The catastrophe struck Wanda Wickizer on Christmas Day 2013. A generally healthy, energetic 51-year-old, she suddenly found herself vomiting all day, racked with debilitating headaches. When her alarmed teenage son called an ambulance, the paramedics thought that she had food poisoning and didn’t take her to the emergency room. Later, when she became confused and groggy at 3 a.m., her boyfriend raced her to Sentara Norfolk General Hospital in coastal Virginia, where a scan showed she was suffering from a subarachnoid hemorrhage. A vessel had burst, and blood was leaking into the narrow space between the skull and the brain.

During a subarachnoid hemorrhage, if the pressure in the head isn’t relieved, blood accumulates in that narrow space and can push the brain down toward the neck. Vital nerves that control breathing and vision are compressed. Death is imminent. Wickizer was whisked by helicopter ambulance to the University of Virginia Medical Center in Charlottesville, 160 miles away, for an emergency procedure to halt the bleeding.

After spending days in a semi-comatose state, Wickizer slowly recovered and left the hospital three weeks after the hemorrhage, grateful to be alive. But soon after she returned home to her two teenage children, she found herself confronted with a different kind of catastrophe. Wickizer had had health insurance for most of her adult life: Her husband, who died in 2006, worked for the city of Norfolk, which insured their family while he was alive and for three years beyond. After his death, Wickizer worked in a series of low-wage jobs, but none provided health insurance. A minor pre-existing condition — she was taking Lexapro, a common medicine for depression — meant that her only insurance option was to be funneled into the “high-risk pool” (a type of costly insurance option that was essentially rendered obsolete by the Affordable Care Act and now figures in some of the G.O.P. plans to replace it). She would need to pay more than $800 per month for a policy with a $5,000 deductible, and her medical procedures would then be reimbursed at 80 percent. She felt she couldn’t afford that. In 2011, she decided to temporarily stop working to tend to her children, which qualified them for Medicaid; with trepidation, she left herself uninsured.

And so in early 2014, without an insurer or employer or government agency to run interference between her and the hospital, she began receiving bills: $16,000 from Sentara Norfolk (not including the scan or the E.R. doctor), $50,000 for the air ambulance. By the end of January, there was also one for $24,000 from the University of Virginia Physicians’ Group: charges for some of the doctors at the medical center. “I thought, O.K., that’s not so bad,” Wickizer recalls. A month later, a bill for $54,000 arrived from the same physicians’ group, which included further charges and late fees. Then a separate bill came just for the hospital’s charges, containing a demand for $356,884.42 but little in the way of comprehensible explanation.

In other countries, when patients recover from a terrifying brain bleed — or, for that matter, when they battle cancer, or heal from a serious accident, or face down any other life-threatening health condition — they are allowed to spend their days focusing on getting better. Only in America do medical treatment and recovery coexist with a peculiar national dread: the struggle to figure out from the mounting pile of bills what portion of the fantastical charges you actually must pay. It is the sickness that eventually afflicts most every American.

What’s less understood is the extent to which our current medical-billing system itself is responsible for the high prices patients are charged. There are, of course, many factors that have led to the United States’ record-breaking $3 trillion health care bill: runaway drug prices, excessive testing and sky-high charges for even the most basic medical interventions. But all of those individual price increases have been enabled — indeed, aided and abetted — by the complex system of billing and coding that underlies bills like those sent to Wickizer. That system, with its lines of alphanumeric codes and arcane medical abbreviations, has given birth to a gigantic new industry of consultants, armies of back-room experts whom medical providers and insurance companies deploy against each other in an endless war over which medical procedures were undertaken and how much to pay for them. Caught in the crossfire are Americans like Wanda Wickizer, left with huge bills and indecipherable explanations in languages they cannot possibly understand.

Disease-classification systems originated during an outbreak of the bubonic plague in 17th-century London — epidemiologic constructs to classify and track causes of death and prevent the spread of infections among populations that spoke different languages. In the 1890s, the French physician and statistician Jacques Bertillon further systematized death reporting by introducing the Bertillon Classification of Causes of Death, the first medical-coding system, which was adopted and modified in many countries. It became an official global effort, which was periodically revised by an international commission. During the first half of the 20th century, the number of entries naturally increased with improved understanding of science, and many countries began tabulating not just causes of deaths but also the incidence of diseases.

In the 1940s, the World Health Organization took over stewardship of Bertillon’s system and renamed it to reflect a new, broader focus: the International Statistical Classification of Diseases, Injuries and Causes of Death (ICD). The codes became an invaluable tool, a common language for epidemiologists and statisticians to track the world’s afflictions. But over the last several decades in the United States, codes gradually took on a bedrock financial function as the basis for medical billing. In 1979, the government decided to use what by then were called ICD-9 codes — which specify the patient’s diagnosis — in adjudicating Medicare and Medicaid claims, with some modifications added specifically for that purpose; the United States version was called ICD-9-CM. (The country has recently moved to a new iteration, ICD-10-CM.) For its beneficiaries, Medicare pays a fixed fee for inpatient hospitalization based primarily on the ICD-CM code, which is translated into a DRG (diagnosis-related group) code — which is the immediate basis for reimbursement.

Other insurers followed in making codes the basis for billing. Coding systems begot new coding systems, because few hospitals wanted to be paid according to Medicare’s relatively low DRG standards. And because strategic coding meant increased payment, that begot coding specialists and coding courses and coding degrees. There are now different increasingly complex coding languages that define payment for different kinds of services: CPT codes, for office visits delivered by doctors, as well as HCPCS, ICD-PCS-CM and DRG, for charges that are incurred in the hospital. There are tens of thousands of codes in each lexicon that have become increasingly specific. For example, there are different codes for in-office earwax removal depending on the method used (irrigation or instruments), different codes for delivering different vaccinations and a code for each injection delivered in the hospital. Different insurers also use different coding systems. While Medicare would have most likely considered Wickizer’s brain bleed as DRG 021, if billed to a commercial insurer, it could result in more than a dozen ICD codes and hundreds of HCPCS entries.

Seemingly subtle choices about which code to use can have large financial consequences. If after reviewing a hospital chart of, say, a patient who has just had a problem with his heart, a hospital coder indicates the diagnosis code for “heart failure” (ICD-9-CM Code 428) instead of the one for “acute systolic heart failure” (Code 428.21), the difference could mean thousands of dollars. “In order to code for the more lucrative code, you have to know how it is defined and make sure the care described in the chart meets the criterion, the definition, for that higher number,” says one experienced coder in Florida, who helped with Wickizer’s case and declined to be identified because she works for another major hospital. In order to code for “acute systolic heart failure,” the patient’s chart ought to include supporting documentation, for example, that the heart was pumping out less than 25 percent of its blood with each beat and that he was given an echocardiogram and a diuretic to lower blood pressure. Submitting a bill using the higher code without meeting criteria could constitute fraud.

Each billing decision, then, can be seen as a battle of coder versus coder. The coders who work for hospitals and doctors strive to bring in as much revenue as possible from each service, while coders employed by insurers try to deny claims as overreaching. Coders who audit Medicare charts look for abuse to reclaim money or fraud that needs to be punished with fines. Hospital coders teach doctors — and doctors pay to take courses — to learn how they can “upcode” their charts to a more lucrative level with minimal effort. In a doctor’s office, a Level 3 visit (paid, say, at $175) might be legally transformed into a Level 4 (say, $225) by performing one extra maneuver, like weighing the patient or listening to the lungs, whether the patient’s illness required that or not.

While most hospitals and insurers set their own rates for each level of care, adding a step when interacting with a patient can also bring windfalls. E.R. doctors, for example, learned that insurers might accept a higher-reimbursed code for the examination and treatment of a patient with a finger fracture (usually 99282) if — in addition to needed interventions — a narcotic painkiller was also prescribed (a plausible bump up to 99283), indicating a more serious condition.

Toward the end of the 20th century and into the next, as strategic coding increased, a new industry thrived. For-profit colleges offered medical-coding degrees, and internships soon followed. Because alphanumeric coding languages are as distinct from one another as Chinese is from Russian, different degree tracks are necessary, along with distinct professional organizations that offer their own particular professional exams, certifications and licensing. Hospital systems and insurers — which have become huge, Hydra-like enterprises — now all employ roomfuls of coding-program graduates to perform these tasks. Membership in the American Academy of Professional Coders has risen to more than 170,000 today from roughly 70,000 in 2008.

Individual doctors have complained bitterly about the increasing complexity of coding and the expensive necessity of hiring their own professional coders and billers — or paying a billing consultant. But they have received little support from the medical establishment, which has largely ignored the protests. And perhaps for good reason: The American Medical Association owns the copyright to CPT, the code used by doctors. It publishes coding books and dictionaries. It also creates new codes when doctors want to charge for a new procedure. It levies a licensing fee on billing companies for using CPT codes on bills. Royalties for CPT codes, along with revenues from other products, are the association’s biggest single source of income.

Patients with good health insurance are often blissfully unaware and mostly unaffected by the jockeying that goes on over how to code their bills. But uninsured patients like Wickizer, or (increasingly) those with high deductibles, are stuck with no insurer to argue on their behalf. Her experience with the University of Virginia Medical Center is not unique: Studies have shown that hospitals charge patients who are uninsured or self-pay 2.5 times more than they charge those covered by health insurance (who are billed negotiated rates) and three times more than the amount allowed by Medicare. That gap has grown considerably since the 1980s.

When Wickizer arrived home from the hospital in January 2014, she had trouble concentrating and finding words; she spoke deliberately, slowly. She remembers nothing before February, she says, but relied on help from her parents, who live nearby, and her boyfriend, who is retired from the Navy. She did her best to address the onslaught of bills that began appearing in her mailbox.

First, she took stock of her finances. She paid the rent for the Norfolk apartment that she and her children lived in by renting out a townhouse that she and her deceased husband had bought in Virginia Beach; after paying property tax, insurance and maintenance on the townhouse, she just broke even. She also received about $2,000 a month in Social Security survivor benefits because of her husband’s death. In addition, she had about $100,000 from her husband’s life insurance in a retirement account, which she was also hoping would help pay for her children’s college. With medical bills totaling nearly $500,000 and no health insurance, the numbers didn’t add up. “My dad said: ‘They’ll never expect you to pay that,’ ” Wickizer told me. “But they did.”

As a sign of good faith, she quickly paid $1,500 to the hospital and $1,000 to the doctors and sought to make sense of the bills. Patients today are told to be good medical consumers, but they are asked to write checks for thousands of dollars — in this case hundreds of thousands — with little explanation of what they’re for. Wickizer did what she would have done with a credit-card statement: She contacted the hospital and requested an itemized bill. Her idea was that if she could understand how much she was being charged for each procedure, she could compare the fees with the reimbursements that Medicare or another insurer would pay for those services and begin some kind of negotiation.

A month later, on March 19, the hospital finally sent a list of charges, using medical abbreviations and terminology but not revealing the all-important alphanumeric codes. Despite being 60 pages long, the tally seemed incomplete, leaving out doctor’s charges and including other fees that seemed incidental, like charges for catheters, wires and oxygen. Room charges were vastly different on different days.

Nearly simultaneously, she received a one-page bill for the hospital portion of her care, broken down only into the broadest categories, including $111,162 in room charges, $34,755.75 for pharmacy, $19,653 for labs, $8,640 for the operating room, $8,325 for anesthesia, $1,143 for the recovery room, $44,524 for medical supplies and $40,489 for radiology services, totaling $356,884.42. The bill informed her that the medical center was prepared to offer her its standard 20 percent discount for patients who are uninsured, leaving a “what you owe now” fee of $285,507.58. It noted that the hospital could offer some additional financial assistance, but only if her household of three had assets of less than $3,100 (“such as bank or retirement accounts”), which disqualified Wickizer and very likely most Americans who have ever held a job.

Next, she did her best to find out what Medicare or another insurer would have paid for her hospitalization, hoping to offer the hospital that amount from her retirement account. To understand the Medicare codes, she had to learn a bit of coding language. Would her hospitalization count as Medicare DRG 020 or 021? She estimated that in 2013, her subarachnoid hemorrhage (most likely coded, she determined, as “intracranial hemorrhage or cerebral infarction disorders, DRG 021, with procedures and major comorbidities or complications”), would have been reimbursed by Medicare for about $80,000. Had a member of the armed services experienced the same condition, Tricare, the military insurer, might have paid closer to $70,000. But to know how much a commercial insurer would have paid, she would have to figure out what HCPCS codes the hospital used to calculate her bill, and the hospital did not send those. Hospitals tend to treat their billing strategies — codes and their master price list, called a charge master — as trade secrets vital to their business. State laws and judges tend to respect that as proprietary information.

When the billers called insisting on payment of the full $285,507.58, Wickizer explained, “I don’t have this kind of money.” She offered the hospital and its doctors the $100,000 in her retirement account. They declined and suggested that she sign up for a payment plan of $5,000 a month to the hospital — and a second $5,000 plan for the physicians’ group. It was an untenable amount.

In October 2014, a sheriff affixed a summons to Wickizer’s front door, saying that the university was suing her for nonpayment. Eric Swensen, a spokesman for the University of Virginia, declined to answer questions about the case, citing patient privacy, as governed by HIPAA rules. But he noted that the university provides $270 million worth of free care to patients who meet its criterion for assistance and sets up interest-free payment plans for those who don’t.

After receiving the summons, Wickizer resorted to a technique followed by many a frustrated customer: She went on Facebook, posted her story and solicited advice. (The Facebook group Paying Till It Hurts, where she posted her story, was created in 2014 in connection with a New York Times seriesthat I wrote with the same name.) A handful of experts — patient advocates, billing professionals, lawyers and a coder — volunteered their help pro bono to try to get more information from the medical center and translate the coding that yielded the unaffordable figure. (One notable aspect of our commercialized health system is that for every person who is pushing to profit, there is another who is doing his or her best to protect patients.)

In vetting Wickizer’s bill, the experts encountered roadblocks from the medical center at every turn in a contentious battle that lasted for over a year. Multiple legal requests to review Wickizer’s chart and complete bill — with its coding elucidated — were refused. Nora Johnson, a retired hospital bill-compliance auditor from West Virginia who volunteered to help Wickizer, noted that not revealing the billing codes constituted a violation of federal law. No insurer would have paid the bills without seeing them, allowing at least a rational attempt at negotiation. As Wickizer’s team wrote to the University of Virginia in one of their letters: “No Codes = No Pay.” The University of Virginia Physicians’ Group, which independently charged Wickizer $54,000, eventually turned over its billing codes. Wickizer’s experts were able to use the bill fragments they had received in discovery, supplemented by those codes, to get a better idea of what medical procedures Wickizer received during her three-week hospitalization. From there, they tried to extrapolate how the hospital had, perhaps, coded her case. By examining the cost reports the University of Virginia hospital must file with Medicare, which indicate the amount it spends delivering certain types of care, Christine Kraft, another medical-billing expert, estimated that even by its own calculations, the medical center spent less than $60,000 treating Wickizer.

The stealth battle between hospitals and insurers over bills for each hospitalization, office visit, test, piece of equipment and procedure is costly for us all. Twenty-five percent of United States hospital spending — the single most expensive sector in our health care system — is related to administrative costs, “including salaries for staff who handle coding and billing,” according to a study by the Commonwealth Fund. That compares with 16 percent in England and 12 percent in Canada.

That discrepancy comes, in part, from the prolonged negotiations over payment and the huge number of coders, billers and collectors who have to be compensated: Their salaries and loans from those years of training in obscure languages are folded into those high charges and rising premiums. In addition, as is often the case in warfare, the big conventional army can be at a disadvantage: The insurance companies and government seem to be always one step behind the latest guerrilla tactics of providers’ coders.

For years, creative coding has been winning over what the government calls “correct coding,” meaning coding that gives providers their due, but without exaggeration. Indeed, each attempt by the government to control questionable coding to enhance providers’ revenue has seemed to only fuel more attempts. In 1996, for example, Medicare’s National Correct Coding Initiative made it clear that certain codes couldn’t appear on the same bill because they were inherently part of the same procedure. As a rule, an anesthesiologist could not, for example, separately bill for anesthesia and checking your oxygen level during your surgery. But the government created Modifier 59 — a code that could be appended to other codes to allow doctors to take exceptions to that rule in unusual cases. Modifier 59 could be used to allow for two payments in certain situations, such as when an oncology nurse needed to insert two separate IVs for two different purposes — one to administer chemotherapy, say, and another hours later because the patient seemed dehydrated. Such cases were expected to be exceedingly rare.

But just as entrepreneurial corporate tax lawyers search each new tax code for economic advantage, entrepreneurial coders and billers find loopholes to exploit at the edge of the law. An investigation by the Health and Human Services Office of the Inspector General in 2005 found many instances of Modifier 59 abuse. Forty percent of code pairs billed with Modifier 59 in 2003 were not legitimate, resulting in $59 million in overpayment. Similarly, when Medicare announced that it would pay only a set fee for the first hour and a half of a chemotherapy infusion — and a bonus for time thereafter — a raft of infusions clocked in at 91 minutes.

Like nearly every area of medicine, coding science has advanced — though not to the patient’s benefit. Commercial computer “encoder” programs maximize income from coding and make helpful suggestions (“That could be billed for Level 3,” or “Did you forget Code 54150,” indicating a circumcision on a bill for a male newborn). Today many medical centers have coders specializing in particular disciplines — joint replacement or ophthalmology or interventional radiology, for example. Advanced coding consultants advise lesser coders. The Business of Spine, a Texas-based consulting firm with a partner office in Long Island, advises spine surgeons’ billers about what coding Medicare and commercial insurers will tolerate, what’s legal and not, to maximize revenue. The evolution of this mammoth growth enterprise means bigger bills for everyone — whether through increasing premiums and deductibles on insurance policies or, as in Wickizer’s situation, depleting the savings earmarked for children’s college.

Like many medical centers, the University of Virginia Health System has turned at least some of its billing and debt collection over to professionals, third-party contractors who have no pretense of the charitable mission espoused by the University of Virginia, founded by Thomas Jefferson in 1819 to educate leaders in public service. The collectors are often paid a percentage of the money they recover. They tend not to care whether a procedure was coded well or poorly. Their task is usually to go after the total sum the hospital says it is owed.

In Wickizer’s case, the hospital brought in a law firm that specialized in debt collection, then called Daniel & Hetzel and based in Winchester, Va. For a year and a half, Wickizer’s team of experts dissected the bills and negotiated with the hospital and its representatives at the law firm over its charges and coding strategies — just as insurers do behind the scenes on patients’ behalf. The experts laid out their logic for what might constitute reasonable payment in a detailed report based on what they could discover about Wickizer’s care: how it could be coded and what other hospitals and insurers would have paid. They helped her local lawyer, Kelly Roberts, write motions for discovery and legal letters and made offers of payment between $65,000 and $80,000, which they calculated should provide the hospital a profit on the services rendered to Wickizer.

But the hospital did not accept any of the offers. In a letter, Peter Hetzel, an attorney at the firm, said his client would accept only just over $225,000, saying the University of Virginia Medical Center was “the victim here.” He noted, too, that the small rental property that Wickizer owned — appraised at $90,200 in 2014 — was considered fair game for the hospital to seize as payment. Swensen, the spokesman for the university, said that it decides on a case-by-case basis whether or not to report nonpayment to credit agencies or to pursue civil cases against patients in court. He added: “If we obtain a lien on real estate, we do not seek to sell the property if it is the patient’s primary residence.”

In February 2016, Wickizer received a letter from the state of Virginia saying that the medical center would be dunning money from any tax refund she might get. At one point, in exasperation, Wickizer wrote to her group of experts: “More than likely I am going to have to declare bankruptcy by the time this is all said and done, and I just would like to have everything settled. I want to pay them what I have and what is fair.”

968COMMENTS

By then, Wickizer was recovering physically and had married her boyfriend. But she was still struggling with stress from the uncertainty of the mammoth bills hanging over her. With court dates scheduled and postponed, motions filed and denied, she and her pro bono lawyer from Chicago, Tom Osran, along with her local lawyer were finally scheduled to face off in court with the University of Virginia Medical Center on April 29, 2016. The day before trial, after Osran was preparing to book his plane ticket to Virginia, and after I called the hospital inquiring about attending the court session, the case was dismissed. The terms of the settlement are sealed.

Nearly a year later, Wickizer remains exhausted by the ordeal. Her speech, which was hesitant when I first spoke with her more than two years ago, sounds fluid now, and she is funny and thoughtful, though she says she still occasionally needs to search to find the right word, a form of a condition known as aphasia. Now working part-time as a clerk in a small store, she would like to go back to her previous work as a bookkeeper, she told me when we spoke in March. But she has failed to secure a job; she worries that her barely noticeable speech problems make her job interviews less than optimal. Or perhaps, she frets, the problem is her credit rating, which (unknown to her at the time) dropped more than 200 points after the doctors who cared for her reported her unpaid bills to credit agencies. That black mark will remain until 2021, even though her legal case is resolved and she now has military health insurance through her husband. And, she notes with a sigh of resignation, “I’m the kind of person who’s always tried to do everything right.”

Not all Twitter bots are trying to spam, hack or peddle you fake news. Some are works of creativity, programmed to tweet diagrams of imaginary bird migrations or haikus composed of words scavenged from surveys of marine mammals.

I follow Twitter bots for serendipitous notes that take me out of my day, if just fleetingly. There’s comfort in witnessing a narrative take form outside the direct control of humans and oblivious to the breakneck clip of the internet.

Science-themed Twitter bots come in various forms, doling out humor, factual information and galactic perspective. There’s @shark_girls, which casts two geotracked great white sharks as travel writers, quoting from the writings of Virginia Woolf and the poet Hilda Doolittle. There’s @the_ephemerides, which juxtaposes raw images taken by outer planet probes like the Cassini spacecraft with computer-generated poems. There are many more. Below are some of my favorites.

@grow_slow

Nicole He programmed this bot to tweet a photo of her fiddle-leaf fig, a common houseplant, at 10:17 every morning. Over time, you can see the plant birth new offshoots and shuffle its leaves ever so slightly.

“It’s striking how often people tweet encouraging, nice things at the plant about how shiny its leaves are or how fast it’s growing,” wrote Ms. He, a programmer and artist, in an email. “I think I’ve discovered that a humbly tweeting plant is actually the secret to world peace (or at least kindness on Twitter).”

@usinjuries

This bot tweets descriptions of emergency room visits from a government database that tracks about 100 hospitals across the United States.

Its parent, Keith Collins, a reporter at Quartz, emailed that he didn’t expect to laugh out loud when he first looked at the data. But most of the injuries are minor, he said, and there’s something about the way they’re written in the “pithy style of a rushed E.R. doctor.” Noticing a glut of entries about patients who punched walls, he charted the age distribution of wall-punchers and found that 15-year-olds were most prolific.

Many of the bot’s retweets come with comments like “same” or “it’s me,” he said. It “gives us a chance to laugh at ourselves.”

@birdcolourbot

Each tweet from @birdcolourbot is a bird name followed by a swath of colors resembling a paint chip. Each band’s width is determined by the probability of a given bird of that species being that hue.

“I’m red-green colorblind, so I’m interested in color perception and how different people see birds (or anything really),” emailed David L. Miller, the bot’s creator and a statistician with affiliations at the Woods Hole Oceanographic Institution in Massachusetts and the University of St. Andrews in Scotland.

@mothgenerator

Also dedicated to winged creatures, this bot tweets make-believe moths of all shapes, sizes, textures and iridescent colors. It’s programmed to generate variations in several anatomical structures of real moths, including antennas, wing shapes and wing markings.

Another program, which splices and recombines real Latin and English moth names, generates monikers for the moths. You can also reply to the account with name suggestions, and it will generate a corresponding moth.

Inspired by naturalist illustrations, such as those of Ernst Haeckel, the programmers designed their bot to create moths stroke by stroke, with each insect composed of tens of thousands of individual strokes. “At its core, the moth generator is a wildly byzantine drawing machine in the shape of a moth,” said Katie Rose Pipkin, an artist at Carnegie Mellon University who created the bot with Loren Schmidt.

@i_find_planets

“They have discovered a planet. A guinea pig-like creature lives there and creeps through the valleys. It is something of a mystery.” Such are the snapshots offered by Newfound Planets, a bot that tweets about fictional distant worlds.

The human behind it, Charles Bergquist, who directs the public radio program Science Friday, wrote via email that he thinks people enjoy the bot because they yearn to know what it’s like on another planet — “how big it is, might there be water or what might the sunrise look like.”

@unchartedatlas

This bot also tweets about fictional lands, but based on actual erosion science. Martin O’Leary, a glaciologist at Swansea University in Wales, programmed it as part of National Novel Generation Month, a spinoff of National Novel Writing Month that challenges people to write an algorithm that writes a novel (here’s his accompanying novel).

The bot starts with a random initial terrain, then simulates how water would flow over it to create channels, valleys and coastlines. Cities are placed away from each other and near running water, and Dr. O’Leary wrote another program to name the cities.

To him, Twitter bots are a piecemeal form of science outreach. He said a lot of publicly available information about science amounted to, “Look at this amazing thing right now!” But Twitter bots “worm their way into your life, sit there and slowly give you this drip of stuff. They’re gentle.”