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 II — When They Still Understood Us

Part I looked at the slow transformation taking place beneath the daily controversies surrounding Big Tech. This second part goes back to the beginning, asking what these companies originally understood about people and how the metrics that once measured their success gradually became the definition of it.

It has become fashionable to tell the story of Big Tech as though it were always destined to end here. In retrospect, it is easy to portray the rise of the major platforms as the inescapable march of surveillance capitalism, monopolistic ambition, and unchecked technological power. That narrative is emotionally satisfying because it offers clear villains and a comforting sense of inevitability. It also happens to be incomplete.

Cory Doctorow has given this deterioration a memorable name: “enshittification.” His argument is that platforms initially serve users well, then, once users and business customers are locked in, progressively shift value away from both toward shareholders, degrading the service in the process. It is a powerful account of how platforms become extractive. But it is not quite the argument here. The deeper problem is not simply that Big Tech has learned to extract more from us but that it has become increasingly convinced that because it can measure and predict our behavior, it understands us and, by extension, the societies it has come to mediate. The failure is therefore not only economic; it is also epistemic. The metric has become a substitute for the mission.

One cannot understand why these companies now appear increasingly disconnected from the societies they helped shape without first acknowledging that, for a remarkably long time, they understood those societies exceptionally well.

Technology succeeds when it solves technical problems. It changes the world when it solves human ones.

That was the genius of the first generation of internet platforms. Their founders did not invent friendship, curiosity, creativity or community. They simply recognized that the internet had reached a stage where these deeply human instincts required new forms of expression. The web of the late 1990s was exhilarating, but it was also fragmented, uneven and, for many people, intimidating. Finding information often required patience. Discovering interesting websites depended on chance as much as design. Publishing demanded a degree of technical literacy that excluded far more people than it empowered. The internet was open, but openness alone does not necessarily produce accessibility.

The great platforms emerged not because they sought to replace the internet but because they made it intelligible. Google transformed an expanding wilderness of information into something navigable. Wikipedia demonstrated that knowledge could be organized through collaboration rather than hierarchy. YouTube lowered the barriers to publishing so dramatically that expertise escaped universities, broadcasters and production studios. Facebook addressed an even more fundamental challenge. It recognized that the internet was no longer simply about information; it had become about people. Until then, maintaining relationships online had been surprisingly cumbersome. Email was too formal, instant messaging too ephemeral, personal websites too static. Facebook reduced social interaction to something almost frictionless. Its success lay not in technological sophistication but in psychological intuition.

This is easy to forget because the platforms that dominate our lives today bear only a partial resemblance to the ones that first captured our imagination. Facebook did not begin as an endless stream of algorithmically selected content. It was, in essence, a digital address book enriched by photographs, conversations, and the ordinary rituals of everyday life. It became valuable because it mirrored existing relationships rather than attempting to manufacture new ones. There was comfort in discovering former classmates, following the lives of distant relatives, or organizing gatherings that would otherwise have required dozens of emails and phone calls. The platform expanded social life without yet attempting to redefine it.

YouTube offered a similarly modest promise. It was not originally designed to maximize engagement or optimize watch time. It functioned more like an immense public archive whose value derived from its unpredictability. One could arrive searching for a lecture on astronomy and leave having discovered a forgotten jazz performance, a documentary on Greek history, or a repair manual for a washing machine. Recommendation existed, but it remained subordinate to curiosity. Users still felt as though they were exploring rather than being guided.

Even Twitter (now X), before it became a battleground for politics, culture wars, and performative outrage, captured something important about the changing nature of public conversation. It collapsed distance between journalists, academics, politicians, and ordinary citizens in ways that would have seemed extraordinary only a few years earlier. For all its imperfections, it suggested that expertise and authority might become more accessible rather than less.

Looking back, what united these companies was not simply technological innovation but a particular philosophy of the internet. They assumed that openness generated value. The more people connected, the richer the network became. Every new participant increased the possibilities for everyone else. Economists describe this as a network effect, but the phrase barely captures its cultural significance. Participation itself became the source of optimism. The internet appeared to be validating one of the oldest liberal ideas: that societies flourish when individuals are free to exchange ideas, collaborate voluntarily, and build institutions from the bottom up.

It is difficult to overstate how persuasive this vision became. Most governments celebrated the digital economy as an engine of innovation. Investors poured unprecedented sums into technology because the opportunities seemed limitless. Civil society organizations embraced online platforms as tools for democratic participation and global advocacy. Even critics of globalization often regarded the internet as an exception, a domain where openness appeared to distribute power rather than concentrate it.

For a brief historical moment, these interests aligned. What was good for technology companies often appeared to be good for users, for markets, and, in many respects, for the internet itself. The incentives reinforced one another. Companies grew by making the network more useful. Users benefited from larger communities. Developers built new services on open standards. The web expanded because success depended on drawing people further into its richness rather than confining them within a single destination.

History, however, has an inconvenient habit of changing the problems that institutions are asked to solve. The sociologist Robert K. Merton once observed that organizations often become prisoners of their own success. Practices that were rational under one set of conditions gradually harden into routines, and routines into orthodoxies. Institutions continue refining the solutions that once made them indispensable even as the environment around them evolves. Success breeds confidence; confidence breeds certainty; certainty eventually makes adaptation more difficult than persistence.

There is no reason to believe technology companies are exempt from this pattern. If anything, their extraordinary success may have accelerated it.

The platforms that once competed to help users navigate an open internet eventually found themselves managing ecosystems of unprecedented scale. Their priorities changed almost imperceptibly. The models they developed during the internet’s age of expansion proved astonishingly effective at connecting people, organizing information, and lowering the costs of participation. The metrics through which they evaluated success, such as growth, engagement, scale, and network effects, were not arbitrary inventions of venture capital. They reflected a period during which connecting more people genuinely created more value for everyone involved. The problem is that the world changed while the metrics remained stable. A measure that once indicated success gradually became the definition of success itself.

There is an obvious parallel here with Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure. The principle, first articulated by economist Charles Goodhart in the context of monetary policy, describes what happens when an indicator that works as a proxy for an underlying objective is turned into the objective itself. But the problem here is slightly different. The issue is not simply that platforms began gaming their metrics, or that users learned to optimize for them. It is that the metrics gradually became so deeply embedded in the companies’ understanding of success that the distinction between the measure and the mission was lost. The metric did not merely distort the objective; it quietly became the objective.

None of this happened because a group of executives gathered in a boardroom and decided to undermine the open internet. Institutional change is rarely so theatrical. More often, it emerges from countless rational decisions made in pursuit of perfectly reasonable objectives: improve the user experience, reduce friction, personalize recommendations, increase safety, remove inconvenience. Each adjustment appears modest in isolation but collectively they alter the character of the system itself.

This is where history becomes quietly ironic. The companies that had once understood the internet better than anyone else slowly began to forget what had made the internet exceptional in the first place. They continued to believe they were connecting the world, even as they increasingly replaced the world with carefully curated representations of it. They accumulated unprecedented quantities of information about human behavior while becoming progressively less attentive to the human condition.

The distinction is subtle, but it may prove to be the defining story of this technological era. To observe behavior is not the same as understanding experience. A platform can know how long we hesitate before clicking a link, which videos hold our attention for an extra seven seconds, or what sequence of images is most likely to keep us scrolling late into the night. It can infer preferences with astonishing accuracy. It can predict patterns that would have been unimaginable a generation ago. Yet prediction, however sophisticated, remains an impoverished form of understanding. It reveals what people do. It says far less about why they do it, what they fear, what they hope for, or what kind of society they are trying to build together.

That difference, almost invisible at first, is where the story begins to change.

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

Filed Under: big tech, enshittification, goodhart’s law, history, metric not mission, open internet

Anthropic now allows you to run Claude Code via cloud sessions without signing up for the research preview, and it’s offering up to $250 in free promotional credits.

Cloud sessions run Claude Code on Anthropic’s infrastructure instead of your own computer, so you can start a task, leave it running remotely, and return later to review the work.

“Cloud sessions run on Anthropic-hosted infrastructure, so the work keeps going even without your computer running,” Anthropic explained.

In our tests, we observed that you can start a cloud session from claude.ai/code, the Code section of the Claude mobile app, the desktop app, or the CLI using claude --cloud.

Claude Code’s cloud sessions have been available to some users via research preview, but now they’re officially available to eligible subscribers, and Anthropic is offering free usage credits to encourage existing subscribers to try them.

Anthropic explains how you can claim free credits

Anthropic is giving eligible Pro users $100 in promotional cloud-session credits, while Max subscribers get $250.

However, it is worth noting that the credits are separate from normal Claude usage limits and are applied automatically when you start a cloud session.

“If you hit a limit locally, keep going in the cloud until your credit runs out,” Anthropic said.

If you like the idea of Claude Code’s cloud sessions, you can claim the offer from Claude’s website by October 7, and any remaining balance expires on November 4.

Once the promotional balance is exhausted, cloud sessions go back to counting against your normal plan limits. There’s no separate charge for the cloud container itself.

The offer is limited to individual Pro and Max subscribers who had an active subscription when the promotion began on September 23.

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

• Hacktron chained a libheif heap overflow, reached through ImageMagick on OpenAI’s Discourse forum, leveraging an OpenAI SSO flaw to briefly take over employee ChatGPT and Codex accounts
• The researchers themselves invoked XKCD #2347, whose 2020 alt text happens to name ImageMagick as the dependency that will one day break
• ImageMagick served as only the pathway to the actual vulnerable component, libheif, an obscure decoder pulled in indirectly across Slack, Meta, and GitHub Enterprise amongst other mediums

When Hacktron AI recently disclosed its months-long libheif research, the researchers reached for a familiar picture that also, to some degree, hints at what let them break into OpenAI in the first place.

They pointed readers to xkcd #2347, Randall Munroe’s 2020 iconic web cartoon of all modern digital infrastructure balanced on a single load-bearing block that some random person in Nebraska has been thanklessly maintaining.

The Cricut StickerPix Print and Print + Cut could birth a generation of sticker moguls. 

Rolling down a hallway with a shirt slung over its neck, Goose looks less like a factory demo and more like a small appliance that decided to grow a personality. Prototype 1 from San Francisco’s The Robot Works is a white, low, rounded base with a long articulated neck, camera eyes set in a bonnet so you can see where it is looking, and a bright orange beak that doubles as a gripper. A blue polka-dot wrap gives the machine a costume instead of a chassis. Founders posted the first public clip this week with a simple brief: tidy, play, live.

Richard Wei has spent months sitting down with families to discuss what they really need from a robot in their house, and the team has used that information to create something they would like to have themselves. Humphrey Hu, who formerly worked with Anki on making highly expressive robot faces, will join him on this project. After considering all of the jobs that tend to build up around the house, as well as how a machine might interact with children and grandparents, they decided against a human-shaped body. Instead, the design features a wide stance close to the floor to keep the creature low to the ground, as well as a long neck that allows it to reach for items such as a sock, a wall outlet, a door handle, or a pile of laundry without having to stand on two legs and tower over the space.

LEGO Disney & Pixar Wall-E & EVE Building Set for Adults, Ages 18+ – Home Office, Book Shelf, or Room…

• LEGO SET FOR ADULTS – The WALL-E and EVE (43279) building set offers adults 18 years old and up an immersive construction challenge featuring…
• 4 DISNEY PIXAR CHARACTERS – Builders can create iconic robots WALL-E, EVE, M-O and Hal from the hit movie—each with authentic functionality like…
• MINDFUL BUILDING EXPERIENCE – This detailed construction set lets builders practice advanced construction techniques for an immersive and relaxing…

They also kept the hardware minimal on purpose, resulting in a robot that rolls, looks around, and then moves on to complete its job. It analyzes its blind spots before moving, and the motors are meant to give way if something like a foot, a chair, or a child gets in the way, and the ability to halt and send it somewhere else was purposefully built in, so it wasn’t an afterthought. A video released shows the robot performing things like placing a man’s foot onto an ottoman, gathering clothes off the hardwood floor, reaching up to a kitchen counter, inserting a plug into a wall outlet, and even pulling open a door when some relatives arrive with children.

Some early write-ups also highlighted some low-key jobs for older individuals who may be less mobile, such as fetching something you’ve tagged, bringing a walker over to the bedside when it’s time for bed, lighting the route to the toilet, keeping an eye out for falls, and sending a ping to family members. Some early elevator and door operations still need a human on the other end of the line, but on the positive side, autonomy is real in some areas and borrowed in others. Which is reasonable for the first version of the product.

The concept is for the robot’s capabilities to evolve as a library rather than coming with a predefined set of talents when you buy it, and they intend to vet each new software before it is released, allowing you to choose what it can see and do. The company is now selling services such as a full-room tidy, checking in on Grandpa, being a companion during story time, taking a look around the house while you’re away, checking the stove, and bringing a parcel inside. The pitch is that life evolves, thus the machine should be able to continually adding new jobs to its repertoire rather of becoming locked in one specific role. That’s the whole point of “built to grow.” They also claim that any household data will remain under your control, rather than being a default action that occurs automatically when the robot arrives.

Unfortunately, orders are currently closed because the initial prototype is still pre-production, with no price or commercial warranty available. Wei says the company is looking for “foster families” that will let Goose learn on real floors, real clutter, and real schedules. You can contact them at therobotworks.ai, while more information will be provided as the device gets closer to launch.

If there was one thing that was obvious this from Meta Connect this year, it’s that the social media giant is all-in on its burgeoning line of smart glasses.

Indeed, the glasses were pretty much everywhere at the annual event, where Meta shows off its newest hardware and AI products. Both Meta staff and the flocks of influencers who frequent the event seemed to arrive with the glasses glued to their faces. Most of the demos the company offered this week also involved the glasses.

I swiftly joined the bespectacled masses and found myself trying on pair after pair of Meta’s high-tech specs.

One of the more notable products I had an opportunity to demo were Meta’s new, still-unreleased, audio-only smart glasses. News of these glasses emerged not long after the company weathered accusations that it was selling “pervert glasses,” with critics alleging that its camera-equipped specs could be used for nefarious surveillance purposes.

The new glasses come equipped with six microphones, but no camera, and they have no native way to record your surroundings, which should go a long way toward easing privacy concerns. They are significantly lighter than any of the other smart glasses I’ve worn, and they were quite comfortable.

The Meta staffer I talked to emphasized the glasses’ entertainment and communication options: you can listen to music easily (they’re more comfortable than earbuds) and take phone calls without reaching for your phone.

The glasses will also integrate with Muse, Meta’s personal agentic system, which can carry out tasks on the user’s behalf. The new integration, which isn’t yet available to the public, lets wearers speak to Muse and give it commands verbally. Meta also plans to personalize the agent, letting users pick from an assortment of cute digital avatars that can represent the agent when it communicates with them.

I was given an opportunity to try this out, which was fun but also somewhat comical. You have to be very direct with the agent, and you can’t talk to anyone else at the same time or it will get confused. Problematically, I kept chatting intermittently with the Meta staffer who’d given me the glasses, and the agent kept thinking I was speaking to it, so it would talk over her.

Still, in the right context, it’s easy to see how this new integration could be incredibly useful. If you don’t mind being seen in public talking to your own sunglasses, you’ll be able to ask Muse to take care of various digital tasks for you, like sending emails or reciting to-do lists, and it will handle them while you’re out grocery shopping or having a beer. You can also ask the glasses anything, and like a mobile version of ChatGPT, they’ll spit out an answer. (I asked the glasses a question about World War II, and they gave me a succinct and historically accurate response.)

Finally, I also got to test-drive a second audio-only pair: Meta’s new glasses for the hearing impaired. Given that hearing loss affects many families (some 50 million Americans are said to have some level of hearing loss), this device, unlike a lot of other modern tech gadgets, serves a clear and practical purpose.

I spoke briefly with a member of Meta’s research team who said the glasses had been in development for some five years, and he pointed out a price difference that could make them appealing: whereas a lot of hearing aids can run as high as $1,600, the glasses will sell for $150.

The experience of wearing these glasses was interesting. Meta had me put earplugs in before trying them on to simulate the hearing loss that the glasses are meant to help users overcome. Once the glasses were on, they seemed to amplify the voice of the person I was talking to. Users can switch the amplification from focused, which zeroes in on the person in front of them, to omnidirectional, which picks up sound from all around them, depending on how they want to experience their surroundings.

Mark Zuckerberg has made it clear that he believes smart glasses are the future, and it’s evident that his company is doing its best to fulfill that mandate. But smart glasses remain a niche that has yet to truly find its footing. What was most obvious at Connect is that Meta has — in an attempt to succeed where others have failed — cast a very wide net, trying to make its glasses stylish, functional, and, most of all, useful.

Is this the future? Unsurprisingly, everyone at Connect seemed to think so. I suppose we’ll have to see if the rest of the world follows suit.

The company is still paying for the Cambridge Analytica scandal.

Tesla has begun volume production of the Semi at Gigafactory Nevada with first deliveries this week, nine years after showing the truck, and rates it at 500 miles of range and 1.2 megawatts of charging. It has already installed more than 100 megawatt-capable stalls at 21 European sites, although the truck is not due in Europe until 2027.

Tesla has started volume production of the Semi at Gigafactory Nevada, with the first deliveries this week, InsideEVs reported. The Class 8 truck was first shown in 2017, nine years ago. It charges at up to 1.2 megawatts, which Tesla says adds 60% of range in 30 minutes.

The range comes with conditions.

Tesla rates the Long Range at 500 miles and the Standard Range at 325, both at 1.7 kWh per mile, and publishes the conditions alongside them. They are measured at 82,000 pounds gross, 55 miles an hour, 20 degrees Celsius, on flat ground, with a tandem axle aero kit and a compliant dry van trailer. Drive power is up to 800 kW from three motors.

Europe waits until 2027.

Tesla is installing more than 100 megawatt-capable stalls across 21 European sites in Germany, France, Britain, Sweden and the Benelux countries. The truck itself is not due here until next year. Both truck and charger use the MCS 3.2 connector.

European law asks for less.

Charging points of at least 350 kW must sit every 60 km along the core European network for heavy goods vehicles, as the ICCT has documented. The megawatt layer above that is commercial rather than required. Tesla built it before it had a truck to sell here.

The battery figures are not Tesla’s.

A filing with the California Air Resources Board gives 822 kWh of usable capacity for the Long Range and 548 kWh for the Standard, neither of which Tesla has published. The Semi uses Tesla’s own 4680 cells from Gigafactory Texas. Those cells have underperformed supplier cells in its passenger cars.

The biggest buyer is Swedish.

Einride ordered 500 Semis last month, the largest deployment announced anywhere, to run freight in California, Texas, New Jersey, Illinois and Georgia. A Swedish company is putting American trucks on American roads. Amazon holds warrants over about 12% of it.

Diesel is doing the selling.

Pump prices are averaging over $6.50 a gallon in the United States, a record driven by the wars in Ukraine and Iran. A coalition of shipping companies ordered 2,500 Semis this week, which would roughly double the American electric truck fleet. Tesla wants 50,000 a year once Nevada is at full capacity.