Looking at the dim, dark gray room before you, nearly devoid of anything, you can only sink into deep despair. This is a shadow forgotten by civilization. The darkness here has substance — it clings to your skin like thick ink. Wrapped in the musty darkness, rust stains, and heavy walls, as long as you still have senses, you’ll realize you are a prisoner. The barred window is the only place where light enters — a small, square grate high above, teasing you. You’d better not dream of escaping through it. The occasional passing guard won’t regard you as human — he is the embodiment of violence. Fortunately, he himself isn’t too keen on frequenting your shabby, miserable little space.
This is a description of a traditional prison. When I try to construct a prison scene in my mind, I think of the color palette from the video game Buckshot Roulette: the space is suffused with oxidized reddish-brown and deep ochre, the light seemingly full of coarse noise. It’s a heavy, chaotic, crude, violent, and lifeless palette. The image is so vivid that I feel like thinking about it any longer would give me tetanus.
Now continue your story. The guard comes and tells you it’s time to change cells. You’re brought to a white cell. There are no rust stains, no musty smell. It looks very modern, fully equipped, brightly lit, with the cell door replaced by a massive horizontal floor-to-ceiling window. The scene somewhat resembles Le Corbusier’s Villa Savoye. The guard says he’s not in charge of managing this area, and no guards will come for inspections. You feel as though you’ve suddenly gone from hell to heaven — you can stretch your limbs, do what you want, fully enjoy your solitude. At this moment, you’re suddenly grateful for modern civilization’s humanitarianism. Only at the same time, you notice that in the distance, more than a hundred meters from the floor-to-ceiling window, stands a tower.
That tower stands a hundred meters away, like a black nail driven into the pure white earth. It is silent to the point of being sacred, with no rotating radar or blinding searchlights — just a ring of dark, emotionless slits, gazing in this direction in the afternoon’s intense light. You begin repeatedly calculating the angle between that dark window and yourself. Is there really someone in the tower? Would they see you doing something private? Once this thought takes root, it’s planted in your brain like a virus, impossible to shake. Sometimes you want to turn off the lights. But you discover that in this minimalist, civilized, divinely-inspired space, you cannot find a single switch to control the light.
Where you are now is the Panopticon. This is an architectural model conceived by the 18th-century British philosopher Jeremy Bentham, later developed by French philosopher Michel Foucault into a metaphor for social power. It is an extremely efficient model: without dense guard patrols, theoretically just one person (or even no one) can monitor thousands of prisoners. It requires no complex escape-prevention facilities, and physical suffering is reduced. If continuous surveillance can make criminals develop habits of diligence and law-abidingness, then society gains cheaper labor and a safer environment. From a utilitarian perspective, it’s truly perfect — no wonder its creator Bentham proudly said it had extreme versatility, and this model could be directly transferred to schools, factories, hospitals, and other places, improving the efficiency of anything (such as learning, production, recovery).
I take a sip of coffee and look up. My gaze turns toward the window — there’s no tower out there, no observer. It’s midnight, and on my screen are a distributed systems project and a technical blog post. The blog discusses the evolution of software system architecture — from monolithic, to microservices, to containerized orchestration, to serverless architecture. I’m astonished to discover that power structures have evolved in exactly the same way.
The Panopticon that Foucault described exists within a disciplinary society. Its characteristic is the enclosed space. Individuals continuously transition from one enclosed environment (such as family, school, factory, hospital, prison) to another. Each environment functions like an independent mold, shaping individuals with its unique rules and discipline. This exercise of power is discontinuous — entering a new environment means starting over.
French philosopher Gilles Deleuze, a contemporary of Foucault, pointed out that Foucault’s theory has limitations. In reality, these spaces are losing their boundaries. He proposed the concept of the Society of Control: the control society no longer relies on physical enclosed spaces but operates through open, continuous, and networked systems. Power becomes invisible, fluid, and omnipresent, penetrating walls — a form of continuous modulation. Unlike the disciplinary society where “numbered individuals are subjected to discipline and punishment,” in the control society, short-term but continuous, ever-changing control and modulation replace discipline. This approach is more flexible, more boundaryless; it appears to offer more freedom but comes with more refined, more comprehensive surveillance and regulation. In the control society, you can never truly complete anything: learning doesn’t end with graduation, the end of one work project is the beginning of another. The current architecture is serverless — no fixed physical servers, only fluid, on-demand data flows.
I realize that my own work objective seems to be becoming the most powerful warden possible. To construct my system, I constantly record, schedule, decouple, atomize, and practice distributed governance. Although I’m currently using simulated data and hypothetical business scenarios, once I work at a real company and touch production environments, isn’t everything I do essentially dismembering living human beings with my own hands?
In the code before me, I see only JSON objects or metadata — strings of decoupled user_ids, consumption preferences, geographic locations, and genders. When I split user data across different microservices for system throughput and high availability, am I ontologically destroying the integrity of “personhood”? I don’t care whether a person exists here; I only care about this dataset. These metadata are like Deleuze’s concept of the “Dividual”: no longer a complete, indivisible subject, but something disassembled into computable, storable, schedulable data points. I had always believed that what I was building was a form of common good. I’m conserving maximum resources, squeezing performance to its limits, ensuring the system is reliable and efficient. But does this system truly benefit humanity? Or am I merely acting out of egoism, deepening the system’s power just to claim my share in this controlled-to-death society? Am I also coveting the technological power this system possesses?
The expansion of this power relies far from merely on coercive violence. In fact, it is based on a more covert form of complicity. In digital ecosystems, there are countless “default behaviors” and “protocols” that users rarely notice, yet these determine users’ existential conditions in the digital world. In classical political philosophy, social contracts imply that equal individuals voluntarily surrender certain rights (in the digital age, privacy rights and data ownership) for mutual security or benefit. But in reality, “Super Apps” have constructed enclosed digital ecosystems. Within these ecosystems, this contract is invalid — even fraudulent.
Take China’s super app WeChat as an example. This app simultaneously bundles social networking, payments, transportation, identity verification, and a series of other functions. It is no longer a tool but a life infrastructure — one could even say it is an extension of the human body’s organs. In China, virtually everyone uses WeChat for social interaction, with monthly active accounts reaching 1.4 billion. People don’t need to use multiple different applications like Paypal, Uber, Slack, or WhatsApp — WeChat alone suffices: teachers may use WeChat to assign homework, friends communicate via WeChat, during the pandemic people used WeChat’s health codes for identity verification, and going out you can pay directly with WeChat. The number of “Mini Programs” within WeChat has exceeded 4.3 million — by comparison, the App Store’s total number of apps is roughly half that. Under these circumstances, rejecting the terms means “digital death,” or very nearly “social death.” Users have no power to opt out. Therefore, that smooth clicking action, carefully designed and guided by UI, is no longer an expression of free will but a helpless surrender locked in by technology.
This inescapable situation has caused a collective psychological pathology. We — both I as the architect and they as users — are all trapped in what Sartre described as bad faith. Users pretend that clicking “Agree” is an act of free will, pretend that the privacy surrendered for convenience (facial data, social networks, identity information) is merely insignificant digital debris. And I, as a developer, use the technical terminology of “optimizing user experience” and “reducing latency” to mask the reality: I am exploiting monopolistic positions to strip them of their bargaining power, enslaving their attention through dopamine mechanisms, and using technological evolution to conduct large-scale, distributed modulation and discipline upon them. In each other’s self-deception, we collectively maintain the operation of this prison without walls.
But this isn’t even the worst of it. To maintain this operation, to make data flow faster through my distributed system, the system must eliminate all “friction.” In Silicon Valley’s lexicon, “Frictionless” is regarded as the highest virtue. From a utilitarian perspective, this means maximizing efficiency. With the development of the internet, everything is accelerating in turnover and circulation. From the early internet’s method of actively inputting keywords and browsing pages, it has evolved into passive algorithmic recommendations and swiping — any pause or break in the information-feeding pathway is erased. From the payment method of pulling out a wallet and counting bills and coins, it has evolved into “just look” or “just tap” payment methods via FaceID or NFC. UX designers have meticulously designed everything so you consume information endlessly and guilt-free within software. You no longer have moments of contemplation or staring at a blank screen, nor the silence and tangible feeling before making a payment — everything comes and goes weightlessly. Technology has filtered out all your hesitation, reflection, and resistance, leaving only that “click” action that can be precisely captured and highly scheduled by distributed systems. The subject becomes fluidized in this process, and information disintegrates into fragments. When the processing and distribution of information becomes ever smoother, complex, “friction-laden” contexts (such as premises, sources, lengthy arguments) are inevitably stripped away. When content goes viral, leaving its original audience and entering different cultural circles, it inevitably suffers a degree of misunderstanding. A complex social event may be compressed into a 15-second short video, a three-sentence Twitter post, a meme with only emotion and no facts, or even a piece of brainrot aimed purely at absurdity and curiosity. The system cannot waste time allowing “explanation” to occur — Reality implodes into Hyperreality, the original has completely vanished, leaving only Simulacra entirely detached from reality. In this sense, what I’ve helped build is not merely a control system, but an automated factory producing misunderstanding and prejudice.
When we shift our gaze from the existing distributed architecture to the future that is already unfolding — that new era dominated by Large Language Models (LLMs) and generative AI — an even more profound paradigm shift is occurring.
If in the microservices era I was still a “warden” making rules, knowing the logic of every line of code, then in the AI era I am becoming a “black box” keeper. The latest AI systems no longer merely analyze data — they generate reality. In the vision of on-device AI, we attempt to have algorithms penetrate users’ most private chat histories, extracting intent and predicting needs. In utilitarianism’s calculations, this is the ultimate efficiency, but ethically, it means machines are no longer merely “surveilling” subjects but “acting as proxies” for subjects. As of now, AI’s limitation is that it possesses statistical perfection yet lacks any existential pain; it can mimic human language but cannot understand human integrity.
In class, we study HCI (Human-Computer Interaction) and Human-Centered Design. These philosophies teach us to design products with empathy, attempting to inject human warmth into cold algorithms. However, their efforts are limited. They can only stumble along in niche research ivory towers, while industrialized, commercialized HCI still serves big companies’ UX and product philosophies, still following the logic of technology.
Real change takes time. Rebuilding failed social contracts, establishing data boundaries, defining algorithmic accountability — all of this requires prolonged social reflection, intense public debate, even protracted litigation. This is a human kind of democratic time, laden with “friction.”
However, technological development does not wait for the verdict.
There exists a fatal “speed gap” here. AI’s evolutionary speed is exponential, while the construction speed of human ethical norms and legal systems is linear. This speed disconnect leads to a continuous deficit in the common good. While we are still debating the definition of “privacy,” AI has already produced a new ontological reality; while we are still discussing algorithmic bias, technology has already iterated through three generations. This “leaving behind” effect fills me with dread and anxiety: are we creating a future that humanity is destined to never catch up with or understand? In the final form of the control society, the most terrifying thing may not be that machines have developed malice, but that machines run too fast, leaving the meaning of being “human” far behind.
As an engineer standing at this crossroads, I feel a Sisyphean helplessness. I cannot immediately stop this accelerating machine. What I can do is preserve a measure of doubt in the crevices of code, a vigilance toward “efficiency.” Technology philosopher Albert Borgmann argued that the deep philosophical root of modernity’s “torment” lies in a critical split caused by the European Enlightenment: the separation of ethics (values) from physics (facts). This split caused science to lose moral guidance while morality lost its objective foundation, ultimately trapping modern humans in a vacuum of meaning, oscillating between endless consumption and nostalgia. Borgmann advocated that we must reunite these two domains and establish a unified cosmology. In this view, scientific facts and moral values are no longer mutually exclusive but together constitute a meaningful whole, jointly guiding human life.
Is this really the way out? Borgmann’s “moral cosmos” is a grand philosophical vision that provides a direction for thought and effort, aimed at freeing humanity from modernity’s predicament. However, he also acknowledged that this is primarily a framework and posture of thinking, and does not provide a detailed implementation plan. Moreover, Borgmann’s “focal practice” presupposes a possibility: that you can temporarily exit the system and establish your own life rituals in the margins. But in the age of super apps, does this space for withdrawal still exist? When WeChat has become a “digital organ,” when algorithms have permeated every crevice of life, could “focal practice” be merely a luxury of the privileged class?
Perhaps we need a different strategy. Japanese thinker Akira Asada’s concept of “escape” in his Theory of Escape offers a different perspective. Unlike Borgmann’s attempt to “root” oneself in concrete practice, Asada advocates a Deleuzian “nomadic” resistance. In the control society, power controls by encoding, classifying, and scheduling subjects — therefore, the way to resist is to refuse being scheduled. This isn’t escaping outside the system (which is already impossible in the contemporary world), but maintaining fluidity within the system, constantly sliding out of algorithmic predictions, becoming that “incomputable variable.” In the Deleuzian sense, this is a “line of flight”: through constant movement, transformation, and creative deviation, one evades power’s encoding and capture.
This thinking resonates with French philosopher Paul Virilio’s “dromology” (speed politics). Virilio argued that the core of modern power lies in speed. Whoever controls speed controls reality. AI’s exponential acceleration is precisely a form of speed violence. But speed itself is a double-edged sword: when the system runs too fast, it also becomes more fragile, harder to precisely control every variable. In this sense, “escape” means using the system’s speed to create chaos, sliding in directions it cannot capture before the algorithm has time to re-encode.
I must acknowledge a cruel reality: in the company, in production environments, it’s very difficult for me to refuse to do anything. When the product manager demands optimization of recommendation algorithms, when the system needs more precise user profiles, when business logic requires me to implement functions that I know will deepen control, it’s hard for me to say no. A cog cannot decide the direction of the machine’s rotation. Those talks about “preserving humanity in technical practice” and “designing interfaces with friction” are, in the face of real workflows, deadlines, and KPIs, often just the wishful thinking of intellectuals.
This makes me re-examine my position as an engineer. Perhaps the question was wrongly framed from the beginning. “I’ve been asking how to resist this system,” but Deleuze would say the real question is “how to generate new possibilities within this system.” The horror of the control society lies not in creating chaos, but in trying to eliminate all chaos — through algorithmic precision prediction, through frictionless user experience, through incorporating all contingency into probabilistic models. What it craves is a completely transparent, completely computable world.
But chaos in Deleuzian philosophy is not an enemy to be overcome — it is the source of all creation. True becoming always occurs at the system’s edges, in those zones the algorithm cannot yet fully encode. The key is not refusing technology or fleeing the system, but finding those spaces within technology not yet fully colonized, cultivating unpredictability in the crevices of control.
Deleuze would say that escape is not about making fine adjustments within the existing system, but about changing the trajectory itself. The real “line of flight” may not lie in how you write this code, but in whether you have the ability to choose not to write it. This makes me think of those developers who devote themselves to the open-source movement and decentralized protocols. When Vitalik Buterin created Ethereum, he wasn’t optimizing the existing financial system — he was attempting to build a fundamentally different financial infrastructure, distinct from any previous financial institution. When developers build decentralized social protocols (like ActivityPub, AT Protocol), they aren’t improving super apps — they’re challenging the concept of “app” itself. They aren’t applications but languages. Although they still have barriers (they may be slower, harder to use, harder to promote, while requiring users to understand technology and abandon their existing social environments), perhaps recognizing the existence of boundaries is itself a successful “quarantine.” They prove that another kind of technology is possible.
Asada reminds us in Theory of Escape that “escape” is not a once-and-for-all solution but a continuous process. Power learns, evolves, and re-encodes your escape paths into the system. Today’s niche platform becomes tomorrow’s super app; today’s innovation becomes tomorrow’s standard configuration. Therefore, “escape” must be an eternal movement, a posture of refusing to solidify. This also means that as an engineer, my ethical responsibility, like my engineering responsibility, is not about “getting it right” once, but continuously maintaining vigilance, continuously creating those small but persistent “glitches” within the system, as Vitalik Buterin has done. In this sense, I begin to re-understand Borgmann’s “focal practice.” It need not be a rejection of technology, but can be a practice of escape within technology, generating new possibilities outside the control system.
Indeed, not everyone can be Vitalik Buterin — disrupting at age 19 an already-disruptive complex system (blockchain), learning Chinese in a few months, while maintaining simplicity in life and consumption. In most cases, what people can do is weaker, more personal. My “escape” may happen in the moment I shut off my computer, in choosing not to devote all my time to technical optimization, in the moments I read, write, think about questions that cannot be measured by GPA, KPI, FLOPS, or P99 Latency. This sounds passive, but Deleuze and Asada would say: maintaining thought itself is a form of resistance. When the entire system demands you become an efficient, optimizable, completely transparent production unit, maintaining a part of your opacity — maintaining those unquantifiable confusions, those “inefficient” contemplations and experiences, those questions about the system’s logic — is itself a refusal to be fully encoded. Writing, especially this kind of writing that attempts to understand and expose the system’s logic, is a practice of “deterritorialization”: it lets you temporarily detach from the identity that demands you be merely a “developer,” a “user,” a piece of JSON, and return to being a person trying to understand the world.
I sit before the screen. Outside the window, there is no tower, but surveillance is everywhere. Yet precisely in this moment when there is nowhere to escape, “escape” becomes possible — maintaining nomadism within, surviving in chaos. This may not be victory or rebellion, but it can serve as a posture of survival: learning to be nomadic under the tyranny of speed, seeking crevices in the network of control, becoming, within a system one cannot flee, that subject forever in “becoming,” forever impossible to fully capture.