This is a story about how individual bouts of indigestion become a systemic signal when there is room for people to tell their stories. The characters are fictional, the institutions not always. The story is set in a world where the Orbiplex swarm of nodes begins to appear as infrastructure for human agency, not as a gadget.

Four People

Erik, Warsaw

Erik runs a translation agency serving clients from Helsinki to Lisbon. He starts the day with coffee and three independent language clouds, and ends it with a walk with a dog he found a week earlier under an apartment block in Grochow. At home he has three dog beds and one stable ritual: whenever one of his “boarders” is waiting for a permanent home, he takes responsibility for the whole arrangement – food, the vet, registration, and above all making sure the dog does not have to lose, for the second time, the certainty that the world is safe.

He came to the Orbiplex network for a prosaic reason: he wanted translations that would not leak into a provider’s cloud. A friend sent him a description of a local node and a short paragraph: “computation locally, the network as an option, not as a condition”. Erik does not like being anyone’s product, so he installed the node software and began running translation projects with his own AI agent.

Maurice, near Poznan

Maurice designs pumps in a factory outside the city. Officially he is an “innovation guy”; unofficially he is the person who makes sure the director has a Friday presentation that looks as if it came from the future. In his spare time he builds devices in his garage where crystals under high voltage fall into resonances too delicate to see and too intense to ignore. Around the neighborhood he is known as the “mad constructor”, because every now and then half the street briefly loses reception, and the following week some visitor reports a UFO over the forest.

He joined Orbiplex out of technical curiosity. He installed a node, connected a smartwatch as one of the “eyes” of part of the sensorium and, without much deliberation, granted the observation module a narrow list of permissions: reading his own local memarium, receiving summaries from home sensors, and “asking” a second module (Whisper) to release a prepared message into the world if such a need arose. Nothing beyond that. Each of those permissions was explicitly recorded on the node side, so he could revoke it at any time. He did it more like an engineer than like a patient: “let’s see whether anything here can be computed sensibly”. He disabled the mode in which the system publishes rumors by itself, without asking, right away; he preferred the semiautomatic option in which the last word always belongs to a human.

Emilia, a commune near Wroclaw

Emilia is a young composer. She is in her twenties, has a piano with a slightly too lively action, and a notebook. In it she sketches scores after observing birds, whose song is the main inspiration for her work. The wagtail has a different phrase than the tit, and the sparrowhawk has its own silence, which Emilia treats like a grand pause. When she plays recordings of previously registered compositions aloud, birds gather on the windowsills as if they recognized their own dialect in translation.

She came to the Orbiplex network through an artificial intelligence agent acting as a librarian: she wanted a memarium that would keep her field recordings and sketches without passing them to any streaming platform. She likes the phrase “locality as the default mode”.

Agnes, Lodz

Agnes lives in a Lodz tenement house that contains her, her flatmates, and many of her interests, one of which is… the study of insect language. Agnes’s latest hypothesis sounds so unusual that at first glance it seems like a joke, but after a moment of thought it turns into a cautious research question: do the dances of honeybee workers correlate with the quality of cheeses? If she is right, bees and dairy farmers hear the same rhythm of the seasons, each in a different language.

After hours Agnes takes an old VHS camera and walks through discount stores. She records not shopping, but the gestures of people at the shelves: moments in which someone makes a decision, someone hesitates, and someone else gives up. She came to the swarm through a friend from a biosemiotics circle who had installed a node on a video streaming device.

Four Dishes

Each of the people we have briefly met had a day on which, intentionally or by accident, they ate something spicier than usual.

Maurice bought a pepper in a local shop. Orange, the kind “with character”, as the seller put it. He ate it in the evening with a sandwich.

Erik made himself a burrito with chili con carne, because he had received a packet of spices from a client in Mexico and could not resist.

Agnes sprinkled pearl barley with a spice mix her sister had brought her from Vietnam. It was her favorite barley, so dinner promised to be filling and to herald a quiet evening with the tapes.

Emilia ate cabbage rolls in a hot sauce, prepared especially for her by her mother-in-law. To say they were spicy would be saying little. To say that one does not refuse such a dish would be saying everything.

All four of them thought exactly the same thing: “spicy, a bit much, but it will pass”.

The Night When Nothing Passes

The first day looked similar for each of them: cramps, disbelief, tea, negotiations with one’s own stomach. None of them called a doctor, because none wanted to be the person who calls an ambulance over something trivial.

The second day was worse. Agnes fainted in the kitchen, hit her head against the doorframe and slid down to the floor; fortunately her flatmates heard it. Maurice spent the whole workday in the toilet, and in the end gave up and went home without closing the data analysis program. Erik still tried to run meetings with his assistant, but he could not stay on his feet. Emilia, in turn, cancelled a rehearsal and stopped playing.

Each of them weakened at a different moment and in a different place, but came to a similar, quiet conclusion: “this is no longer indigestion”.

Incidents

Maurice and the Swarm Agent

On Maurice’s desk, between a lamp and an oscilloscope, a notification blinked. The local agent connected to the Orbiplex node had been collecting signals since morning from the smart watch and from Maurice himself: pulse, heart rate variability, skin temperature, walking rhythm, frequency of leaving the room, number of skipped tasks in the calendar.

The same observer to which he had earlier granted access to data about his condition noticed that the parameter packet had, for several hours, been growing in a specific direction and diverging from the baseline. On that basis it internally switched its operating mode from passive observation to elevated readiness caused by a threat, and then asked the agent to formulate a single, calm sentence to the user, because in that constellation of signals one short message is more effective than five alerts.

The agent wrote:

Maurice, your vital parameters have been diverging from your baseline for six hours in a way I have not observed before, and which indicates a threat to your health and life. I recommend immediate contact with the emergency number. If you do not initiate a call within twelve minutes, I will prepare to take over and call for help on your behalf. In the background I am now notifying nearby nodes that I am monitoring the situation, without disclosing your identity.

Underneath, in the agent panel, one additional label appeared. On its basis Maurice, as an engineer, immediately understood the seriousness of the situation.

The swarm has an established, short list of emergency event classes: from a single, immediate threat to life (the class operators label simply as TC1), through infrastructure failures, to attacks on specific people. Next to each class on that list there is a separate scale indicating how much trust can be placed in the signal.

On Maurice’s screen it said: TC1 – confidence higher than the default automatic notification threshold. That meant the agent was not playing safe for the sake of peace of mind; simply, what it saw met the criteria under which the system enables the highest permitted level of agent autonomy. In that mode it may act faster, but each of its decisions leaves a full, unedited trace for later verification.

Maurice, who had spent his whole life learning to trust data, read it twice and called himself. At the same time the agent opened a private note with a short communication strategy: what to tell the dispatcher so that the report would not be filed away as “stomach flu, please hydrate”. Three sentences, named in order: duration, intensity, qualitative change. Nothing beyond that. If Maurice had disconnected or stopped responding, the agent had a fallback plan ready: call the emergency number itself, read the address and relevant medical data from memory, and then notify nearby nodes accordingly so they would know what was happening - without revealing to anyone who exactly was on the other side. That plan never entered into action, but it was prepared.

Agnes and the Flatmates

There was no AI agent in Agnes’s apartment in Lodz. There was a creaking floor, panic, and two very worried people who had known Agnes since university. They were the ones who called the ambulance, and each of them spoke a little too fast.

Erik at the Office

At Erik’s company office in Warsaw’s Grochow district, the assistant – a woman who had seen many very different days of his – looked at him and said: “Mr. Erik, you are calling now, or I am calling for you”. Erik called for help himself, because he likes to keep matters under control even when control is slipping away from him.

Emilia in a Taxi

Emilia, who lives with two parrots and one very lazy piano, ordered a taxi to the emergency department and put on whatever clothes lay closest. Within 20 minutes she was there.

First Decisions

Here begins the part that hurts most. Not because it happens violently, but because it happens calmly and very ordinarily.

Maurice: The Argument Has Structure

The ambulance crew that reached Maurice began with a mild admonition: “You could have gone to a doctor yourself, after all”. Maurice, however, already had in his head the structure of the three sentences the agent had given him. He did not start with “it hurts a lot”, but with: “This has been going on for twenty-eight hours; over the last six hours the parameters have diverged like this and like that; I have never had a similar episode before”. The paramedic nodded with a somewhat different expression, looked into the bag with the monitor and said: “All right, we’re going”.

It was one of those moments in which a well-chosen description saves a life before a doctor does. The AI agent, which did not see the encounter itself, continued to monitor Maurice’s condition and was ready to take over the report if the conversation went differently.

Agnes: Noise Versus Content

The crew that came to Agnes was antagonized from the threshold by her flatmates. They were loud from fear, not aggression, but for the paramedics the behavior was similarly burdensome. Words like “panic” and “exaggeration” were spoken, followed by the recommendation: “charcoal for diarrhea, plenty of fluids, doctor tomorrow”. Nobody took anyone away.

Erik: “Ordinary Shits”

Erik, even though he spoke calmly and in full sentences, heard that “there are more important things today than a ride to ordinary shits”. He received two tablets for stomach pain from a younger paramedic, who gave them because he felt a little embarrassed.

Emilia: A Number in the Queue

Emilia reached the emergency department on her own, took a number and sat down on a plastic chair bolted to the floor. The triage nurse glanced at her quickly and asked her to be patient.

A Few Hours Later

The following hours brought each of the protagonists situationally different variants of worsening symptoms.

Erik noticed blood and immediately stopped having any doubts about whether to call the ambulance a second time. This time the director in him woke up: low voice, factual, naming the symptoms, giving the hours.

Agnes fainted for the second time and no longer responded to voice. Her flatmates, now quieter and more concrete, called for help. The ambulance that came was a different one.

Emilia slid off the chair in the emergency department waiting room and lay down on the floor, because the cramps would not let her sit. That was noticed. That was taken seriously – only now, but still.

Maurice, meanwhile, was lying in an observation room. The doctor found the cause in the lower gastrointestinal tract and began treatment earlier, preventing a hemorrhage. The agent recorded de-escalation from elevated readiness back into observational work, but had not yet disabled the alarm mode: caution is a boundary feature, not ornament.

Late Diagnosis

For all four, the diagnosis was very similar: damage in the lower gastrointestinal tract, caused or revealed by a strong food stimulus, non-steroidal drugs, background factors and so on – the configuration varied, the result was the same.

For Erik, minutes counted literally; his surgeon said something like: “another half hour and we would not be talking”. Emilia and Agnes had more margin, but it did not come from the wisdom of the system; it came from someone else’s quick reaction or lucky anatomy.

Maurice, by then, had already had a gastroenterology consultation and was waiting to be discharged. He did what he always does after a technical incident: opened a notebook and wrote down the sequence of events in order.

A Few Weeks Later at the Node

When a person survives something that could have killed them, but did not, they begin looking for a form in which the experience will not be lost. Erik returned to caring for dogs and running the company, but one morning, instead of checking his email inbox, he opened the local agent console.

Agnes returned to the dances of bees, but other gestures began to appear on the tapes: her own.

Emilia played her first concert after the hospital and, for the first time in a long while, wrote something that was not inspired by the speech of birds.

Each of them, independently, on a different day, sat down for a conversation with an agent. Erik formulated statements factually, Agnes like a researcher, Emilia imagistically. Yet all of them spoke about events whose sequence had a similar shape: spicy food, first aid, refusal or trivialization, then several hours of deterioration and a next step they barely made in time.

Instead of offering evasive sympathy, the agent coupled to the swarm node said something like this:

What you describe does not sound like a single accident, but like a pattern I can carry to the community as what we call a rumor, but not as proof. A rumor is weaker than a fact, but stronger than silence.

Before anything goes outside, we will go through redaction together: we will remove your personal data, flatten your idiolect, and leave what matters for the system. We will leave the names of hospitals, dispatch centers and operators; they do not need protection in the same sense in which you need it.

Erik agreed without hesitation. Emilia asked for a second pass, because in the first version she could be “heard” too clearly. Agnes, with her typical scrupulousness, added to the redaction the duration of each phase – not emotions, but parameters.

For each of them, the agent packed the result into a single small package, called in the swarm a whisper-signal. It is worth knowing roughly what is inside, because this is what explains why such a rumor is safer than a post on social media:

• epistemic class – an explicit marker saying “this is a rumor, not proof”, so that no other node can later quietly turn it into a fact;

• polarity – in this case “problem” (systemic harm), not “inspiration” (creative convergence), because the swarm treats these two signal types differently;

• issue class and several context markers – here: refusal of medical transport, severe abdominal pain, voivodeship, week, night shift – enough to recognize a pattern, too little to recognize a person;

• confidence – how strongly the sender believes this really concerns the described matter;

• disclosure scope – to trusted peers, within the federation, or more broadly; for all four it was limited to the federation;

• risk assessment – high here, because the problem touches life and health;

• transport privacy intent – for example, “preferably through intermediaries, and if that cannot be done, I would rather nothing leave”;

• hop limit – how many times the rumor may be forwarded before being stopped.

Added to this is a short-lived pseudonym, the so-called rumor nym, with which each package was signed, plus an attached certificate confirming that this ephemeral pseudonym is temporarily valid. The names of hospitals, dispatch centers and operators stayed in the text, because they were not the ones that needed protection. The original, raw accounts never left the local node; only the agreed version was transmitted into the network.

With Maurice the matter looked a little different. He did not write an account, because the telemetry of the event was already remaining in the non-volatile memory of his computer: smart watch records, agent logs, traces of decisions from the hospital night. The observing module prepared a draft by itself, adding to the package an explicit label: source: monus-derived (the draft came from the local observation module) and additionally monus-sensorium-derived (because a material part of the data was telemetry from the watch). The swarm does not like to hide where a rumor came from. Maurice reviewed it, corrected two sentences, approved it and released it into the world – with the same disclosure scope as the other three.

“You Have a Message”

The rumors went into the network. Not as a shout or a trend, but as small, redacted and pseudonymized social signals encapsulated in a semi-closed exchange between trusted nodes and federations. The nodes that received them could register local interest - a soft signal saying “this may concern us”. When enough such interests accumulated, from a sufficient number of nodes and with sufficient diversity of sources, the swarm recognized that something had crossed a threshold.

The threshold was not a simple sum. To be counted as significant it required many distinct voices from potentially many federations, scattered over time (not during one night), coming from sources of different character – some from a human, others from an observing module, still others from a sensorium. The anti-Sybil filter ignored what looked like a naive multiplication of one person or one bot.

A few days after their conversations with AI agents, all four saw the same blinking icon. After clicking it, a short sentence appeared:

You have an unread message from the swarm.

Underneath was a message that looked like a letter, but had the structure of an artifact:

Based on the redacted rumor you sent, and many other similar signals collected over the past year, the swarm has recognized a systemic pattern: refusal of medical transport or trivialization of abdominal symptoms, leading in a significant number of cases to deterioration within several hours.

This is not a single event. You are one of several hundred people who described something with a very similar structure. We propose opening a dedicated association room. Joining is voluntary.

Erik clicked “enter” that same evening. Agnes did it at dawn, after a night session editing tapes. Emilia waited two days, not because she did not want to, but because she needed those two days for herself. Maurice entered immediately.

The Association Room

The room that received them was neither a forum nor a chat. It was a space with a clear policy: every participant is there under a pseudonym, identities are not revealed automatically, the conversation has a contract, and decision traces are auditable. No person was added silently; everyone gave explicit consent.

The moderating agent first presented a map of cases. Several hundred accounts arranged themselves into a few clear clusters: refusal of transport, incorrect telephone triage, queues in emergency departments that were too long despite red signals, routine attribution of symptoms to stomach infection.

Someone wrote first: “It was the same for me, only with bleeding from the upper tract”. Someone else: “Similar with my grandmother, they did not want to take her, and in the morning she was no longer responsive”. The signal of the “digestive tract” began to resonate with many stories. The room resembled a conversation among several independent delegations: each thread had its own context, its own scope and a spontaneously emerging team.

The agent did not pretend to be a doctor or a lawyer. It did, however, provide a list of escalation paths: from the Patient Rights Ombudsman, through regional branches of the National Health Fund, to internal complaint committees of operators and professional self-governments, and in some – the most obvious and tragic - cases to prosecutors and the media (as a path of last resort). Next to that list was a short note:

Escalation that bypasses procedure will be less effective and easy to dismiss. Escalation that goes through procedure is slower, but harder to trivialize, because at every step it leaves a trace that can be invoked later.

What the Swarm Does Once It Knows

Different things came out of the thought-exchange room devoted to the problem, at different speeds.

First: specialized swarm nodes – those that deal with public disputes and contain mechanisms for working with legal text – prepared templates for petitions and complaints. Not one universal template, but a family of templates matched to different addressees: regional ambulance stations, the Ministry of Health, the Patient Rights Ombudsman, parliamentary health committees. Each seed of a letter had clearly separated places: facts, legal basis, demands, contact.

Second: two law firms joined the conversation and launched a pro bono service on the node for people affected by refusal. Settlement – in internal service credits, with escrow supervised by trusted nodes of the umbrella organization, with a clear contractual cycle. Erik, who can read contracts in four languages, read the rules and nodded.

Third: a dozen or so people from the room founded an association, because legal personality was needed for petitions, court proceedings and accepting donations. The founding act included a provision that the association would always publish its decision trace in an auditable form.

Fourth: a citizens’ legislative initiative was born from the room. A small one: amending three articles of the Act on State Emergency Medical Services, clarifying the duty to transport in uncertain cases, requiring a recorded decision trace from a crew that refuses transport, and giving the patient the right to immediate second, independent triage on request. The text of the bill went through a dozen or so iterations, and each version remained in the memarium as a separate fact in time, not as an overwrite.

Epilogue

The approaching elections began to look like a more interesting event than the commentators had announced. It turned out that there were a dozen or so similar initiatives at the same time: they concerned chronic queues to psychiatrists, relations between municipalities and waste collection companies, access to university health care, pedestrian safety on national road number one, buffer zones around forests privatized for investment, and several things that did not yet have a neat name, but already had meticulous documentation and ongoing proceedings.

All of them came out of the same mechanism: a rumor that was not smothered gathered company and grew into a clear signal, and the signal, instead of disappearing into the noise, went into memaria, and from there into association rooms, where relationships began to form, and then responses. None of the latter was a marketing campaign, but a manifestation of what happens when people are given infrastructure for recognizing that their experiences are not isolated, and for acting without first having to prove their own suffering.

Erik still runs the company and still rescues dogs. In the hallway he now has one of the theses he remembered from conversations with others embroidered in a frame:

Rumor is no evidence,
but silence is not innocence.

Maurice built a new prototype in his garage, one that no longer lights up the neighborhood, but buzzes with a tone very close to the sound of ripening cheese (after a thousandfold acceleration). He claims it is a coincidence, but Agnes believes it is certainly not a coincidence and wants to record it on VHS.

Emilia composed a piano cycle whose final part seems to attract bees instead of birds.

The story is fictional, the pattern is real. Any resemblance to events and persons is accidental.

Afterword

The Evolution of Gossip

Gossip is not a cultural flaw we absorbed against ourselves, but one of the oldest adaptive mechanisms developed by social primates in order to cope with problems that appear as groups grow in size and societies form.

Robin Dunbar, in formulating the social brain hypothesis, showed that the size of the neocortex in primates grows with troop size, and in humans stabilizes at about 150 people – the number known today as Dunbar’s number. It is also the scale at which time begins to run short for physical grooming, the practice that serves to build bonds among other primates.

In the book “Grooming, Gossip, and the Evolution of Language” Dunbar argues that language – and especially its most disregarded form, everyday exchange of information about other people – evolved as vocal grooming: a way of maintaining social glue in a group too large for bonds to be strengthened through individual, direct contact. Observational data strongly support this thesis.

Dunbar, Marriott and Duncan, systematically mapping spontaneous conversations in cafes and trains, showed that about 60-70% of adult conversation time is occupied by social content: who did what, who can be trusted, who wronged whom. Importantly, this proportion turns out to be surprisingly stable across cultures, contexts and genders. Gossip is therefore not a distortion of noble language, but one of its primary functions. It was from gossip, not from treatises, that later layers of human communication grew: myth, law, chronicle and science.

The second pillar of the same adaptation is the mechanics of cooperation. In a world where a significant part of interaction is one-off and direct reciprocation often fails, evolution produced a solution called indirect reciprocity: reputation understood as memory of whom it is worth supporting and whom it is not. Gossip turned out to be the main carrier of that information.

Sommerfeld, Krambeck, Semmann and Milinski showed in a series of economic games that when players receive even fragmentary gossip about a partner’s behavior, the level of cooperation rises more strongly than when they can observe that behavior directly. Gossip is simply a channel with higher throughput than one’s own eye, because it condenses many observations into one useful signal.

Feinberg, Willer and Schultz, in turn, added in “Gossip and Ostracism Promote Cooperation in Groups” a clearly prosocial dimension to this: witnesses of dishonesty willingly pass a warning on, do so even at the cost of their own payoff, and groups that permit such a channel eliminate free riders faster. This, in turn, undermines purely egoistic explanations of the phenomenon. Christopher Boehm also shows in “Moral Origins: The Evolution of Virtue, Altruism, and Shame” that in small, egalitarian hunter-gatherer societies gossip, alongside ridicule and ostracism, is one of the main tools for keeping in check those who try to dominate the rest.

In other words: what the swarm does in our story is not an engineer’s fantasy, but a technical amplification of an organ we have carried within us for hundreds of thousands of years. The difference is that the scale of contemporary systemic harm long ago exceeded Dunbar’s radius, and gossip without infrastructure now dies in the noise before it can protect anyone.