Subsection: 15 May 2101 Up Section: Mercury

29 May 2101

We arrive into elliptical Mercury orbit (modulo a pellet elevator throwing warnings) and scan the poles on our first pass. There are three undocumented objects in orbit; the one with a transponder is tagged as “derelict”, and the traffic control station identifies them as the remains of a putative comsat network that failed to survive in local conditions.
(14 October 2020)
We establish resupply arrangements with Gustave Lallier, and I am invited aboard the EU flagship here by Commodore Penn. Mid Addams brings us in neatly, and I shuttle across, with Patel, Jane, Addams and Stewart visiting the station.
It seems that the Chinese are reporting incidents of sabotage, roughly one per three weeks, without details; there have also been four in other facilities, over the last 14 months or so. Information about attacks on EU facilities should be fairly reliable: one was a crawler transporting a devourer swarm (all open source designs and TSA software, which means nothing in itself). The other was a bioroid copy of someone who liked surface rambles, and who is still missing.
There’s little interconnection between settlements, and no general Mercury web or positioning system.
There’s generally a physical component to the espionage: physical bugs placed near wireless networks, that sort of thing. Chinese and Russian sources claim capture and execution of human spies, but without detail.
Commodore Penn is interested to see whether Lidar comparisons might show up slow small movements on the surface. His working theory is a hidden base somewhere near the south pole.
We move over to Lallier and make use of their park facilities, then join the Commodore for dinner before returning to Alacrity.
We set up a plausible mapping orbit. Of course, that path is public information, so we also set up a RATS and a sabotage drone with separate orbits to run passive sensors over the ground before our planned passes.
Initial comparisons show a remarkably high amount of surface activity from the Brazilian contingent. Nothing immediately useful for comms interception, except that they’re heavy on soft humans compared with infomorphs. The US group is surveying, probably working on an ice lode.
There’s activity in the same crater as the Chinese base, though some way off and up the crater well, not moving though sometimes visible during the Lidar passes, spotted moving on the offboard passive sensors at other times. There’s some traffic between this and another site on the wall.
There’s nothing as obvious near the north pole, though we’re able to map some traffic patterns between main and outlying domes.
After several days we establish that there’s regular traffic from the crater wall site to the Chinese base, smaller than human based on where it hides during the Lidar passes. But we can’t confirm a directionality. Eddie and Zaphod put together a higher-resolution passive sensor.
With higher resolution, the movers are small tracked robots. At the second crater wall site there is assembly going on, hidden from the base by a rock wall. Jane determines that this is a sensor array of some type, with foundations sunk into the rock. Clearly this will be passive, and will be able to track anything visible in the sky – which in practice means anything in a polar orbit, the usual path for spacecraft and antimatter factory satellites.
The robots appear to be taking steps to remove obvious tracks, brushing dust and varying routes, and never getting closer to the base than about 1.5km. The first site appears to have underground facilities rather than anything on the surface.
We consider possibilities for a lower pass for our passive observer. We do indeed get a closer look; the robots are working on the sensor array, which appears to be designed to be active and phased-array. This will clearly be detectable from orbit while in operation; I wonder whether it might be close enough to the base to appear to be a Chinese project?
(11 November 2020)
We start considering sabotage elsewhere. Information on the sabotage in the US base is very limited, but we understand it was an attempt at life support, swarms attacking external radiators. And we have no information at all about the Russian incident.
We send a summary to Earth, and I report in person to the Commodore, where we consider the options:
We’ll do more traffic and movement analysis on the north polar sites, looking for something similar – washing out the official information about movements, and seeing what’s left. We can pick up suit transponders with the passive ECM array, which helps matters.
After a few days, we have something interesting: what might reasonably be the same sort of small stealthy bot, operating at the collapsed wall intersection of the EU and Russo-Brazilian craters. The hub, or hubs, seem to be concealed somewhere in rough terrain.
Another pass with the passive sensor seems indicated, but there’s a larger target footprint. Instead, we put a jump RATS on the surface to approach stealthily, with real-time laser link to the ship in orbit. That gets a look at one of the surface robots: brushes, caterpillar tracks, no obvious sensors or manipulators.
We follow. It moves along the crater wall, then turns into the EU side of the crater and rummages around briefly in a junk pile using manipulators stored behind the hatch, not obviously adding or removing anything. It moves back along its previous course; when it meets another robot it makes brief arm contact with it, then each continues as it was before. Eventually our target reaches something like a cave entrance (“point N”), just about large enough for a crawling human and definitely tight for a RATS.
The RATS conceals itself. Two hours later, a robot of similar design comes out, quite probably the same one, and leaves on what might be the same route. Half an hour after that another one emerges…
After 24 hours of observation we have six distinct robots operating through point N, taking between two and eight hours in the cave and dispersing in different directions.
We put another RATS down to watch the junk pile. It looks as though there’s roughly one visit per 24 hours.
We deploy a snakebot into the cave, and remote cameras around the area. When the snakebot returns, we have a curving tunnel into a large chamber containing a fabricator, material stocks, a power network and some inactive robots. The fabricator’s generic, but the power system looks Russian.
Some of the robots pick up the output of the fabricator; it’s not entirely clear, but we get the impression that they’re working with explosives. They also load spools of cable (possibly for power) and swarm hives.
This starts to look like potentially quite large-scale sabotage. We don’t know where the explosives are being planted, and we’d really like to. So it’s time for another snakebot probe, this time with instructions to come out when it sees explosives being loaded.
The one goes for a mile or so and enters a small cave, which turns out to be an explosives store.
The RATS units are set to trail the robots and establish other active locations; they add four more to the tally. Altogether, we have four explosives dumps, one site for cable, one for swarm hives.
We report again. The solar arrays and power feeds seem like a good target for an attack with explosives. That’s something we’ll need to tell the civilian authorities about…
(25 November 2020)
Orders are sent back from Earth, suggesting seismic probing. We will run Royal Marine exercises separately on both sides of the area, using the “inactive” monitoring stations on the other side; given the location, it seems polite to invite the Russians to cooperate with us, so we have an attack/defence setup.
Patel and Jane drop in the shuttle to set up the umpire/monitoring stations. The exercise scenario is an initial orbital bombardment to cover the advance to the first objective, followed by further advances on foot. Jane takes up a spot with a line of sight to point N. Stewart is on guns, Addams at the helm.
The initial stealthy advance is less so that one might like, but sufficient RATS units “survive” to take the objective point. Russian counterfire “explodes” some of the expendable tracked drones, and the RATS are stopped at the edge of the Russian dome.
On the way back, Jane spots a subject robot trying to hide in shadows, and off its normal route; she tries to look as though she hasn’t noticed.
Analysing the data: there are definitely caves (which shouldn’t be here at all) near point N, leading towards the crater wall solar arrays. There’s also an area unconnected to tunnels, showing as partly filled in with something solid; perhaps a 5m cube, under 3m of cover, near the edge of a crater but not in the wall.
There’s no sign of communication from the subject entities in response to the exercise.
If the tunnels were to be used to sabotage solar arrays, the EU base would be most affected.
We cross-check various touristic points to look for a plausible excuse for someone to walk across the cube. That won’t happen for a few days. Soft-landing a swarm seems technically challenging.
Addams reckons that the power loss is being hidden in the natural decay of solar arrays over time. There’s not a lot of data on the Mercury environment, though I dig out a paper and consequent correspondence (and check with the antimatter facility panels, which are built to the same process), which can be calculated back to a divergence point something like two years ago.
The woman replaced by a bioroid was known for walking on the surface; she would have had access to significant research and the ears of important people. She was caught planting malware, which seems an unusually wasteful use of a major asset.
We report on the state of play.
(9 December 2020)
And hear back from Admiral Wreford. Other assets are going into the cube; we concentrate on the sabotage and bots.
It seems to be worth trying to make contact with the Chinese internal security services, but we have no personal connections. One of the planetologists at the EU base, Henri Gavascon, knows one of the Chinese ditto, which isn’t much but it might be a start. I take Addams to meet him in person, and when I confirm that this might be connected with what happened to Maria he agrees to make the introduction to Chiang Lung. He is concerned, and thinks he might be able to produce a more cooperative interaction than a formal approach, though of course it will need a visit in person.
At the Chiense base, Chiang Lung introduces me to Yui Chen, who certainly has the air of an experienced operative, and I share the relevant data. The next day I am privileged to observe a ground exercise by Chinese RATS-equivalents, following seismic surveys. A target crawler slags itself when approached too closely. The antenna construct has no active components: it’s all locally-synthesised plastic. The Chinese operators disassemble and recover it, as well as collecting the target crawlers after the self-destruct.
During a slow spot in the operation I point out the solar panel decay curves – though this base isn’t particularly close to any solar arrays. Following thermal trace leads to a small fission plant and a decent-sized fabricator, rather more hardware than anyone’s happy to see here. The infiltrators do get a decent amount of information before there’s a full shutdown and wipe, and they use this to try to get into a machine that hasn’t yet been alerted – though unsuccessfully.
The initial package seems to have been reactor, fabber, and a few specialised units; it starts to look as though some of the Chinese sensor systems may have been subverted. Construction here concentrated on swarms (devourer, disassembler, construction, etc.) and hives, and crawlers to carry them. The purpose of all this is still unclear, and it’s a major investment of resources. Fomenting distrust is possible, but this is a very expensive way of doing it.
We return to the ship with a bottle of bijiu. Next day there’s news: a party of “prospectors” has found a “hidden base”, the five-metre cube, containing what appears to be Maria and an unidentified second person in stasis pods, and some long-wrecked bioroid-construction and HyMRI equipment. The operatives in question are proceeding with extreme caution.
On analysis it looks as though there were some tweaks from the TSA-standard specification, which may account for the wrecking, which goes some way to explain why it might have been worth doing. The second body is proving hard to identify, but it looks as if it hasn’t been used; it is a bioroid, though its computer brain appears to be blank of personality. Releasing the appearance gets a reaction from the Brazilians; it’s a copy of José Manuelo, a programmer on their base who specialises in life support sysadmin. But he isn’t inclined to go out on his own… and the original is still in the Brazilian base.
It seems that the changes to the machines are set up to do a better job of conjoining bioroid parts; the manufacturers are impressed and interested.
This feels like the same kind of pattern we’ve seen in some other incidents.
We start to consider the possibility of a low-impact intrusion: a small unit which could get aboard a crawler and make an internal connection.
(6 January 2020)
The plan is for something like a surveillance worm to get aboard a crawler, make a copy of its memory, and then drop off, wait, and call for pickup; this should avoid obvious high-bandwidth communications. Eddie and Stewart work on customising a robot chassis.
We drop several off and wait for one to report in. The crawler has rather more memory than expected… and the worm wasn’t able to understand its operating system. Looks oddly custom. And the hardware of a crawler that’s self-destructed, with a bit of close inspection, isn’t generic TSA either… someone’s been applying software-type optimisation techniques to the hardware, making it less generic but better at the specific jobs it’s doing here: wiring is incorporated into structural panels, and so on.
More robots report back and we start to assemble the AI data model. There’s some similarly interesting customisations here… and it’s been sanitised of tooling marks, which is odd given how distinctive it is.
The whole thing gives the air of having been customised or written from scratch, from someone with relatively little access to e.g. standard parts catalogues but with substantial mental resources used to compensate. That’s very unlike current TSA operations, always our first guess.
Our evaluation is circling closer to “emergent AI”. Where its processing may be going on is another question.
One project is to try to emulate the robot’s NAI mind on general-purpose hardware, with a virtual environment and debug views and breakpoints and things. This isn’t perfect but can start to give us a window into what is likely to trip the self-destructs.
We also start looking at patterns of landings and vehicular movements in the relevant area – and, alas, patterns of impacts too. There’s a remarkably low-energy impact event, so we send a RATS to take a look, to find what might well be trace from a braking rocket and soft impact. Tracking data aren’t great, but it looks like something out of the main belt that took about two years to get here… and arrived about six months before the solar panel degradation accelerated. Solar flare observatories didn’t spot anything interesting at the time, though there was one witness report not regarded as significant.
Command reports an increase in network attacks, starting a few hours after the discovery of the nanostasis vault.
There are some candidate impacts near the Chinese base, 12-18 months ago. This time there’s a gamma flare for one of them; their positional sources are different from each other and from the one we’re looking at. We pass this information on to Yui Chen – who confirms obliquely that these sites are also somewhat radioactive; and that they will be happy to assist in the destruction of whoever it is.
The original Maria is revived; her last memory is of walking on the surface perhaps half a mile from the nanostasis vault, so it looks as though she hasn’t lost a lot of time. Physical evidence is consistent with the shadowing. We now have a solid date for her replacement, and the bioroid doesn’t seem to have made any major decisions differently from what she might have done. When her DNA was taken… it looks like a visit to Lallier for medical treatment, and the medic had a misbehaving LAI that was then restored from backup.