it."

Put that way, it's clear that the whole business of needing to force the troops back through the gate is a limited one. Not because it wouldn't work at accomplishing those goals, but because once you start actually thinking about it like a riddle it's obvious what other kinds of solutions could work there. They could take the gate itself instead of pushing troops through it, and fortify it. It's much more obvious how to non-lethally keep people from storming a position than to dislodge them from one, though taking the structure without casualties won't be trivial; with fortifications inside and out of the ring, absent additional capabilities there wouldn't be more threats to dislodge. They could seal off the portal without taking it. It would be more difficult, in the face of the demolition techniques the aliens have thus far demonstrated and any more they might keep in reserve, but rendering a place approximately impossible to traverse is not actually beyond their means. They could simply retreat. Losing Schelling Point would be a very expensive blow, but there's a lot of expensive things that are worth doing if it saves enough lives. They've seen how quickly the aliens can take ground when advancing unopposed, and it would not actually be difficult to keep people away from that for weeks while they work on better solutions without even needing to fortify, especially now that the first combats have given everyone still in the city time to escape.

"Or, you know. We've seen what they prioritize when able, above killing or capturing prisoners or almost anything short of immediate strategic objectives, and there's one kind of exchange you don't actually need much of a shared language for. What's stopping us from bribing them?"

That'd be giving in to threats.

That's deeply unclear. It's not obvious what parts of this course of action they expected to benefit them and how it weighs off against costs, much less that they did it specifically expecting to be bribed to leave in a way we don't want to incentivize. Decision theory says you shouldn't let yourself be exploited, sure, but it doesn't mean you're obliged to blow yourself up whenever you're confused. Besides, how much would you have been willing to pay if they'd negotiated opening a portal in advance, conditional on believing it? Ignoring the killing you'd have made on prediction markets from being the only one to know it was going to happen, I mean.

Division of labor, then. One group will get the balls rolling for prep work on these plans, creating the necessary inventory list and filling out documentation and setting up the prediction markets so they can go live promptly, while a second group consisting of the people who aren't necessary for that goes back to think of alternatives or refinements. They're not expecting much out of it, not now that they've already done a whole effortful team planning session, but it remains the case that it's often worth taking unlikely bets when the potential payout is large enough and this qualifies.

The only real refinement ends up being that they can potentially deliver a lot of payloads by ballistic arc if they get the right launchers in place, which is pretty much inherently going to be less accurate than the drones but hedges against a repeat of whatever took down the manned helicopters. Mostly they end up thinking of specifications for possible products that probably don't exist but would solve things if they did, like some kind of distance-deliverable concrete spray to render the gateway impassible; the vaguely plausible ones they take a moment to specify enough that they can post a bounty on it to the net in case someone happened to have it in their back pocket.

And then the third team - or perhaps more properly a half dozen teams, each split up by which plan they're on and who they're talking to - sets out to pitch the emergency services and the public on their ideas. They're in competition to each other to an extent, but for the most part they'd all prefer the other plans win over people instead of status quo ante, and many of the exceptions are still willing to work together if it still means better odds their own plan (here referring to the plan they champion, which everyone involved has done their best to decide for reasons other than who first proposed it) gets traction. In another world they might be worried about accidentally locking things into place with decision paralysis, but Civilization works hard to keep their decision-making infrastructure robust against that sort of issue.

We're not saying no, but we're not saying yes either just yet. It's not that we don't appreciate last minute arrivals with a plan to fix everything that nobody else thought of, we're still human, but this is serious, so let's talk specifics here. How long will each option take, 80% and 99% confidence intervals? What estimated risks are we taking into account, at what rates, for what payouts? Give us some specifics we can use to evaluate tradeoffs once we run prediction markets to double check your numbers.

They've got those prepared ahead of time, of course. (It's facially obvious that you would need those, even six year olds know better than pitch ideas without numbers, though this level of sophistication usually comes later).

An air raid with commandos on the gate structure would take a few hours to arrange, and is not really a novel strategy here except for implementation details, so they can borrow numbers from some of the extant markets as a baseline. They'd be risking the deaths of everyone dropping in, plausibly including their true deaths, and any pilots involved, plus a number of casualties among the defenders if things went wrong, but significantly fewer if things went right; and there are a number of potential problems that could come through a gate that are readily resolvable via people on the ground ready to deal with it and not otherwise. For their preferred version mean projections are about 17 paratroopers dead and four times that in defenders, about 2 and 6 of which are true deaths; the median case is about 7 but 80th percentile version is about two dozen* and 99th percentile is hundreds dead. Here's three other plans elsewhere on the pareto frontier and why they picked this one, presented in case someone has a compelling case on the tradeoff they didn't see. Yes, it's more casualty prone than the already rejected aerial missions it's being compared to, but they're pretty confident it would kill significantly fewer aliens and that's worth it. If Civilization survives this coming year, they're confident it'll be willing to pay the rates it would take to match people's cheerful price for the mission in order to save that many lives; it won't be able to make that trade then unless they can close that deal with their future selves now, which means there's a sizable amount of value up for grabs here. They've got a growing list of investors who agree with them and are willing to buy shares in the value Civilization will later pay out at the following rates, and while it's not enough to fund things yet emergency services won't have to take on all the risk.

*3 groups of 8 in baseline, technically.

Just attempting to do area denial on the gate and its surroundings is significantly lower investment, and could be done inside an hour (from now, not from when they first started getting prepared for it). A lot of the necessary work to make it go well is already in place among the various military forces in Schelling point right now, and there's also less people at risk from the operation. Their median and modal case projections are both at zero dath ilani casualties, and even taking into account the 99th percentile outcomes where it provokes a response the estimated casualties above replacement doesn't pass four. However, it's inherently harder to be careful when you're fighting at that kind of a remove, so the estimated risk profile for the aliens looks roughly similar to an airdrop, and it's significantly less resilient of a strategy. Especially so in the case of unknown unknowns, since having humans at the front will go a long way in terms of providing them with options, but outside of that too. Even if they bring out the really scary stuff like napalm or turn the whole area into an impassable resiny mess or bury it in rubble, it wouldn't keep determined humans out long term without constant upkeep; their counterparts are too capable for anyone to want to rely on their bizarre logistics issues generalizing.

And then bribery is, of course, both the biggest and least risk. The least risk, in that even the worst case scenario has between one and three people in danger and their estimates of any death are low enough that it doesn't even make the 99th percentile estimations. Capture and imprisonment seem possible, of course, but it's not difficult to find qualified volunteers who don't consider that particularly aversive. And perhaps more encouragingly, unlike either of the other three options at hand there aren't any anticipated deaths if things go as planned.

And the most risk, in that it doesn't actually guarantee anything. It's that same fact which so consumed their debates and kept it from being adopted as the consensus platform; as much as it would unarguably be amazing to resolve this without any further conflict, but there's a lot of places it could go wrong in a lot of different ways. They could seem like a trap, and fail to so much as get a meeting. They could be wrong about what their counterparty values, and fail to successfully offer anything, or succeed too much and convince them that attacking dath ilan is more rewarding than previously imagined. They could be entirely mismodeling the alien psychology at work here, or dealing with a scenario where contrary to appearances the decisionmakers behind the invasion are not approximately far mental-distribution humans, and simply fail to communicate the attempt at all. And of course there's the possibility that time doesn't actually favor them. They think it's more likely than not it does, and the markets* seem to agree, but it's far from a sure thing, and they could very well end up frittering away what advantages they have in the conflict beyond their ability to regain. Right now, as bad as the invasion is, it's still looking like one of the absolute best case scenarios for getting attacked by surprise by an enemy with superior capabilities; just because they regularly rehears what to do in the case of alien invasions doesn't mean anyone in dath ilan is *unaware* of what the expected outcome is if that actually happens. It's a reference class that includes a billion deaths as table stakes.

A lot of the worst possibilities would end up that way with most of the other plans, and some might get uniquely disarmed here. But everyone involved would really prefer there was less superheated uncertainty about literally everything.

*Markets of this sort use a somewhat different method than normal prediction markets, since anyone who thought civilization was underrating the chance of their defeat wouldn't expect to collect their winnings. Instead there's a proxy based on the current bond rate, where people who expect the world to end can take out loans to spend now, and repay if their prediction is falsified. It's better than nothing, but there's enough confounders that people tend to have less confidence in the results, especially over very short time horizons.

 ...They're gonna have to think about this, though to preserve option value they'll parallelize as much of the prep work on their end as possible.

When legate Cattaneo wakes up the next morning there are no new fires. Around the ring of his camps, though, there are reports of lots of native soldiers mustering, and fortifications that sprung up overnight in time for the sentries to double take at dawn.