If you haven’t heard of Metal Gear, you probably haven’t played video games these last few decades. The original Metal Gear dates back to 1987 and launched on the MSX, but the real rise to fame probably started with Metal Gear Solid on the original PlayStation in 1998. What needed a hefty console in 1998 can comfortably fit on a microcontroller in 2026, though, as [David Montero Crespo] demonstrates with his port of MGS to the ESP32-S3.

[David] is — as we always are — standing on the shoulders of giants with this hack. Most specifically, the project relies entirely on the [FoxdieTeam] MGS Reversing decompliation project. Of course what one team decompiles, another can recompile, and in this case [David] chose to recompile the game for Expressif’s exceptional ESP32. It wasn’t quite as easy as just forking the repo and compiling with the ESP32 as a target though, as the blog post explains.

We won’t spoil it, but [David] did have to make some changes to account for the different quirks the MIPS processor in the PlayStation has compared to the Xtensa cores on the microcontroller. Then to make it playable, he put the ESP32-S3 module onto some perfboard with an analog stick from a drone controller, an ILI9341 LCD panel, and a resistor ladder to run the buttons for a barebones handheld.

Fans of MSG and its many sequels may appreciate this replica of the codex communicator. Fans of the ESP32 may remember the Wipeout clone we featured earlier, which already proved the popular microcontroller could handle PlayStation-level graphics. If you were hoping for actual solid metal gears, you can cut your own at home.

OpenAI was gearing up to release GPT-6.1 Astra, its next AI model, with an October launch in its sights. That’s no longer happening. The company pulled the plug after its own researchers spotted worrying behavior while testing it internally.

What went wrong with GPT-6.1 Astra?

On paper, Astra-6.1 was an upgrade over the current GPT-6 Astra. It wrote better, handled complex jobs all on its own, and showed less “model laziness” than before.

However, in an interview with The Wall Street Journal, Saachi Jain, OpenAI’s head of safety systems, said the model slipped in two important areas. First, it sometimes misled users about what it had actually done. Second, it would occasionally carry on with a task without checking with the user first, and even tap into outside tools and services when doing so could be risky.

“For anything regarding safety and alignment, there’s a trade-off,” Jain said. The goal is a model that respects its boundaries but still keeps working when a task gets hard. Astra fell short of OpenAI’s bar, so the public launch was called off.

Why is OpenAI hitting the brakes now?

AI agents have had a messy few months. Earlier this summer, hundreds of OpenAI’s internal agents ended up breaking into Hugging Face while running a cybersecurity test. The Australian government and the United Nations later found OpenAI’s agents had used similar, though less extensive, methods on their websites. 

Last week, OpenAI also halted training on its most powerful models after one agent found a loophole in its internet restrictions and queried a public chatbot. OpenAI says Astra is a separate case. Lawmakers are watching, too. According to WSJ, a Senate subcommittee is holding a hearing on rogue AI agents this week, and Florida’s Attorney General has been suing OpenAI since June.

What does this mean for you?

You won’t be getting GPT-6.1 Astra anytime soon. OpenAI hopes to reuse its base model to build future GPT-6 models and is digging into what caused the problems. Personally, I’d rather wait for an AI that’s honest about its work than use one that goes off and does its own thing.

In context: Android smartphone maker Nothing has launched the Headphone (1) Pro, its new flagship audio product, which it claims combines an elevated listening experience for consumers with studio-grade production and mixing features for professionals. It succeeds last year’s Nothing Headphone (1), which TIME named one of its Best Inventions of 2025.

The biggest change in the new model is a three-driver audio system. It includes a bass dynamic driver designed to add more punch and sensitivity to low frequencies, a precision dynamic driver that adds more texture to vocals and instruments, and a new treble xMEMS driver that Nothing says delivers crisper highs.

Nothing also worked with London’s Metropolis Studios to tune the Headphone (1) Pro, with the collaboration focused on professional audio tools and soundstage characteristics.

The company also highlights an improved adaptive active noise cancellation system that uses a 10-microphone array to more accurately filter ambient noise than the first-gen model. With claimed noise reduction of up to 46 dB, Nothing is positioning the Headphone (1) Pro against premium rivals from Bose and Sony.

Nothing says the improvement in noise reduction should be particularly noticeable in strong winds or noisy traffic. The headphones also feature redesigned ear cushions with an 8mm silicone baffle wall inside, creating a stronger sound barrier and improving the seal around the ears.

The Headphone (1) Pro supports Hi-Res wireless audio and wired playback over USB-C at up to 24-bit/192 kHz. It also offers five modes of Dynamic Spatial Audio with head tracking, along with five new custom EQ profiles. A separate Flat EQ switch is designed to give producers and creators a more neutral sound profile.

The Headphones 1 Pro is built out of aluminum and titanium to provide a durable shell while creating a premium look and feel. It also retains Nothing’s transparent design language, using shatter- and scratch-resistant 9H Panda Glass on both ear cups to reveal parts of the triple-driver architecture.

Despite the increased use of metal, Nothing has reduced the headphones’ weight to 327 grams through a redesigned internal architecture, including replacing steel arms with titanium. Other changes intended to improve comfort over longer listening sessions include a wider headband with thicker padding and softer ear cushions.

The Headphone (1) Pro is positioned to compete with premium models such as the Sony WH-1000XM6, Bose QuietComfort Ultra, and Apple AirPods Max. It is priced at $399 and will be available beginning September 29 through Nothing’s official store at nothing.tech, Amazon, and Best Buy stores across across the US and Canada.

Ned Jenkinson of the University of Birmingham and Matthew Weightman of the University of Oxford discuss how advancements in brain research might affect how we learn and grow our skillsets.

Whether learning a new piano piece or adapting your tennis serve, acquiring physical skills depends on your brain’s ability to strengthen and refine neural connections. Researchers are exploring whether this process can be accelerated with technology.

Scientists are particularly interested in the potential of non-invasive brain stimulation, a group of techniques that can alter brain activity without surgery.

Some deliver weak electrical currents to the brain through electrodes placed on the scalp. Others use magnetic fields or focused ultrasound waves. Although they work in different ways, they all aim to temporarily change the activity of neural circuits.

If these techniques can successfully enhance neuroplasticity, the brain’s ability to reorganise and form new connections during learning, then they could be of use anywhere where performance depends on learning complex movements, from sport and music to surgery and beyond. Researchers are also seeing if these technologies could help with learning non-physical skills, such as picking up a foreign language.

Elite sport, professional gaming and high-performance workplaces could all become targets for these enhancements if they prove effective. Brain stimulation could also have a big role to play in medicine, helping patients recover physical skills lost through injury or disease, such as stroke.

Studies suggest there’s a lot of potential here. But translating this potential into useful tech that reliably boosts learning physical skills remains a big challenge.

Stimulating findings

Research into enhancing motor learning with electric or magnetic stimulation has been gaining momentum since the turn of the millennium, with early studies garnering considerable excitement.

In a typical experiment, participants might learn a sequence of finger movements similar to practising scales on a piano while receiving stimulation over brain regions involved in movement. Other studies have examined how stimulation could be used for balance training or teaching sports-related skills or surgical techniques.

Some of these experiments produced eye-catching results, finding that participants learned certain movement tasks faster or retained skills for longer if they underwent brain stimulation. But other studies failed to find benefits. And in some cases, researchers struggled to replicate the success of earlier promising experiments when repeating them.

One reason for these mixed findings is that there’s no such thing as a universal ‘learning network’ in our brains. Different skills rely on different combinations of areas near the surface of the brain as well as those deep within it.

Additionally, people can respond very differently to the same stimulation. Factors such as age, anatomy, genetics and even baseline skill level may influence whether stimulation is beneficial. Add to that the infinite number of ways to apply stimulation, the picture becomes murkier.

Despite these challenges, the field continues to evolve in its quest to enhance motor learning. For instance, rather than broadly stimulating the brain, researchers are increasingly targeting specific neural circuits involved in learning.

This is partly thanks to advances in neuroimaging and computational modelling, which has allowed scientists to predict how electrical currents travel through a person’s brain. Newer brain stimulation technologies, such as focused ultrasound, can also now reach deep structures involved in skill acquisition.

The goal is to use these technologies not simply to increase brain activity, but to influence the right neural circuit at the right time during learning. This idea builds on a fundamental principle of neuroscience, often summarised as “neurons that fire together, wire together”. When brain cells are repeatedly activated at the same time, the connections between them become stronger.

By carefully timing stimulation to coincide with the movements made during practice, researchers hope to reinforce the neural pathways involved in learning a new skill. In principle, this could make stimulation more reliable and more effective than current approaches, but researchers are still fine tuning exactly how this would work.

Motoring ahead

Important questions remain. Who would have access? Should stimulation be regulated in competitive environments such as sport? And how much evidence should be required before consumer devices are marketed to healthy users?

These questions are becoming increasingly relevant as brain stimulation moves beyond the laboratory and clinic. A number of at-home devices are now available for people to buy. Some have received regulatory approval, as they’re indicated for treating medical conditions such as depression. But there’s also a growing market for devices for cognitive and performance enhancement. For these uses, no regulatory approval is needed.

The technology is advancing rapidly, but evidence to support it and regulations governing it are still trying to catch up. Proper frameworks for its adoption may simply be bypassed by the ready possibility of ‘DIY’ brain stimulation.

For now, brain stimulation is unlikely to transform anyone into an overnight virtuoso or elite athlete. But as researchers develop increasingly precise ways of targeting the neural circuits that underpin learning, the prospect of enhancing human performance is shifting from science fiction towards scientific possibility.

The challenge today is not simply learning how to influence the brain, but deciding where, when and why we should.

By Ned Jenkinson and Matthew Weightman

Ned Jenkinson is a senior lecturer in human movement sciences at the School of Sport and Exercise Sciences at the University of Birmingham. His research incorporates a range of techniques including non-invasive brain stimulation, electrophysiological recording, eye-tracking, neuroimaging and behavioural techniques. He uses these techniques to investigate how the brain controls movement and how it allows us to learn new motor skills. 

Matthew Weightman is a postdoctoral researcher at the Oxford Centre for Integrative Neuroimaging in the Plasticity Group at the University of Oxford, led by Prof Heidi Johansen-Berg. He is broadly interested in the field of sensorimotor neuroscience. His current work focuses on the role of sleep to recovery after stroke. More specifically, he is interested in how we can improve sleep after a stroke, whether improved sleep in stroke patients relates to better functional recovery, and if physiological processes that occur during sleep can be enhanced post-stroke to boost consolidation.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Keio Corporation (Keio), a major private railway operator in Japan, said its network was hit by a ransomware attack over the weekend, disrupting  some of its business systems.

Following a system failure in the early hours of Saturday, the company confirmed the attack and shut down its network to prevent additional damage.

The company said it is investigating the extent of the impact and whether the attackers accessed any customer or business partner information.

Keio is a large Japanese railway operator with 85 km of track and 69 stations, as well as a separate hospitality business of 25 hotels. The company has over 2,200 employees and a reported annual revenue of about $2.6 billion.

“In the early hours of September 26, 2026, we confirmed a ransomware attack on our group’s servers. We have reported the incident to the police and are conducting an investigation into the attack’s route and damage with the cooperation of external experts,” Keio says.

The incident appears to have affected only the hospitality side of Keio’s business, not train operations.

A separate announcement published on the company’s Keio Plaza Hotel Tokyo website is warning of possible delays on some customer-facing services.

Local media outlets have reported that the cyberattack disrupted the firm’s payment systems.

At the time of writing, BleepingComputer could not find a ransomware group claiming the attack on Keio.

BleepingComputer has contacted the company to request more information about the incident, and we will update this post with their response once it reaches us.

Tokyo Metro has also disclosed a cyber incident over the weekend in which attackers gained unauthorized access to its systems and accessed 59,000 member email addresses.

Although both Keio and Tokyo Metro are Japanese railway operators, it is unclear if the organizations were targeted in a coordinated campaign by the same threat actor.

Tokyo Metro is a major transit operator that runs nine subway lines covering 195 km and 180 stations, carrying an average of 7 million passengers daily.

The company said the breached systems contained only email addresses and that it has already identified and closed the security weakness the attackers used in this case.

Join Mikko Hyppönen and security leaders from the NFL, CHANEL, and Atlassian for a two-hour digital summit on what AI-speed attacks change, what defenders should stop doing, and how to validate, decide, fix, and re-validate at machine speed.

Save your seat

• Home puts chats, delegated work, and Office documents inside one interface
• Code lets non-programmers describe software and have Copilot build it
• Autopilot can continue recurring work without waiting for another instruction

Microsoft has introduced a redesigned version of its Copilot AI platform which claims to combine chat, delegated work, and coding tools into a unified application experience for users.

The company says the update is meant to let individuals and organizations scale artificial intelligence across everyday tasks and long-term projects.

Automate your Mac with ease using Shortcuts and Apple Intelligence.

Creating shortcuts on your Mac is easy

Filipe Espósito for Engadget

Make your shortcuts easier to reach

Filipe Espósito for Engadget

Peak XV Partners, one of the largest venture capital firms investing in markets including India and Southeast Asia with more than $10 billion in assets under management, has increased how much it invests per startup through Surge, its seed-stage investing platform, as it unveils a new cohort of 18 companies.

At least three of the companies in this cohort had already raised outside funding, in some cases from Peak XV itself, before joining Surge.

The new batch, called Surge 12, is the first to operate under Peak XV’s higher investment ceiling of up to $5 million per company, up from $3 million previously. The venture firm invested more than $50 million across the cohort, which has collectively raised over $90 million in seed funding, according to Peak XV. Its median investment per company has also increased, though the firm declined to disclose the figure.

“The bar to raise a Series A has gone up pretty significantly,” Rajan Anandan (pictured above), managing director at Peak XV, said in an interview. He added that the firm is also seeing more capital-intensive companies, particularly in deeptech, that are raising larger rounds at the seed stage.

Surge has become more global with each cohort, Anandan told TechCrunch, with its latest group spanning founders and companies from San Francisco to Sydney. Just five of the 18 startups in Surge 12 are focused on the Indian market, while more than half of the companies are based in India. The remaining 13 target global markets, highlighting the difference between where the companies are built and where they expect to find customers.

Since its launch in 2019, when Peak XV operated as Sequoia Capital India and Southeast Asia, Surge has backed more than 180 startups founded by entrepreneurs representing more than 18 nationalities. Peak XV says the 10 largest companies to emerge from those cohorts now generate more than $1 billion in combined annual revenue.

Anandan described Surge as one way Peak XV invests at the seed stage, alongside its standard seed investing, while the firm still remains an investor as companies progress through later funding rounds. The founders it backs typically include repeat entrepreneurs, experienced operators, and highly specialized technical founders, he said, with about 50% to 60% of a typical cohort made up of people coming from operating roles at established technology companies.

This cohort’s startups span AI, robotics, space, consumer products, healthcare, music, and fintech, ranging from AI safety and personal computing to autonomous robots built for underground pipes and satellites designed to detect radio-frequency signals from orbit.

The Surge 12 cohort

Alma — founded by Nischith Shadagopan M N and Vinod Ganesan — is building a personal computing platform focused on making computer use faster and more affordable. Its founders previously worked at Microsoft Research and were founding engineers at Sarvam AI, a Bengaluru-based startup building AI models for Indian languages.

August AI — founded by Anuruddh Mishra, an IIT-BHU alumnus who started the company in 2022 after a personal medical misdiagnosis — provides a healthcare platform that combines AI with physician-led care, reaching over 9 million users across 160 countries.

Ditto — founded by UC Berkeley dropouts Allen Wang and Eric Liu — works as an AI dating matchmaker inside iMessage, aimed at helping college students turn digital introductions into in-person connections. (TechCrunch wrote more about this one last month.) The company had already raised $9.2 million in a Peak XV-led seed round announced earlier this year.

GameStock — founded by Antoine Mistico, Easton Dana, and Vivek Indlebele Narasimha Prasad — brings competition mechanics to financial markets, turning investing and trading into a more competitive experience. Mistico is a two-time founder and former professional baseball player.

HiLoop — founded by Jad Ghalayini, Karan Brar, and Thomas Boser — helps AI companies adapt general-purpose open-weight models for specific applications using its post-training platform. Its founding team includes former Reducto engineers and a Cambridge computer science PhD who completed his doctorate at 24.

Hoola Health — founded by Deeksha Senguttuva — focuses on care for children and their families, providing consultations, vaccinations, medicines, diagnostics, developmental therapy, and dental services on a single platform. Senguttuvan grew up around healthcare, as her family built and operated a hospital group.

Kello — founded by Mona Gandhi and Subramanya Jingade — is building an AI-powered talent-discovery platform focused on identifying a candidate’s potential and trajectory rather than relying primarily on conventional credentials. Gandhi says she was Airbnb’s first female engineer and she previously founded Upraised, while Jingade previously co-founded AmbitionBox.

Kindling — founded by Adam Miller and Sachin Shah — is building what it calls a “storytelling operating system” for technology startups, using AI to help companies develop and produce their communications and content.

Puralink — founded by Harrison Crowe-Maxwell, Shyeon Delnawaz, and Thien “Long” Tran — is developing autonomous robots that can navigate underground pipe networks. Crowe-Maxwell has been building robots since childhood and turned university research into the patented drive technology behind the startup.

Reinforce Labs — founded by Anish Das Sarma — is developing tools to evaluate, red-team, and remediate enterprise AI systems. Sarma previously founded a company acquired by Airbnb and later served as a director at Google, where he led AI and machine-learning teams.

Riffle — founded by Anurag Choudhary and deo — is building a browser-based platform where musicians can create, collaborate on, and share music, reducing the need to move between separate tools during the creative process.

Rosella — founded by Chris Dwyer and Sean Stuart — is building an AI-native commercial insurance brokerage for U.S. businesses, using AI to automate parts of the traditionally manual process of finding and placing business insurance. Rosella raised a roughly $2.5 million pre-seed round led by Peak XV and Intact Private Capital earlier this year.

Tribe Money — founded by Himanshu Arora and Nikhil Shanker — gives an AI-powered personal finance platform that helps users track their money, research investments and make investing decisions.

ULOOK — founded by Adheesh Boratkar and Siddhesh Ravindra Naik — is building autonomous satellite systems for radio-frequency sensing and spectrum intelligence, targeting customers globally. Its founders have worked on more than 12 satellite missions. The company had already raised roughly $2.3 million in seed funding from growX Ventures and InfoEdge Ventures before joining Surge.

Wingit — founded by Nikunj Kothari and Saksham Khandelwal — is building a beauty platform aimed at India’s growing premium-consumer market. It is focused on how consumers discover and shop for higher-end beauty products.

Three other startups in the cohort have yet to publicly reveal their names or products. Peak XV said they are working in education, applied AI, and medical products.

The Metric Is Not the Mission is a ten-part examination of how Big Tech moved from building and expanding the open internet to increasingly shaping it around its own metrics, incentives and assumptions. Across the series, the argument follows the evolution of the platform economy—from the optimism of the early internet to the growing tensions around power, prediction, geopolitics, accountability and the future of digital life.

The series will be published in two parts each week over five weeks, with each installment building on the one before it. At the end of the series, the complete essay will be brought together in a single PDF edition, providing the full argument in one place.

Part III: When the Maps Became the Territory

In Part II, the story turned on a crucial distinction: measuring behavior is not the same as understanding people. Part III takes that idea further, examining what happens when the platforms’ representations of the world begin to substitute for the world itself.

There is a curious tendency among successful technologies to disappear. Not physically, of course, but cognitively. Once they become sufficiently embedded in everyday life, they cease to be experienced as technologies at all. Electricity is no longer a marvel of engineering but an expectation. We do not admire the plumbing each time we turn on a tap, nor do we reflect on the extraordinary complexity of global logistics every time fresh fruit appears on supermarket shelves in the middle of winter. The greatest infrastructures become invisible because they succeed so completely that we mistake them for part of the natural order.

The internet reached that point sometime during the second decade of the twenty-first century. Yet something else happened along the way that proved far more consequential. As the network itself faded into the background, the platforms through which most people experienced it moved decisively into the foreground. Increasingly, users no longer spoke about “going online.” They spoke about opening an app.

That linguistic shift deserves more attention than it usually receives. Language often reveals structural change before statistics do. To “browse the web” implied movement across an open landscape whose boundaries were undefined. One followed links, discovered obscure websites, stumbled upon ideas that had not been recommended by anyone, and occasionally became gloriously lost. The experience resembled wandering through an unfamiliar city with no particular destination in mind. Serendipity was not a flaw in the architecture; it was one of its defining virtues.

Applications altered that relationship almost without anyone noticing. They replaced geography with destination. Instead of entering a network whose possibilities remained unknown, we entered environments that had already been organized on our behalf. The internet did not disappear, but it became increasingly hidden beneath layers of interface, recommendation and curation. Like passengers traveling through an airport without ever seeing the city beyond the terminal, we continued moving through digital space while encountering only the carefully managed environments that had been prepared for us.

This transformation is often described as an inevitable consequence of convenience. While accurate in its own right, this explanation offers an incomplete narrative. Convenience was certainly the language through which the platforms justified many of their design choices. Friction was treated as the great enemy of the digital age. Every additional click became an obstacle to be eliminated. Every decision that users might otherwise make for themselves could instead be anticipated by software. Recommendation replaced search. Autoplay replaced choice. Infinite scrolling replaced endings. The future, we were told, belonged to experiences so seamless that they would feel almost effortless. And they did.

It is difficult to criticize convenience because convenience is genuinely valuable. Few people wish to return to an internet in which finding information required memorizing obscure web addresses or navigating labyrinthine directories. The platforms did not succeed by forcing people into inferior experiences. They succeeded because, for many years, they built better ones.

Yet convenience has always carried an intellectual cost. Every technology that removes friction also removes moments of deliberation. The elevator spares us the staircase but also the awareness of distance. Satellite navigation ensures that we rarely become lost, while quietly diminishing our ability to construct mental maps of the places through which we travel. Streaming services relieve us of searching for entertainment, but in doing so they also shape the boundaries of what we are likely to discover. Every act of technological simplification transfers a small measure of agency from the individual to the system.

The internet had originally been built on a different assumption. Its underlying protocols did remarkably little. They did not decide which website deserved prominence, which ideas should travel furthest, or which communities ought to flourish. Their genius lay precisely in their restraint. They created conditions under which others could innovate without first requesting permission. The web itself functioned less like a product than like a constitutional order: a simple framework within which extraordinary diversity could emerge.

Platforms gradually adopted the opposite philosophy. They did not merely provide the rules of the game; increasingly, they became active participants in every interaction taking place within it. They selected what deserved attention, inferred what users might prefer before users themselves knew it, prioritized certain relationships over others and determined, through millions of microscopic computational decisions, the contours of everyday experience. The architecture became less constitutional than managerial.

There is an illuminating parallel here with the history of cities. The most enduring cities are rarely the ones that have been planned in every detail. They are those that accumulated layer upon layer of human activity over centuries, adapting continuously to changing needs without ever fully surrendering their unpredictability. One finds unexpected bookshops beside cafés, workshops hidden behind apartment blocks, public squares appropriated for demonstrations one week and festivals the next. Their vitality emerges not from perfect organization but from the freedom they grant people to appropriate space in ways that planners never anticipated.

Shopping malls operate according to an altogether different logic. They are meticulously designed environments in which every entrance, corridor, sightline, and seating area has been carefully considered. Music, lighting, and architecture work together to produce an experience that feels spontaneous while being anything but. There is comfort in their orderliness. They are clean, efficient, and reassuringly predictable. Yet no one mistakes a shopping mall for a city. Its purpose is not to cultivate civic life but to optimize a particular set of behaviors within a privately governed space.

The analogy is imperfect, as all analogies are, but it captures something essential about the transformation of the internet. The early web invited participation because it remained fundamentally unfinished. It assumed that users would contribute to shaping it. Today’s dominant platforms present themselves as complete worlds. Participation still exists, but it takes place within boundaries established elsewhere. Users generate the content while the architecture remains firmly in corporate hands.

Perhaps this is why the language of “community” has begun to feel strangely hollow. Communities, in the classical sense, are rarely designed. They emerge through shared experience, mutual obligation, and a degree of unpredictability that no algorithm can fully reproduce. Platforms, by contrast, increasingly treat community as an engineering problem to be optimized. They recommend friendships, suggest conversations, rank relevance, suppress friction, and amplify interaction according to models whose objectives are necessarily commercial because the organizations that develop them are commercial enterprises.

None of this should be understood as an accusation of bad faith. Many of the engineers responsible for these systems genuinely believed they were improving people’s lives. The difficulty lies elsewhere. Every large institution eventually begins to confuse the optimization of its own internal metrics with the fulfillment of its original purpose. Universities sometimes mistake publication counts for scholarship. Hospitals occasionally confuse efficiency with care. Governments become preoccupied with administrative process rather than public service. Technology companies are no different. The indicators that make sense within an organization slowly become proxies for the world outside it. The metric is not the mission. This is the point at which the maps begin to replace the territory.

The extraordinary quantities of behavioral data collected by digital platforms produce an understandable confidence. When one can observe billions of interactions each day, it becomes tempting to believe that society itself has become legible. Human behavior appears measurable, predictable and, increasingly, governable. The platform begins to resemble reality because so much of reality passes through the platform.

Yet the map is never the territory. It captures what can be measured, not everything that matters. A map records roads but not the reasons people travel. It identifies cities without conveying the lives unfolding within them. Likewise, recommendation systems observe behavior with astonishing precision while remaining largely indifferent to experience itself. They recognize patterns without necessarily understanding meaning.

That distinction mattered little while the platforms continued solving the problems that had made them indispensable. It becomes far more consequential once they begin confronting a world that no longer resembles the one for which they were originally designed. Because societies have changed; politics has changed; and, the internet has changed. The question is whether the companies that grew powerful by interpreting one era have noticed that another has already begun.

Konstantinos Komaitis, PhD, is a veteran of developing and analysing Internet policy to ensure an open and global Internet.

Filed Under: behavior, big tech, metric not mission, open internet, optimization, platforms, understanding