Patrick Stewart as Captain Jean-Luc Picard.

CBS via Getty Images

When Patrick Stewart, who died on Wednesday, first stepped onto the bridge of the Enterprise in 1987 in his defining role as Captain Jean-Luc Picard, the Berlin Wall still stood. The Soviet Union would not fall until the fourth season of that show, Star Trek: The Next Generation, had already begun. Yet Picard’s generation of Star Trek franchise anticipated both events**,** portraying a post-Cold War world just as the Cold War was ending. It was the quintessential show for an age of American hegemony.

For those unfamiliar with the franchise, Star Trek is primarily a story about Starfleet, which serves as both the military arm and as a kind of diplomatic corps for the United Federation of Planets, a benevolent, democratic union of more than a hundred worlds. In Star Trek’s Original Series, the 1960s episodes featuring Kirk and Spock, the Federation stands in the same shoes as the Cold War United States — vast and powerful, but also harried by other galactic superpowers like the Klingon Empire.

Picard’s Federation, by contrast, is the sole dominant power in Earth’s quadrant of the galaxy. The Evil Empire is now an ally. The Original Series’s writers conceived of the Klingons as thinly veiled stand-ins for the Soviet Union, but on Picard’s Enterprise there was a Klingon, Lieutenant Worf, serving directly behind Picard on the Enterprise’s bridge.

Yet, while the Federation was a metaphor for an American empire victorious after the Cold War, it was also an idealized vision of the United States. The Enterprise spent far more time rescuing people from environmental catastrophes or shuttling ambassadors to peace negotiations than it did battling aliens in space. Picard flew around the galaxy in a ship packed with enough armaments to destroy entire planets. But he more often used that ship to distribute vaccines.

In its best moments, The Next Generation was a meditation on the just use of power. And Picard, as Enterprise’s captain, was the voice of 1990s liberalism — known as much for his eloquent soliloquies in defense of universal rights as for his tactical prowess as commander of the Federation’s most advanced weapon.

Looking back, it’s easy to see Picard’s liberalism as a vestige of a lost era. The Next Generation showcased a just and highly capable Federation that bears no resemblance whatsoever to the current United States government. But, while America has fallen far short of Star Trek’s Clinton-era aspirations, it would be a mistake to abandon those aspirations today.

The children of nearly four decades ago watched Stewart’s portrayal of Picard and saw a role model for how the most powerful people in society should behave. Those lessons are just as relevant today as they were then.

The institutionalist liberalism of Jean-Luc Picard

The decision to cast Stewart, one of the preeminent Shakespearean actors of his generation, as the Enterprise’s captain was an inspired choice. Though the show often burdened its actors with heavy-handed dialogue, Stewart delivered his lines with the same aplomb and gravitas he would bring if he were playing Macbeth.

Consider, for example, Stewart’s performance in The Measure of a Man, an early episode focused on the legal status of Lieutenant Commander Data, an android who serves as one of the Enterprise’s senior officers. In it, Data is ordered to report to an expert on artificial life forms who wishes to dismantle and study him in order to learn how to construct more, similar androids, and Picard is tasked with defending Data’s right to refuse this order before a judge that he personally loathes.

The turning point in the episode comes when Picard realizes that he is arguing, not just to save one android’s life, but to prevent the creation of an entire class of people with no more rights than a toaster. “A single Data,” Picard says in the trial that will determine the android officer’s future, “is a curiosity, a wonder even. But thousands of Datas, isn’t that becoming a race?”

“Are you prepared to condemn him and all who come after him to servitude and slavery?” Picard asks the judge in his closing argument.

When this episode first aired in 1989, democracy was ascendant. The Berlin Wall would stand for only a few more months. And the United States had emerged from the 1960s and even escaped the Reagan administration with its commitment to civil rights intact. Just a few years earlier, American liberals outmaneuvered opponents of the landmark Voting Rights Act of 1965 — the law that probably did more than any other to dismantle Jim Crow — and persuaded President Ronald Reagan to sign legislation expanding its scope.

And so the conclusion of The Measure of a Man mirrors this optimistic reality. At the end of Picard’s courtroom defense of his android officer, the judge concludes that Data has the “freedom to choose” not to be dismantled and studied. In the liberal universe imagined by The Next Generation’s writers, institutions are trustworthy and responsive to reasoned argument.

Indeed, for a show about naval officers tasked with securing a hegemonic power, Star Trek spends a surprising amount of time on courtroom drama. In another episode, The Drumhead, Picard battles an admiral who is pulled out of retirement to investigate allegations of sabotage aboard the Enterprise — and who eventually leads a witch hunt into a largely innocent crewman whose only crime is hiding his Romulan ancestry (the Romulans, like the Original Series’s Klingons, are a longtime enemy of the Federation).

Like the trial of Commander Data, The Drumhead is best known for Picard’s speech in favor of universal rights. “When the first link of the chain is forged,” Picard tells the admiral, “the first speech censored, the first thought forbidden, the first freedom denied, it chains us all irrevocably.”

And, in keeping with Star Trek’s belief that institutions can be forces of good, the episode ends with Starfleet’s head of security storming out of the admiral’s inquisition and putting an end to it.

There have been several attempts to revive Star Trek during the Trump era, and they’ve often fallen flat in an age when US institutions can no longer be trusted to defend universal rights and freedoms. Star Trek: Discovery, which launched during Trump’s first year in office and aired for five seasons, featured sinister captains with hidden agendas, and culminated in two seasons about a collapsed, distrusted Federation struggling to rebuild itself.

Indeed, when Stewart revisited his breakout role in Star Trek: Picard, an uneven series that opens with its titular character as a retired admiral, Picard is less a crusader for justice than a relic of a more enlightened age. In the series’s first episode, a reporter asks why he decided to resign from the Federation’s corps of warrior diplomats, and an angry Picard responds, “because it was no longer Starfleet!” He then denounces the Federation’s descent into prejudice against artificial life forms like Data, the very lives he fought to protect in The Measure of a Man.

Picard might as well have been speaking about the United States. In 2020, when this episode first aired, the institutionalist liberalism that defined The Next Generation was hard to defend. Donald Trump, a man who launched his campaign by labeling Mexican immigrants rapists, was president of the United States. He’d already transformed the US Supreme Court into a rubber stamp for his own biases. And the world was about to be plunged into a global pandemic that showcased the US government’s incompetence.

With its focus on artificial life forms as a potential new enemy, moreover, Star Trek: Picard anticipated the techno-pessimism that now defines an age of addictive social media platforms and artificial intelligence models that may take all of our jobs — or worse.

The Next Generation imagined a utopia, but it was a utopia built on military might, advanced technology, and institutions capable of eradicating hundreds of billions of lives. The heroes of Star Trek embrace hierarchy — most of the important characters literally wore their rank on their collars — and they held no illusions about what hegemons must sometimes do in order to maintain peace and prosperity.

In one of the franchise’s greatest episodes, In the Pale Moonlight, a Starfleet captain spends 45 minutes confessing to a murder he committed in order to trick the Romulans into joining a war against an enemy that threatens to eradicate the Federation.

And yet, Star Trek’s message has always been that power can be wielded justly, even by superpowers. And that people who do so, like Picard, are the only ones who can build utopia.

Star Trek is right

One of the hardest parts of living in Trump’s America is that the man never gives us time to grieve. Every day can bring a new executive order that thousands of people must mobilize to fight. Institutions, like the federal judiciary, that exist to uphold universal rights are rapidly being corrupted into tools of the regime. A pandemic is overshadowed by a crisis of democracy. Even Trump’s defeat in the 2020 election could only pause the parade of outrages. We are constantly being shuffled from one crisis to the next, without any time to pause and reflect on what we have lost.

And so, as we mourn Patrick Stewart, we must also grieve the optimism that his portrayal of Jean-Luc Picard embodied. What if the United Federation of Planets was always a fantasy, and Trump’s America is inevitable?

Yet, while the United States may never become the utopia imagined by The Next Generation, Star Trek is not wrong about the role benevolent institutions must play in building a just and prosperous society. In their magisterial book Why Nations Fail: The Origins of Power, Prosperity, and Poverty, the Nobel Prize-winning economists Daron Acemoglu and James Robinson argue that successful nations share two traits. One is “inclusive economic institutions” that “encourage participation by the great mass of people in economic activities that make best use of their talents.”

These nations, the two economists write, feature secure property rights, an unbiased legal system, and public services such as universal education that prepare citizens for economic success.

But it is not enough, according to Acemoglu and Robinson, for a nation to simply aspire to inclusion. It must also centralize power in a strong government that can enforce the rule of law and even, when necessary, use force to prevent the mighty from trampling on the weak. As the economists write, without a monopoly on the legitimate use of force “and the degree of centralization that it entails, the state cannot play its role as enforcer of law and order, let alone provide public services” that require taxation, or “regulate economic activity” by wealthy businesses.

Without a strong federal government, humanitarian US institutions like Medicare or Social Security cannot exist. America owes its prosperity to police forces that protect individuals from theft and assault, and to a military that, like Starfleet, secures us against external adversaries.

Star Trek’s core insight is that there will always be powerful institutions, just as there will always be militaries capable of exterminating human life. Anarcho-communist notions of a world without hierarchy or state-sponsored violence are nothing more than fantasies. There are only institutions, which can be led by good or evil individuals.

Which is why it is important for American children to grow up with fictional role models like Jean-Luc Picard — or perhaps like Atticus Finch, the model of white rectitude taught to so many American schoolchildren — who teach people who are destined to wield power to do so inclusively and justly.

In Trump’s America, such role models have become unfashionable for two reasons. The first, as I’ve already laid out, is that it is hard to tell a hero’s story about good men achieving justice by working within noble institutions when this story is so much at odds with how institutions function in the real world. A closely related reason is that, even before the rise of Trump, there were myriad examples of powerful, well-meaning individuals fostering death in the name of democracy. The Iraq War, and the broader neoconservative foreign policy project, was a failed effort to spread liberal democracy through warfare.

Given these failures of American institutions, it is easy to dismiss fictional heroes of liberalism — like Picard or Finch — as white saviors whose idealism cannot work in the real world, and to conclude that efforts to emulate them will only lead to an Iraq-style disaster. And it certainly is naive to think that evil can always be defeated by reason and good intentions. The world of To Kill a Mockingbird, where Finch ultimately fails to save an innocent man from conviction and death, often bears far more resemblance to the actual world than The Measure of a Man.

But even if just institutions are not inevitable, they are still the only way forward. As Acemoglu and Robinson show, the only thing that has ever built a sustainably prosperous and egalitarian society are strong governments that empower individuals to share their talents with their countries.

And so we should teach children to emulate characters like Jean-Luc Picard, because someday many of those children will wield unimaginable power — and because they can, by wielding that power justly, sustain the closest thing humanity has ever known to utopia.

U.S. workers are earning the smallest share of U.S. GDP on record.Lea Suzuki/The San Francisco Chronicle—Getty Images

As corporate earnings soar and U.S. GDP balloons, the American workforce isn’t feeling the same boom. American workers are taking home less of the country’s overall income, data from the Bureau of Labor Statistics shows, and employment in the U.S. is set to continue to slow.

Labor share, or the portion of the U.S.’s economic output that workers receive through salary and wages, decreased to 52.8% in the second quarter of 2026, its lowest level since the BLS started recording this data in 1947, according to its labor productivity and costs report. In the previous quarter, labor share was at 53.6%. This decade, the labor share average was about 55.6%.

That’s despite corporate earnings skyrocketing, with profits for Fortune 500 companies hitting a record $2.1 trillion in 2025. The U.S. GDP grew 3.4% in the third quarter last year, exceeding economists’ predictions. 

That growth has not only come at the expense of how much of the pie of wealth workers are taking home, but also how many Americans are in the workforce, economists warn.

“That decline in the share of labor has got to be either falling earnings or falling numbers of people,” Raymond Robertson, a labor economist at Texas A&M’s Bush School of Government, told Fortune. “The falling share of income is having to do with the shift towards capital.”

Indeed, there are growing signs that as national income balloons, the U.S. workforce is deflating. Unemployment ticked up to 4.2% in September, and employers added just 584,000 jobs in 2025 compared to 2 million added in 2024. The U.S. added 29,000 jobs in September, about one-third of economists’ projections.

The bifurcation of corporate victories and weak labor data raises concerns among economists of jobless growth jeopardizing the U.S. workforce, as well as a K-shaped economy, where the rich get richer while the poor get poorer, becoming more exaggerated.

“Data right now is very mixed,” Robertson said. “But I think it also all consistently points to this idea that things are getting worse for workers and much better for billionaires.”

What is causing jobless growth?

Robertson attributes weakening labor share averages to the rise in automation, which he noted is displacing workers, with productivity—a metric essentially measuring worker output—continuing to rise. Second-quarter GDP data showed nonfarm productivity growth grew 1.4% last quarter, though the rate has slowed from 2.2% in the same quarter a year ago.

“All these things, bit by bit, are replacing people, and they’re concentrating income and their share of capital,” he said.

Goldman Sachs analysts Joseph Briggs and Sarah Dong estimated in a report, based on Department of Labor job numbers, that AI automation could displace 25% of all work hours. They predicted that over the course of the AI adoption period, a 15% increase in AI-driven productivity would displace 6% to 7% of jobs, and, at its peak, a 1 million increase in unemployed workers, though job displacement as a result of AI has not shown up meaningfully in any economic data.

The displacement could be substantial, the analysts said, but said the impacts of automation will be tempered by a wealth of new jobs created as a result of the technological changes.

Automation is expected to be a boon to corporate profits and GDP, projected to boost GDP by 1.5% by 2035, according to a Wharton brief published in September 2025. Early signs indicate AI could drive productivity gains, with around 80% of employees reporting AI makes them faster at specific tasks, but only 37% of companies seeing an impact of that productivity on operating profits, according to McKinsey data, and AI operating costs continue to constrain its deployment.

Robertson added that growing unemployment, which he expects to see rise over the next few months, keeps wages down, allowing margins and profits to expand.

To be sure, the recent productivity surge has been an “open question,” Morgan Stanley economists have argued, not unanimously attributed to increased adoption of AI or automation. The analysts suggested this increase would be cyclical, or vestiges of pandemic-era habits of companies making more from less.

An Oxford Economists research brief from earlier this year suggested companies are disguising overhiring-related layoffs as a result of AI, but said automation-related workforce reductions have not yet happened en masse. Additionally, while unemployment has been ticking up over the past year, it is still relatively low.

Why is an immigration crackdown backfiring on U.S. labor?

Mark Regets, senior fellow at National Foundation for American Policy (NFAP), sees a different reason for a slowing workforce. He told Fortune President Donald Trump’s immigration crackdown has not done what Trump administration officials, such as White House Deputy Chief of Staff Stephen Miller, said it would in increasing the number of U.S.-born workers. Instead, according to Regets, Trump’s immigration policies have not only decimated the foreign-born workforce, but has also created fewer opportunities for domestic-born workers to find jobs.

NFAP data shows a 1.1 million to 1.4 million-person reduction in the foreign-born workforce since January 2025, consistent with the Congressional Budget Office’s report last year indicating shrinking U.S. population growth as a result of migrants being deported or refusing to come to the U.S. out of fear of hostile polities.

“The data is raising huge red flags that we are losing immigrants of all types that we otherwise would be advancing America’s economy,” Regets said.

Recent BLS data is counter-evidence to Miller’s argument that harsher immigration policy would grow the U.S. workforce. The labor participation rate for foreign-born workers fell from 66.3% to 66% from September 2025 to September 2026, and labor participation for native-born workers has similarly fallen from 61.6% to 60.9% over the same period. Fewer immigrant workers may actually make it harder for U.S.-born individuals to find work.

“A company unable to find the workers it needs for some roles could shut down operations rather than continuing,” Regets said.

He noted skillset diversity in a workplace could boost productivity and justify employing more people. Greater immigration can also increase consumer spending and stimulate businesses, as well as encourage businesses to take advantage of ample labor market availability and seek out their labor instead of offshoring jobs.

How does the U.S. reverse a shrinking labor force?

While friendlier immigration policies could help reverse an exodus of foreign-born workers, Robertson said addressing the workplace automation push would be key to growing the U.S. workforce.

“There are trades that are technology-assisted,” he said. “Those are going to be in higher demand, but you really still have to have a significant investment in skills.”

The young generation of workers are already prepared to adapt to a changing labor landscape. Gen Z are flocking to trade schools in hopes of a finding a job as a carpenter or welder not so easily outsourced by AI, and in 2024, enrollment in vocation-based community colleges increased 16%, according to data from the National Student Clearinghouse. 

Companies have taken it upon themselves to provide reskilling opportunities to employees. A CareerTrainer survey from June found that 72% of hiring managers intended to reskill employees. While the U.S. Department of Labor updated guidelines to encourage states to adapt workplace development systems, Robertson argued the government hasn’t done enough in several decades to imbue the workforce with necessary skillsets for future jobs.

“Democrats and Republicans have not significantly invested in training [or] the retraining or active labor market programs that you need to match workers to jobs,” Robertson said. “That’s the obvious solution.”

Without changes, economists see the pattern of an employment slowdown continuing, but with greater concern about the ability for the U.S. economy to sustain growth.

“We need job growth to have a growing economy, and I think we need job growth to pay our debts,” Regets said. “I don’t know how you have job growth with a shrinking labor force.”

A version of this story was published on Fortune.com on Jan. 13, 2026.

Key Takeaways

• The AI Now Institute released Uber for Nursing Part II on April 21, 2026, documenting how gig platforms including ShiftKey, Clipboard Health, and CareRev are lobbying in at least 17 states to exempt themselves from healthcare staffing rules.
• Platforms use AI-driven dynamic pricing, bidding auctions, and rating systems that can pay nurses different rates for the same shift, while classifying them as independent contractors without benefits.
• New York rejected the gig model in 2025, but bills remain active in Colorado, Illinois, Iowa, Wisconsin, and other states, meaning the rules that govern how nurses are hired through apps are being written right now.

A new AI Now Institute report published April 21, 2026, warns that gig-work platforms marketed as "Uber for nursing" are aggressively lobbying states to rewrite healthcare staffing rules, a push that could leave nurses with less pay, fewer protections, and less control over their shifts, according to The Guardian.

The report, titled Uber for Nursing Part II, identifies apps like Clipboard Health, ShiftKey, CareRev, and IntelyCare as the main players reshaping how hospitals and long-term care facilities fill shifts. According to the report, the ten most prominent gig nursing platforms have raised roughly $1.4 billion in funding, with ShiftKey valued at $2 billion and Clipboard Health at $1.3 billion.

Researchers say the platforms' AI-driven pricing, performance surveillance, and contractor classifications are creating a new category of gig work that looks a lot like rideshare, and may be on track for similar regulatory battles.

Inside the 'Uber for Nursing' Playbook

The report describes a system where nurses download an app, get rated on performance, and often bid against each other for shifts. Platforms use algorithmic pricing, including CareRev's "Smart Rates" system, which recommends hourly shift rates based on market demand and past performance, according to the AI Now Institute.

Some platforms run "Quick Bid" auction systems where nurses bid against each other for the same shift, with the lowest rate winning. The report also raises concerns about what it calls "surveillance wages," or algorithmic wage discrimination, where platforms may use personal data to set different pay rates for nurses doing identical work. A companion Roosevelt Institute brief found that some platforms "use data-brokers to determine nurse debt levels and offer lower wages to those with highest debt loads."

Workers are classified as independent contractors, which excludes them from minimum wage protections, overtime, workers' compensation, and unemployment insurance, according to the AI Now Institute. Nurses also report arriving at facilities with no paid training or orientation, and say shift cancellations and ratings can permanently affect their access to future work.

One 29-year-old St. Louis nurse identified only as Dana, quoted in coverage of the researchers' earlier interviews, said: "I love being a nurse. But I hate being a nurse right now with these greedy, immoral, corporate companies have done to health care."

Where the Fight Is Happening

Since 2022, lawmakers in at least 17 states have introduced bills to carve gig nursing apps out of the rules that apply to traditional healthcare staffing agencies, per the report. Eight states, including Colorado, Illinois, Iowa, Louisiana, Minnesota, Missouri, Nevada, and Rhode Island, have already advanced exemption bills. Six more, including Georgia, Ohio, Pennsylvania, California, Missouri, and Wisconsin, have seen bills aimed at carving platforms out of worker protection laws.

The lobbying is well-funded. Federal lobbying by Clipboard Health and ShiftKey exceeds $1.1 million combined, and the platforms work alongside Amazon and Uber through the Coalition for Workforce Innovation, the report states.

Not every state is on board. In 2025, New York classified gig nursing platforms as healthcare staffing agencies, requiring annual registration and quarterly wage reporting, and blocking independent contractor classification. Following the law, Clipboard Health, Nursa, and KARE ceased operations in the state, the report says.

What Nurses Need to Know

Gig platforms are a fast-growing slice of the nurse staffing market, and the rules governing them are being written right now. The AI Now Institute warns that bidding wars and algorithmic pricing can push profession-wide wages down, while contractor classification strips nurses of the benefits and protections most healthcare workers expect. Healthcare facilities are already spending heavily on these services, with one Colorado nursing home spending $2.9 million in 2022 and a Chicago hospital spending $2.7 million annually on platform staffing, signaling that this model is not going away.

For nurses picking up shifts on an app, the report suggests paying close attention to rating systems, cancellation policies, and whether the facility provides any orientation or training. For unionized nurses, the authors flag risks of job displacement as facilities substitute gig workers for permanent staff.

State legislation is also the place to watch. Bills in Colorado, Illinois, Wisconsin, and other states could determine whether gig nurses are treated as employees or contractors, a classification that affects pay, safety, and accountability when something goes wrong at the bedside.

Whether models trained on different modalities converge to the same representation of the world is central to AI. The Platonic Representation Hypothesis (PRH) conjectures that they do as they scale, but without a proof it offers neither a guarantee nor a recipe. We turn the hypothesis into a theorem. First, we prove a \emph{Platonic Representation Theorem}: Joint-Embedding Predictive Architectures (JEPAs) trained separately on any number of modalities learn the same representation up to one rotation whenever their positive pairs are two observations of one scene--\emph{good views}--and their misalignment is bounded otherwise. Second, we show that next-token prediction is a JEPA with \emph{bad views}, parts of a single document, which explains why language models align with vision only through scale. Third, we validate the theory from 145k captions to the texts of 940M web images, and uncover a scaling law of alignment that predicts unseen runs within seed noise. Good views need 42 times fewer training samples than next-token prediction on the same corpus to reach the same alignment, and a 109M-parameter text JEPA aligns with DINOv2 and CLIP's image encoders better than every language model we evaluated, up to 72B parameters, and than CLIP's text encoder: next-token prediction is a remarkably inefficient route to platonic representations. As a showcase, one rotation turns separately trained encoders into an image captioner (CIDEr .838 on COCO). Even that rotation needs no pair: on the latest pair-free benchmark, our text JEPA beats every zero-shot accuracy reported with language models up to 8B parameters, e.g., 64% against at most 56% top-5 on ImageNet-100. Positive views, and not scale alone, are thus the lever to build platonic representations.

For three decades, white LEDs have depended on rare-earth materials such as yttrium aluminium garnet (YAG) layered onto a blue LED chip (Figure 1). Opto-semiconductor device maker Seoul Viosys Co Ltd (SVC) says that it has now broken that dependence by developing and mass producing a chip that emits white light without rare earths. The chip can be as small as 10µm.

Figure 1. New invention: white WICOP 3.

Figure 1. New invention: white WICOP 3.

Seoul Viosys and its parent company South Korean LED maker Seoul Semiconductor Co Ltd opened up the micro-LED market with WICOP (Wafer Integrated Chip on PCB), which is a chip architecture that eliminates the package altogether. With WICOP, the two companies have already developed and mass produced ultraviolet, blue, green, red and infrared LEDs smaller than 100µm in size. WICOP 3 is the third generation of the technology, extending that capability to white light without rare-earth elements (REEs).

White LED market

White LEDs comprise about $10bn of the $15bn global market for packaged LEDs. WICOP 3 can now replace them with a technology free of rare earths, reckons the firm.

More significantly, WICOP 3 is claimed to be the world’s first white opto-semiconductor to achieve white light in a micro-LED smaller than 100µm.

WICOP 3 uses neither rare earths (a major source of environmental pollution) nor a package. Nor does it contain synthetic silicone resins, plastics or lead frames used in conventional LEDs. The result is said to be a far more environmentally friendly product that costs about half as much to make as a conventional white LED.

Creating new markets

The miniaturized transistor replaced the vacuum tube and created vast new markets in electronics. The company believes micro white opto-semiconductors can do the same.

WICOP 3 generates white light from gallium nitride (GaN) alone. Also, the chip contains no plastics, silicone or other synthetic organic materials. It can hence last for decades. In lighting, invisible micro-emitters could be embedded in wallpaper or window glass, lighting a space with no fixtures at all. This would end the inconvenience of replacing bulbs and the high cost of fixtures. The company expects this to create a new market tens of times larger than today's $10bn white LED market.

Performance

In its first mass production, a micro-LED chip measuring just 25µm x 10µm produced 0.015lm of white light. Its chromaticity falls inside the CIE standard white region (Figure 2). Luminous efficacy is already close to 100lm/W. The company aims to reach 200lm/W within two years. A conventional 1000µm x 1000µm chip achieves 50lm.

Figure 2. WICOP 3 brightness by size.

Figure 2. WICOP 3 brightness by size.

Technical significance

Up to now, WICOP could make chips smaller, but white light still required a package containing rare earths. White chips therefore could not go below 1000µm. This held back cost reduction. White micro LEDs under 100µm (0.1mm) were also out of reach. Removing rare earths changes that. White light is now possible both in large chips of 1000µm and above and in micro LEDs below 100µm.

Future plans

Seoul Group (consisting of Seoul Semiconductor and Seoul Viosys) will start mass production of WICOP 3 in fourth-quarter 2026. It says it will supply first those customers that have respected its intellectual property and relied on its products.

Seeking supply chain partners

Seoul Viosys uses WICOP to produce micro-emitters smaller than 100µm. Micro-LEDs built on WICOP can now carry high volumes of AI data over distances from under 1cm to tens of meters. Over short distances, data has long traveled electrically through copper wire.

Under the data loads of the AI era, copper links slow down, generate heat and consume much more power. Using light instead, Seoul Viosys arranges 1000 WICOP chips of about 10µm each in a grid within an area of 1mm2 or less. This array transmits data 10 times faster at one tenth of the power. Market research firms including Omdia forecast that this technology and augmented reality (AR) together will create a new $60bn market within five years (equal to half of today's DRAM memory market).

Seoul Semiconductor’s founder Chung Hoon Lee says that the company intends to target WICOP 1, 2 and 3 at applications in AR glasses and AI data communications. These technologies work on scales of 10µm or less. To that end, it is seeking trusted upstream and downstream micro-LED supply chain partners in AR displays and AI transmitters.

Openreach has issued a stark reminder to customers that the legacy phone network is being officially switched off on January 31, and says the deadline will not be postponed.

The infrastructure arm of BT Group, Britain’s biggest telecoms biz, circulated the warning this week in a letter to comms providers that use its network to deliver their services.

It informs them that any customers still on legacy public switched telephone network (PSTN) connections are at risk of disruption when these are retired at the end of January, 2027.

The switch-off could affect businesses and consumers alike. Beyond the plain old telephone, systems that rely on PSTN connectivity include point-of-sale terminals, fire and security alarms, and personal care alarms for elderly and vulnerable people.

The move is part of a wider government scheme to get the UK onto all-digital high-speed networks running on fiber-optic infrastructure, and the legacy phone service – with its dependency on copper cabling – was standing in the way.

Many households have now ditched landline handsets and instead use only mobile phones, yet for those who still want a landline, the changeover means plugging the phone into the back of a broadband router to access a Digital Voice service.

Openreach has been planning this move for the best part of a decade. It has already pushed back the deadline at least once, after the government asked telecoms providers not to force vulnerable customers who use phone-network systems such as TeleCare to switch until fixes are in place.

Nevertheless, public awareness seems low that analog phone services will soon be disconnected. "Openreach and all providers need to coordinate a much more cohesive campaign that reaches customers directly," said telecoms analyst Paolo Pescatore at PP Foresight.

"TV, billboards and newspaper advertising should reinforce a consistent message, supported by targeted letters, calls and door-to-door support for older and vulnerable people who need extra help," he told The Register.

"Customers need to understand what is changing, what action to take, and who will help them. Providers must take responsibility for ensuring people remain safely connected, especially those who rely on their landline or a care alarm."

Caroline Abrahams, charity director at Age UK, agreed.

"With only three months to go, the digital landline switchover is entering a critical final phase. Most people have moved over without major problems, but nobody who relies on their landline or TeleCare should be left behind," she told us.

"Anyone with concerns should contact their provider, but the responsibility cannot rest with customers alone. Providers must not give up on people who have not responded. They need to keep trying to reach them and offer practical, individual support, recognizing that some may be unwell, living with dementia, worried about scams or without anyone nearby to help."

Openreach has a webpage with information regarding the switch-off. This advises consumers to call their service provider as soon as possible to ensure service is not interrupted.

The company previously told The Register that customers without a fiber connection will likely be switched to SOTAP for Analog (Single Order Transitional Access Product), where equipment is installed in the local telephone exchange to allow them to continue to use their old phone line as before.

Openreach has warned that commercial customers delaying migration risk significant disruption and unnecessary costs - leaving it until the last minute could make it harder to secure the requisite engineering support before the deadline hits.

"Businesses and organizations that delay action are making things harder for themselves," said James Lilley, Openreach Director of All-IP.

"The straightforward migrations have largely been completed. What's left are the complex connections that underpin critical operations, requiring careful planning, supplier coordination, and thorough testing. We are urging businesses to act now, audit their remaining analog services, and work with their communications provider on a migration plan today."

Earlier this year, the firm tried to drive home the message by announcing it was stepping up its charges to service providers and companies still operating legacy business lines, who will by now be paying double the rental cost compared to last year. ®

I recently finished my second "novella".

pdfinfo-wc.png: Screenshot of the output of 'pdfinfo' and 'wc': About 50 pages / 15k words for the first document, 120 pages / 40k words for the second one.

Don't even think about it: I won't publish this. Nor send it to you privately, which is mostly the same as publishing it.

The first story is roughly 50 A5 pages long (~15k words), the second one is about 120 pages (~40k words). (Both barely fit the definition of a "novella", apparently, with the first one being a bit too short and the second one a tiny bit too long.)

It feels pretty absurd and surreal that I have done this at all. In school, anything related to "writing" was pure horror for me -- unless it was related to maths, physics, or other "technical" stuff. But here we are.

How long did it take? Well, the first one took over seven months. But the second one, the longer one? Ten days.

Some time last year, probably around July, I started reading some of the books that were lying around in my appartment for decades. And this was much more enjoyable than I thought. No idea why. It's not like I never read a book before, but it was always a somewhat boring thing to do and it took me ages -- unless it was something technical. I'm only referring to fiction and novels here. Like, one book per year, tops. Yet, I somehow got hooked last year and now I've read about 30 novels since July 2025. Some faster (a couple of days), some slower (three months).

It wasn't that much of a stretch to try to do this myself.

This blog is still technical in nature, so I'll walk through the software that I used.

groff

Both of my books (they're not really "books", but I'll keep referring to them as such for the sake of simplicity) are typeset in groff. You know, the thing that usually renders manual pages in your terminal on Linux. groff/troff is a full typesetting system, though, and it can be used to produce real books, which I've already mentioned a couple of times in previous blog posts.

troff-books-1.webp: Photo of my copies of 'The C Programming Language', 'The UNIX Programming Environment', and 'UNIX Network Programming'. troff-books-2.webp: A page of 'The C Programming Language'. troff-books-3.webp: A page of 'UNIX Network Programming', showing some diagrams.

In my case, the appeal of groff is that it's relatively small and fast, but still able to produce high-quality output. Making that 50-page document takes about 700 ms on my 13.5 year old PC, and the 120-page one takes 1.6 seconds. This includes some pre-processing which we'll cover later. Given that most modern software these days feels like walking in mud, this is refreshingly quick.

Because it's more fun that way, I made my own little macro package. There really isn't a lot to it, though. It's tiny. Not a lot is needed for a simple novel / novella -- no table of contents, no cross-references, no diagrams, no code blocks, no lists, ... Just nicely flowing text and sometimes a word gets emphasized.

Disclaimer: You'll note some TODO items in the text below. groff is a full-blown typesetting system and these things are complicated. I'm a long way from having mastered every aspect of it.

Basic formatting

There are only three macros:

.\" Section header / chapter title
.de SH
.   sp 1v
.   ps +5p
.   ce
\f[HR]\\$1\fP
.   ps
.   sp 4v
..
.\" Normal paragraph with hanging indent.
.de PP
.   br
.   ti +2m
..
.\" "Long" paragraph, longer empty space.
.de LP
.   sp 2v
..

(The name SH is a spill-over from manual pages. I've written quite a few of those and this name somehow stuck. CH for "chapter heading" or CT for "chapter title" might have been better choices. Or, you know, .CHAPTER.)

You start defining a macro with .de -- anything until .. is part of that macro.

• .SH: Go down one vee, increase type size, request centering, print the macro's first argument with the HR font (Helvetica), return to previous type size, add another four vees of space.
• .PP: Just break the current line, then request a temporary indent for the next output line. This creates a hanging indent, which is customary in German books.
• .LP: Just some empty space between paragraphs. Used for things like intra-chapter time jumps.

You can look up the individual commands in man 7 groff.

And that would already be enough to produce a document that somewhat resembles a book.

Put the snippet above in a file called macros.groff. I made one file per chapter. For example: chapter01.groff. You can then make a PDF using a command like this:

groff -Tpdf -dpaper=a5 -P-pa5 macros.groff chapter01.groff >book.pdf

A5 is roughly the page size of a "typical book".

Footers and page numbers

You'll note that there are no page numbers by default. You could actually just decide to not have them if all you want is a digital PDF. They're useful for printing, but you don't really need them in a file, if we're being honest.

Except: The thing gets confused about page breaks. Create the PDF using the example above and you'll see:

wrong-page-break.webp: PDF viewer screenshot: Last line of text outside of the page.

Looks like the baseline of the last line of output is just outside the page. Not sure what's going on here -- it's related to setting the page size to A5. In other words, I probably missed another spot where this has to be set. It'll be fixed implicitly by the following.

The whole page number thing consists of three parts: First, how large is the footer (the area where the page numbers live)? I defined this in a register (which is not really necessary but I find the "code" easier to read this way) -- the name is arbitary:

.nr _FooterSize 5v

We than plant a trap at the bottom of the page:

.wh -\n[_FooterSize]u _FooterTrap

\n[_FooterSize] "gets" the value of the register called _FooterSize. There's a minus sign in front of it, it's negative and relative to the bottom of the page. The documentation says:

A general-purpose register (one created or updated with the nr request) is implicitly dimensionless, or reckoned in basic units if interpreted in a measurement context.

I'm not 100% sure why I need to add the u. I haven't read every page of the manual, yet. Needless to say, it doesn't work without it.

Okay, so the trap gets installed at 5v before the end of the page and it'll call the _FooterTrap macro. It looks like this:

.de _FooterTrap
.\" Push current environment on stack and switch to `_FooterEnv`.
.\" Parent env will keep partially collected lines.
.   ev _FooterEnv
.\" Copy settings from default environment.
.   evc 0
.\" Don't do this for pages 1 and 2, they're title pages in my actual
.\" "book".
.   if \\n[%]>2 \{\
.\" Set vertical position, explained in the text below.
.       sp |\\n[.p]u-2v
.\" Print % register centered.
.       ce
\fR\\n[%]\fP
\}
.\" Go back to previous environment.
.   ev
.\" Request page break.
'   bp
..

It's much easier to read without the inline comments:

.de _FooterTrap
.   ev _FooterEnv
.   evc 0
.   if \\n[%]>2 \{\
.       sp |\\n[.p]u-2v
.       ce
\fR\\n[%]\fP
\}
.   ev
'   bp
..

Why is it 'bp instead of .bp? We'll get to that.

What I'm unsure about: I set the vertical position to |\\n[.p]u-2v. That's relative to the top of the page, it's the page length minus two vee. I would have expected |-2v-1v (including the "minus one vee" mentioned in the docs) to also work, but that leads to an infinite loop (probably triggers the same trap again and again). Why?

Header

For symmetry, I also want some empty space at the top of the page:

.nr _HeaderSize 5v
.de _HeaderTrap
'   sp |\\n[_HeaderSize]u
..
.wh 0v _HeaderTrap

And another TODO for me: This isn't 100% symmetrical. The length of the header is a bit larger than the footer. An issue with the page size again?

A closer look at traps

Let's have a closer look at traps. I'm not 100% sure on this -- this is to the best of my understanding.

Actually, let's take a step back from the example above. Let's start with this:

This is a little excerpt from Mary Shelly's Frankenstein.

These visions faded when I perused, for the first time, those poets
whose effusions entranced my soul and lifted it to heaven. I also became
a poet and for one year lived in a paradise of my own creation; I
imagined that I also might obtain a niche in the temple where the names
of Homer and Shakespeare are consecrated. You are well acquainted with
my failure and how heavily I bore the disappointment. But just at that
time I inherited the fortune of my cousin, and my thoughts were turned
into the channel of their earlier bent.

Six years have passed since I resolved on my present undertaking. I can,
even now, remember the hour from which I dedicated myself to this great
enterprise. I commenced by inuring my body to hardship. I accompanied
the whale-fishers on several expeditions to the North Sea; I voluntarily
endured cold, famine, thirst, and want of sleep; I often worked harder
than the common sailors during the day and devoted my nights to the
study of mathematics, the theory of medicine, and those branches of
physical science from which a naval adventurer might derive the greatest
practical advantage. Twice I actually hired myself as an under-mate in a
Greenland whaler, and acquitted myself to admiration. I must own I felt
a little proud when my captain offered me the second dignity in the
vessel and entreated me to remain with the greatest earnestness, so
valuable did he consider my services.

Save it as test.groff and make a PDF:

groff -Tpdf test.groff >book.pdf

Looks like this:

traps-1.webp

Install a trap at 10v. It's empty so far and does nothing -- insert this at the very top of the file:

.de MyTrap
..
.wh 10v MyTrap
.\"

The PDF looks the same.

Now what happens when we change the trap macro to this?

.de MyTrap
FOO
..

This is the result:

traps-2.webp

When the output reached the word "fishers", it sprang the trap. "FOO" comes after "fishers" in the PDF: "fishers" triggered the trap, then "FOO" got inserted. I think it's as simple as that, isn't it?

Suppose we want to put "FOO" at the beginning of the line and we want nothing but that word on that line. "fishers" shall be put on the next line.

If I understand correctly, switching environments is key: The environment holds the currently collected line (among many other things). Switch to a new one and we get a new buffer/line -- one that's empty so far. We can then flush/break this line to have it included in the PDF, and then switch back to the previous environment:

.de MyTrap
.   ev MyTrapEnv
FOO
.   br
.   ev
..

Looks like this:

traps-3.webp

So far, so good. Now I want there to be an empty line before and after "FOO":

.de MyTrap
.   ev MyTrapEnv
.   sp
FOO
.   sp
.   ev
..

In the new environment, we issue .sp: It breaks and inserts space. Nothing was collected so far, so the break doesn't print anything at this point. Then we have FOO -- this isn't printed directly, it's just collected. The next .sp breaks again (thus printing "FOO") and then inserts more space.

traps-4.webp

Okay. Now. Suppose that printing "FOO" with those two lines around it is all we want to do inside of the trap. We then want the normal text to continue at some absolute position on the page, say at 17v.

So, how about the following? We go back to the previous environment and then issue an absolute location request:

.de MyTrap
.   ev MyTrapEnv
.   sp
FOO
.   sp
.   ev
.   sp |17v
..

Oops:

traps-5.webp

.sp |17v broke the already collected line (the one with just "fishers" in it) and then inserted the space. We don't want that break. (Also note that "fishers" is at the exact same position where we left off inside the trap.)

This can be avoided by using ' instead of .:

'   The neutral apostrophe is the no-break control character, rec‐
    ognized where the control character  is.   It  suppresses  the
    (first) break implied by the bp, ce, cf, fi, fl, "in", nf, rj,
    sp,  ti,  and trf requests.  The requested operation takes ef‐
    fect at the next break.

So let's do this:

.de MyTrap
.   ev MyTrapEnv
.   sp
FOO
.   sp
.   ev
'   sp |17v
..

Result:

traps-6.webp

The example in the documentation doesn't change environments. But I find it much easier to understand what's going when changing them. (And I hope I got it right.)

One final remark: Environment carry settings. .ev foo starts a completely new environment (unless an environment with that name already exists), it doesn't clone the previous one. That's why my macro package also does .evc 0.

As a little exercise, see what happens when you use 'sp |12v, 'sp |11.5v, 'sp |11v, all the way down to 'sp |9v.

Inter-sentence spacing

Only when writing this blog post, I found the following sentence in the docs (Basics):

If you don't start a sentence on a new line, put two spaces after the previous sentence. GNU troff then recognizes punctuation that ends a sentence, and inserts inter-sentence space accordingly.

I didn't do that, because this isn't customary in German. But now I see it. Sometimes, it looks like this in the PDF:

Foo. Bar.

And sometimes there's a little bit of extra space:

Foo.  Bar.

As far as I know, this is usually not done in German books. I disabled it with this request, which keeps "minimal inter-word spacing" at the default value of 12 but sets "supplemental inter-sentence spacing" to 0:

.ss 12 0

This isn't part of my macro package, but it comes first in the "index" file (book.groff, see demo package below).

German hyphenation

Apparently, it's as simple as this:

.hla den
.hpf hyphen.den

den is "de(utsch) new": This includes the "new" German rules from 2006.

Markdown and other pre-processing

I actually wrote both of my stories in plain groff, using the macros I explained above.

Markdown has the big advantage of being "visually pleasing": It mostly looks like text that is intended to be read by a human being. groff isn't, really. I mean, it wasn't horrible, I wouldn't have written over 170 pages in groff if this was pain. It's fine. I use groff in several other places as well. But Markdown is a tiny little bit better, and I assume that it allows me to focus more on the creative process.

So, afterwards, I converted the second story to "Markdown". Didn't bother to do that for the first one. I put that in quotation marks because there are only very few elements, it's not full Markdown:

• # Chapter title at the top of the file.
• Empty lines between paragraphs, --- for "long empty paragraphs".
• *foo* for emphasized text.

And that's it. Makes the conversion really easy.

I could have just use plain Markdown for this whole thing and created an HTML file. I already have the tooling for that -- that's what drives my blog. Maybe this could even produce a PDF file. But what would be missing there, among other things, would be an engine to properly justify text. It hurts my eyes to see justified text in HTML, it just looks like a complete disaster. The results that groff or TeX create are much, much better.

While we're at it, we can also do some basic character replacements: I have the German typographic quotation marks U+201E and U+201C directly on my keyboard, but Guillemets are preferred in books.

I first wanted to do all this in Python but then switched to a little Perl snippet, because it makes it easy to read the entire file and then do multiline-regex stuff on it. It goes something like this:

perl -0pe '
# Guillemets are preferred in books.
s/... German typographic quotes, see demo package .../... Guillemets .../g;

# Markdown "h1" into ".SH".
s/^# ([^\n]+)\n/.SH "\1"\n/smg;

# Markdown "hr" into ".LP".
s/\n---\n/.LP/smg;

# Markdown paragraphs into ".PP".
s/\n\n/\n.PP\n/smg;

# Markdown emphasize into groff font selection:
s/\*([^*]+)\*/\\fI\1\\fP/smg;
'

The output is still groff for my little macro package. But I can now write it like this:

# Kapitel 42: Die Antwort

Da war sie also. Die Antwort auf *alle* Fragen.

Und dann ging es weiter:

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vestibulum
hendrerit blandit suscipit. Maecenas sed nisi nec ante mattis
vulputate. Pellentesque maximus leo id purus vestibulum dictum.
Donec tincidunt mollis viverra. Curabitur at ligula in turpis
pellentesque vehicula.

Duis hendrerit commodo mauris, non posuere leo vulputate ut. Nulla
finibus urna eget vehicula auctor. Mauris in posuere nibh. Vivamus
ut metus vel nisi gravida molestie vitae et tellus. Integer mi odio,
vestibulum eget laoreet vel, auctor nec ligula. Quisque eu placerat
quam, eu tincidunt mauris. Nam vestibulum lectus non leo dignissim,
non sagittis dolor venenatis.

This reads much nicer and there is syntax highlighting in Vim for it. Well, the *emphasized* parts, that is -- the rest doesn't matter:

vim-md.png: Screenshot of Vim with standard Markdown highlighting: Emphasized text uses the terminal's 'standout' attribute.

Speaking of Vim ...

Vim

Goyo

There is one super important Vim plugin for this:

https://github.com/junegunn/goyo.vim

It hides almost all UI elements.

On top of that, I put the terminal window in fullscreen mode to hide my desktop.

And there you go. "Distraction-free writing" indeed. Maybe put on some music, but there's nothing else. No browser, no clock, no nothing. Just a typewriter with an easy undo function.

vim-photo.webp: Photo of Vim with Markdown highlighting and active Goyo, terminal in fullscreen mode.

The font is the 11x22 variant of UW ttyp0.

Reflowing text

There's a popular way to format text in plain-text files:

This is one sentence.
This is another one.
And so on and so on.
One sentence per line, that's it.
Sentences can be short.
Or they can be pretty long, sometimes even exceeding the available width on your screen, thus requiring you to either scroll or have your editor soft-wrap it for you when displaying it.

As we've learned above from the groff docs, this actually carries meaning for groff (properly detect the ends of sentences). And it makes the diffs in VCS's less noisy.

Another popular way would be to put each paragraph on its own line. Like, there's only one "physical" newline character after the whole thing and it's your editor's job to properly display this:

This is one sentence. This is another one. And so on and so on. One sentence per line, that's it. Sentences can be short. Or they can be pretty long, sometimes even exceeding the available width on your screen, thus requiring you to either scroll or have your editor soft-wrap it for you when displaying it.

I used the second version for a couple of years at the end of the 1990ies and the early 2000s, but eventually switched to hard-wrapped text. Like this:

This is one sentence. This is another one. And so on and so on. One
sentence per line, that's it. Sentences can be short. Or they can be
pretty long, sometimes even exceeding the available width on your
screen, thus requiring you to either scroll or have your editor
soft-wrap it for you when displaying it.

I just find this easier to read and easier to navigate in. And it reminds me of the "good" old DOS days, where I used the T editor a lot and all my files looked like that (manually formatted, of course). It does have drawbacks, yes. It's not great for collaboration with other people, but that doesn't matter in my case.

Another drawback is that I have to reformat these paragraphs while editing. Luckily, Vim makes that easy:

" In ~/.vimrc:
set nojs
map Q gq}

I can then just put my cursor at the beginning of a paragraph:

[T]his is one sentence.
This is another one.
And so on and so on.
One sentence per line, that's it.
Sentences can be short.

And hit Q to turn it into this:

This is one sentence. This is another one. And so on and so on. One
sentence per line, that's it. Sentences can be short.
[_]

The cursor will be positioned after that paragraph. To reflow the entire document, I can just keep Q pressed down until it hits the end.

-- edit: Or do :set formatoptions+=a to have paragraphs reformatted automatically while editing it. I usually don't want that, I usually want it to be an explicit operation. But for writing prose, this might be pretty useful.

The nojs option (funny name, huh?) was new to me, only found out about it just now. The full name is nojoinspaces and it prevents the very thing that I should be doing (but which is not customary in German): It does not insert two spaces after a period. Without it, i.e., with set js, the result would look like this:

This is one sentence.  This is another one.  And so on and so on.
One sentence per line, that's it.  Sentences can be short.

Some people like that, I don't.

Git

As with virtually everything I do, this is in Git.

But that's the end. I regularly commit the current state, but I hardly care about anything else. No proper commit messages, nothing. It's just a tool for making "backups" easier and to sync the data between my laptop and desktop PC.

demo-book.tar.gz: Putting it all together

As I said, I won't publish the actual "books". But here's a demo of the "scaffold":

demo-book.tar.gz

(I won't update this. This is just a snapshot of the current status.)

The writing process itself

I won't say much about this. I'm a total noob.

I changed course a couple of times. For both stories, I tried to draft the whole thing before I began writing the actual text, but that didn't help much. The first story isn't that great anyway. It has an interesting premise, but I wasn't really "in the zone". That's also why it took so long. The second story has a much more mundane basic idea to it and then I eventually realized how interesting it actually was. There are quite a few twists and turns in it, and I love that.

It was way harder than I thought to keep the facts straight, especially with the course changes. Which character knows what, when, why? What are the motives? Is everything consistent and plausible? Constantly question everything -- does she really need to do that now or is there some other, much more likely way out? Does every sentence she said 100 pages ago still make sense after the big reveal at the end? Stuff like that. I'll have to get better at that. Probably use a little "character inventory" next time, like, a profile for each one.

Building a world with its own set of rules and history and what not turns out to be very fascinating. As is building the characters.

The length of the stories make proof-reading ridiculously hard. Like, I'm not going to read 120 pages in a couple of minutes. I re-read the second story one last time yesterday and it took ... *checks notes* ... three hours. Ouch.

Where to go from here?

Now that I've re-read that 120-page thing again yesterday ... I actually like it. I'm happy with the result. There's obviously a lot to learn: This entire field is completely new to me. I have identified quite a few places where the storytelling isn't great -- the pace isn't right, things progress too fast. (At the moment, I only notice this when reading the whole thing in one go. Oof.) Sometimes the wording isn't great. Not enough detail. And so on. But the story itself is something that I'd enjoy reading myself.

Digging deeper into groff is on my TODO list, also for obvious reasons.

What I really loved about this whole thing is that I ended up completely in the tunnel. On one of those ten days that it took me to write the 120-page thing, I wasn't even actually typing on a keyboard: We were travelling across the country, lots of idle time, I didn't have a laptop, but I basically "wrote" at least one chapter and then memorized (most of) it. Went back and forth, new ideas. Completely absorbed.

I want more of that. :-)

Maybe stuff will eventually get published (as in: I might put it on this website), but not now. I do this for myself. I'm not even remotely interested in my work getting torn apart by rude critics. Leave me alone. :-)

And if you think about suggesting the use of LLMs: Congratulations, you have completely missed the point. "Setzen, 6."

---

Unicode contains a range of symbols which don't get much use. For example, there are separate symbols for TradeMark - ™, Service Mark - ℠, and Prescriptions - ℞.

Nestling among the "Letterlike Symbols" are two curious entries. Both of these are single characters:

• Telephone symbol - ℡
• Numero Sign - №

What's interesting is both .tel and .no are Top-Level-Domains (TLD) on the Domain Name System (DNS).

So my contact site - https://edent.tel/ - can be written as - https://edent.℡/

And the Norwegian domain name registry NORID can be accessed at https://www.norid.№/

Copy and paste those links - they work in any browser!

Is this limited to TLDs?

No! This works ANYWHERE in a domain name. Copy and paste these examples:

• Script https://ℰ𝒳𝒜ℳ𝓟ℒℰ.𝒞𝓞ℳ/
• Math Bold https://𝐞𝐱𝐚𝐦𝐩𝐥𝐞.𝐜𝐨𝐦/
• Fraktur https://𝖊𝖝𝖆𝖒𝖕𝖑𝖊.𝖈𝖔𝖒/
• Math bold italic https://𝒆𝒙𝒂𝒎𝒑𝒍𝒆.𝒄𝒐𝒎/
• Math bold script https://𝓮𝔁𝓪𝓶𝓹𝓵𝓮.𝓬𝓸𝓶/
• Double struck https://𝕖𝕩𝕒𝕞𝕡𝕝𝕖.𝕔𝕠𝕞/
• Monospace https://𝚎𝚡𝚊𝚖𝚙𝚕𝚎.𝚌𝚘𝚖/
• Super script https://ᵉˣᵃᵐᵖˡᵉ.ᶜᵒᵐ/
• Sub script https://ₑₓₐₘₚₗₑ.cₒₘ/ NB not all characters supported
• Math sans bold https://𝗲𝘅𝗮𝗺𝗽𝗹𝗲.𝗰𝗼𝗺/
• Math sans bold italic https://𝙚𝙭𝙖𝙢𝙥𝙡𝙚.𝙘𝙤𝙢/
• Math sans italic https://𝘦𝘹𝘢𝘮𝘱𝘭𝘦.𝘤𝘰𝘮/
• Math Squared https://🄴🅇🄰🄼🄿🄻🄴.🄲🄾🄼/ NB the dot must not be squared
• Circled https://ⓔⓧⓐⓜⓟⓛⓔ.ⓒⓞⓜ/ NB the dot must not be circled

There are a whole bunch more miscellaneous characters you can use:

Christoph Päper 🇪🇺
@Cr1ss0v

Replying to @lambdalynx @lambdalynx @edent Wait, so one can use any of
㍳ ㏃ ㏇(!) ㏈ ffffifflfifl ㎇㎓㎬㏉ ㏋㍱㎐ ㎄㎅㎑㏍㏎㎸㎾ ㎃㎆㎒㎫㎹㎷㎿㎽ ㎁㎋№㎵㎻ ㍵ ㎀㎩㎊㏗㏙㏚㎴㎺ ₨ ℠ßst㏜ ℡㎔™ ㏝
ÅℬℂℭℰℱℐℑKℒℳℕℙℚℛℜℝℤℨ and more to leet-code URLs?
@urlstandard

Replying to @edent @edent Things used to retrieve web pages rather than web browsers
curl 7.59, Linux - Yes
wget 1.19, Linux - Yes

Replying to @edent @edent Yes, Chrome/Safari/Firefox running on Mac. The TEL however changed from superscript to normal text. If I copied/pasted into Word and then into the browser, the superscript is preserved and it no longer resolves (takes you to the google page with this page being the first hit)

How does this work?

Magic! Which is to say, I think it is the browser doing the conversion. DNS Servers don't successfully reply to queries about .℡ domains.

The browser sees the .℡ and then follows the IDNA2008 process listed in RFC5895 to normalise it:

map characters to the "Simple_Lowercase_Mapping" property (the fourteenth column) in <http://www.unicode.org/Public/UNIDATA/UnicodeData.txt>, if any.

The ℡ entry is:

2121;TELEPHONE SIGN;So;0;ON;<compat> 0054 0045 004C;;;;N;T E L SYMBOL;;;;

U+0054 is T, U+0045 is E, U+004C is L.

You can test this in Python using:

python -c 'import sys;print sys.argv[1].decode("utf-8").encode("idna")' "℡"

Does this work?

Yes! I asked people on Twitter whether they could access my website using a .℡ - and it appeared to work on every modern browser and operating system.

Terence Eden is on Mastodon
@edent

Hey gang! I have a little experiment for you 🙂

Does this URL resolve in your browser?
https://edent.℡/
(That's https:// edent. ℡ /)

If it does or doesn't, could you let me know which browser and operating system?

THANKS!

It even works on command line tools like wget and curl.

Mike
@6byNine

Replying to @edent @edent Things used to retrieve web pages rather than web browsers
curl 7.59, Linux - Yes
wget 1.19, Linux - Yes

It does fail in some circumstances:

Ricardo Sueiras (@094459@hachyderm.io)
@094459

Replying to @edent @edent Yes, Chrome/Safari/Firefox running on Mac. The TEL however changed from superscript to normal text. If I copied/pasted into Word and then into the browser, the superscript is preserved and it no longer resolves (takes you to the google page with this page being the first hit)

What are the limitations?

Two main ones:

• Sites like Twitter and Facebook don't recognise it as a valid URl and refuse to auto link it.
• Some command line tools like dig and host don't understand it

Copied Bash to 📋

dig edent.℡

; <<>> DiG 9.10.6 <<>> edent.℡
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NXDOMAIN, id: 55282

Is this useful?

Obviously yes. This may be the most important discovery of the decade. You get cool looking URls and get to save a couple of characters on specific domains, at the minor expense of working inconsistently.

It could also be used for evading URl filters.

Every modern browser supports these "fancy" domain names - but most websites won't automatically link to them. So sharing on Facebook doesn't work.

Where can it be used?

Here are the single characters which can be normalised down to a valid TLD. They're mostly country codes, but there are a few interesting exceptions:

• ㏕ - US Military
• ℡ - .tel registry
• № - Norway
• ㍳ - Australia
• ㍷ - Dominica
• ㎀ - Panama
• ㎁ - Namibia
• ㎃ - Morocco
• ㎊ - French Polynesia
• ㎋ - Norfolk Island
• ㎏ - Kyrgyzstan
• ㎖ - Mali
• ㎙ - Federated States of Micronesia
• fi - Finland
• ㎜ - Myanmar
• ㎝ - Cameroon
• ㎞ & ㏎ - Comoros
• ㎰ - Palestine
• ㎳ - Montserrat
• ㎷ & ㎹ - Republic of Maldives.
• ㎺ - Palau
• ㎽ & ㎿ - Malawi
• ㏄ - Cocos (Keeling) Islands
• ㏅ - Democratic Republic of Congo
• ㏉ - Guyana
• ㏗ - Philippines
• ㏘ - Saint Pierre and Miquelon
• ㏚ - Puerto Rico
• ㏛ - Suriname
• ㏜ - El Salvador
• ℠ - San Marino
• ™ - Turkmenistan
• st & ſt - São Tomé and Príncipe
• ㎇ - Great Britain ( Obsolete)
• ß - South Sudan ( Not available)
• ㏌ - India and Indiana (subdomain of .us)
• Ⅵ & ⅵ - Virgin Islands and Virginia (subdomain of .us)
• fl - Florida (subdomain of .us)
• ㎚ - New Mexico (subdomain of .us)
• ㎵ - Nevada (subdomain of .us)
• ㍵ - As part of .ovh

If you can find any more, please stick a comment in the box below.

You can always reach this blog post at:

https://🅂𝖍𝐤ₛᵖ𝒓.ⓜ𝕠𝒃𝓲/🆆🆃🅵/ _