An Uber driver works a city centre on a Tuesday afternoon. For three consecutive days, they have rejected low-paying assignments and accepted only premium fares. The platform notices the pattern. It responds by incrementally raising the offered payout on standard trips until the driver yields. Within days, the offers begin to decline again — in random decrements of five and ten cents, each engineered to sit below the threshold of human perception. The driver is soon working longer hours for less money, unable to identify the mechanism that produced the outcome. Legal scholar Vina Dubal named this practice algorithmic wage discrimination. It is not a malfunction. It is a feature, enabled by a capability that defines the modern internet: the ability to alter pricing, rankings, and payout structures at any moment, at scale, without consent. Technology writer Cory Doctorow gave the broader process a name in 2023. He called it enshittification.
Key Takeaways#
Enshittification is a three-stage structural lifecycle — user capture, business-customer extraction, shareholder harvest — that recurs across digital platforms regardless of founding culture or stated mission.
The mechanism is enabled by “twiddling”: the cloud-computing capacity to modify business logic continuously, invisibly, and without consent.
Google’s 2024 DOJ antitrust proceedings produced internal memos confirming that the deliberate degradation of search quality was a top-level revenue strategy, not an emergent accident.
Users remain on degraded platforms not from loyalty but from collective action failure: exit requires coordinated departure that the platform’s architecture is specifically designed to prevent.
Asymmetric accounting — rewarding measurable cost reductions while obscuring qualitative degradation — is the managerial logic that converts enshittification from an external structural pressure into an internal operational preference.
The enshittification of platform narratives mirrors the enshittification of platform services: Apple’s 2022 privacy campaign blocked competitor surveillance while expanding proprietary surveillance, demonstrating that the mechanism operates at the level of reputation as well as product.
When the Algorithm Learns Your Breaking Point#
The technical term for what happened to the Uber driver is twiddling. It refers to the infinite flexibility granted by cloud-based business logic: the capacity of a firm to reach into its operating system at any moment and alter a cost, a ranking, a payout, or a recommendation without notifying the party affected. A nineteenth-century factory owner seeking to calibrate wages to each individual worker’s desperation would have required a room full of accountants tracking every movement, every refusal, every moment of financial pressure. The surveillance cost would have made the scheme economically unviable. Cloud infrastructure performs the same surveillance automatically, continuously, and at near-zero marginal cost. The economics of individualized extraction flipped. What was once prohibitively expensive became the default operating mode of consumer-facing digital services.
Twiddling did not remain confined to gig-economy wage suppression. It became the universal operating logic of the attention economy. Search rankings are twiddled. Recommendation feeds are twiddled. Advertising auction prices are twiddled. The terms under which a small business reaches its customers are twiddled. The transformation of the internet from what Doctorow calls a “magic discovery machine” into an engine of continuous value extraction follows directly from this single capacity: the ability to adjust, invisibly and without consent, the terms of every transaction in real time.
A Lifecycle, Not a Malfunction#
Enshittification is a three-stage structural process that transforms user-centric services into extraction engines. The stages are not accidents or results of individual moral failure. They follow from a specific configuration of structural conditions: network lock-in sufficient to make exit costly, cloud infrastructure enabling frictionless adjustment, and accounting systems that reward measurable cost reduction while concealing qualitative degradation. Identifying the pattern requires no attribution of unusual greed. It requires reading the incentive structure that governs any platform that has achieved dominant market position without facing credible competitive threat. The men who run these companies are no more or less extractive than the railroad barons of the nineteenth century. The difference is the environment in which they operate.
The Mechanism of Managed Decay#
The Three-Stage Playbook#
Every major platform follows the same trajectory. In stage one, the platform behaves with exceptional generosity toward end users. The goal is not profit — it is capture. Google search in its early years was a genuine instrument of discovery. Type approximate terms into a minimal interface and precise answers returned, unencumbered by commercial interruption. Google minimized advertising in its first years and maximized engineering investment in result quality. The product was not charitable. It was bait. Between 2000 and 2021, Google spent tens of billions of dollars securing default placement on browsers and mobile operating systems, ensuring that the user encountered no friction and no alternative. The user did not choose Google because they had compared it to competitors and found it superior. They chose it because the path to a competitor had been purchased away. The trap was set structurally, before the product began to decay.
Once users are sufficiently locked in — once the cost of switching platforms is prohibitive in time, social capital, or institutional dependency — the platform enters stage two. It begins degrading the user experience to extract value for business customers: advertisers, publishers, vendors. For Google, this meant filling the search interface with paid placements, shrinking the visual distinction between advertisements and organic results, and demoting content that did not serve the advertising ecosystem. The degradation is gradual enough to fall below the threshold of individual response but cumulative enough to transform the product fundamentally over a period of years.
Stage two contains a structural trap for the business customers as well. A small retailer that depends on Google’s organic search for even ten percent of its revenue cannot afford to be algorithmically demoted. It becomes a hostage. It pays for placement to restore visibility it previously received for free. The supplier, like the user, is now locked in. Stage three follows logically: once both users and business customers are trapped, the platform extracts maximum value for its own shareholders, leaving behind the minimum residue of utility required to prevent a mass exodus. Doctorow’s term for this residue is precise: homeopathic. The dose is calculated, not incidental.
The Internal Memo and the Collective Cage#
The 2024 Department of Justice antitrust proceedings against Google produced documentary evidence confirming that the mechanism is understood and managed internally. Internal memos revealed that by 2019, Google had achieved over ninety percent of the global search market and had exhausted the pool of new users to convert. Wall Street’s structural demand for continuous growth remained operative regardless of this saturation. The solution proposed by Prabhakar Raghavan, then head of search revenue, was structurally precise and qualitatively destructive: deliberate degradation of result quality. A search system calibrated to return a suboptimal answer requires the user to search again. Each additional query generates an additional page view. Each page view generates an additional advertising impression. The revenue model is improved by engineering failure into the product.
Ben Gomes, the engineer who had built Google’s foundational data infrastructure, reportedly opposed the proposal on product-quality grounds. Raghavan’s faction prevailed. Today, a typical Google search returns an AI-generated summary of uncertain provenance, followed by multiple paid placements, sitting above a page populated largely by content produced specifically to game algorithmic ranking rather than to serve user queries. The degradation is not subtle. It is the structural output of a deliberate internal decision documented in corporate communications and confirmed in federal proceedings.
Users remain despite this degradation not from loyalty but from collective action failure. Individual dissatisfaction is widespread and documentable. Coordinating a simultaneous departure is a different problem. Departure requires agreement on a destination, a mechanism for reconstructing social networks built on the incumbent platform, and a critical mass of simultaneous movers sufficient to make the new platform viable. The incumbent holds the social graph — the accumulated record of connections, communications, and shared history that constitutes the user’s actual stake in the system. Exiting means abandoning that graph. The platform’s most durable competitive advantage is not its search algorithm or its recommendation engine. It is the cost it has imposed on leaving.
Apple’s 2022 privacy campaign illustrates that enshittification operates at the level of narrative as readily as at the level of product. Apple’s App Tracking Transparency feature, introduced in iOS 14.5, allowed users to opt out of cross-application behavioral tracking by third parties. Ninety-six percent of Apple users activated the opt-out. Apple marketed the feature as a privacy landmark. Simultaneously, Apple shipped a parallel system that expanded the company’s own first-party behavioral surveillance — tracking application usage, search behavior, and purchasing patterns to feed Apple’s proprietary advertising network, without opt-out functionality and without disclosure. The company that positioned itself as the privacy-respecting alternative to the surveillance economy was running a comparable surveillance operation under a different brand. The product of the digital age, confirmed by this episode, is not the person who fails to pay. It is anyone who can be productized. Everyone gets twiddled.
The Spreadsheet Always Wins#
The advertising executive Rory Sutherland describes the organizational logic that produces enshittification from the inside with a metaphor he calls the doorman fallacy. A hotel employs a doorman at an annual salary of fifty thousand dollars. A consulting firm, engaged to identify inefficiencies, observes that the doorman’s measurable function is opening the door. It recommends replacement with an automatic motion sensor. The hotel saves fifty thousand dollars. The consultants log a success. Five years later, the hotel has quietly lost its most valuable clientele, the lobby has become unwelcoming, and the accumulated revenue loss exceeds the saved salary by an order of magnitude. The executives who authorized the sensor are not held accountable for the revenue destruction, because the causal chain between a staffing decision and a reputation decline five years later is too diffuse to attribute. The cost saving, however, is immediate, numerically precise, and career-advancing.
Self-checkout technology in retail follows the same logic at industrial scale. Introduced as a convenience for small purchases, kiosks were gradually converted into an obligation — the labor displacement producing a visible payroll reduction on quarterly reports. The immediate result was a significant increase in retail theft, with major chains reporting inventory discrepancies far exceeding prior loss rates as customers — intentionally and not — exploited systems with minimal supervision. The retailer traded a reliable workforce for an adversarial relationship with its own customers. The payroll line improved. The inventory loss line deteriorated. The quarterly report captured the former. It distributed the latter across product categories where attribution was difficult.
The asymmetric accountability that the doorman fallacy describes is not a management pathology confined to particular firms or industries. It is the predictable output of any measurement system that can capture cost reductions precisely and attribute qualitative degradation only diffusely. When executives are evaluated on headcount reduction and short-cycle revenue, the system will consistently generate outcomes that destroy the conditions of long-term value. Enshittification at the platform level and the doorman fallacy at the organizational level are the same mechanism operating at different scales.
Diagnostic, Not Nostalgic#
Enshittification is not a product of unusual corporate greed. It is the predictable output of a specific structural configuration: network effects that make exit costly, cloud infrastructure that makes continuous extraction frictionless, and accounting systems that reward measurable cost reduction while obscuring qualitative degradation. The Uber driver who cannot trace the source of declining wages, the Google user who cannot identify why search results became unreliable, and the retail customer who cannot explain why the checkout process became adversarial are experiencing the same mechanism from different structural positions.
The historical comparison clarifies the stakes. The Sherman Antitrust Act of 1890 was not a moral condemnation of nineteenth-century industrialists. It was a structural response to a structural problem: that concentrated commercial control over daily necessities produces outcomes incompatible with the conditions required for a functioning market and a non-coercive social order. The legislative reasoning was explicit — if political autocracy is illegitimate, commercial autocracy over transportation and daily goods should be equally illegitimate. The capacity to twiddle represents a form of commercial autocracy over necessities of daily life — social communication, commercial search, economic participation in the gig economy — that did not exist in 1890 and for which no equivalent institutional response has yet been constructed.
The mechanism is now named and documented. The question is whether the structural response will match it. Part 2 of this series examines the three constraints that once disciplined corporate behavior — competitive pressure, regulatory enforcement, and organized worker leverage — and identifies the specific policy decisions that removed each one.
