We’re All Agents of Change: Towards a More Just Technological Future

It was pitch dark when the alarm pierced through my room, yanking me out of the sweet nothingness of deep sleep. I pressed ‘ACK’¹, noticed the time (3:29am), filled a glass of water, moved to my desk, read the issue brief on my phone while waiting for my laptop to boot… all in rapid succession, an on-call ritual I’d cultivated over the years. By the time my monitor and screen were fired up with code, reports, and dashboards, barely a minute had passed.

With preparation and experience, anyone can be put on standby and respond to an ill-timed emergency at a moment’s notice. Actually extinguishing a production fire, however, proves to be much more demanding. While on-call, I was responsible for all the components that my team built—12 engineers’ brain children—and despite understanding the broad architecture of the system, sifting through unfamiliar code snippets and pipeline data was always necessary. Once in the zone, I acted like an archaeologist, excavating pages of error messages to locate my treasure: the culprit behind the crash. Needle in a haystack, race against the clock, my task was to stop and restore the rapidly cascading failures. There I was again, before the break of dawn, trying to patch up the infrastructure that powered 135PB² of data throughput on an average day.

For three years, I worked as a software engineer who built and maintained data processing infrastructure for a tech giant. Once every season, I would be put on a week-long 24/7 on-call rotation, responsible for keeping systems up “despite hurricanes, bandwidth outages, and configuration errors”³. Being on-call in tech (also known as Site Reliability Engineering, or SRE in short) is disruptive, intellectually taxing, and often relegated to the bottom of the desirability hierarchy in terms of job specialties. It’s dirty, unwelcomed work—even when everything’s smooth sailing, the prospect of having to work on any exigency within 5 minutes can at best bring feelings of inconvenience, and at worst become immediately unnerving. To make matters more complicated, it’s hard to predict how much investigation and effort are required before the system can be brought back to functional equilibrium. Some of the trickiest corner cases that my team encountered, which also kindled the most persistent outages, necessitated all of us working in tandem for hours on end.

To the outsider, however, our labor is eclipsed by the fact that the infrastructure is ostensibly, perpetually functional. Tech companies have put extremely high fault tolerance—the ability for systems to operate fully despite errors—at the forefront of SRE requirements. Thanks to an abundance of data centers designated for duplicate computation, built-in failover mechanisms, and conservative paging thresholds, it’s indeed rare for data pipelines to be actually unavailable. While software engineers who build glamorous new features are often given the lenience of ‘fail fast’⁴, site reliability engineers are held to an entirely different standard and at the minimum, expected to maintain five 9s system availability—99.999% of system uptime: less than six minutes of  failures per year. Anything that falls short of this onerous goal would lead to ‘postmortems’—solemn and accountability-centered meetings where perfectionism prevails. It is no wonder that Andrew Russell, co-founder of The Maintainers—a global research and practice network interested in the concepts of maintenance, infrastructure, repair, and the myriad forms of labor and expertise that sustain our human-built world—remarked that “the most unappreciated and undervalued forms of technological labor are also the most ordinary: those who repair and maintain technologies that already exist”⁵.

I am tempted to build a case for all maintenance labor with my experience in SRE in mind: how, across contexts and professions, “there’s [always] an unsung correlation between the perpetuity of upkeep and society’s reliance on it”⁶, and advocate for more consideration and recognition from those of us who are not directly involved in maintenance, and more broadly, care work. However, to venture in that direction would be uncritical and dishonest. Unlike homemakers whose labor has been financially devalued and informalized, sometimes wholly taken for granted, on-call engineers are properly compensated for holding the pager and resolving incidents. Whether paid per emergency or as a percentage of base salary, on-call compensation is on par with the standards set by the industry (think 6-figure salaries for entry-level jobs). Some would even argue that when on-call compensation is calculated per number of days served, it can feel like passive income, generously dispensed when stretches of time go by without any incident.

Additionally, and in my opinion more problematically, unlike firefighters or emergency medical technicians who rush to save lives, maintenance engineers’ praxis is inescapably wrapped in the logic of profit, exploitation and capitalism: the very antithesis of care and service. The data processing infrastructure my team created for internal use was primarily used to collect, categorize, and analyze the company’s ads data: views, clicks, purchases…. We built a highly intelligent and efficient network that ingests logs⁷ completely illegible to human beings, and converts them to persuasive invoices for advertisers: pay us X amount of money, because we generated Y interactions and Z profit for you. Other divisions of the company had access to our infrastructure too, albeit with a laissez-faire attitude and in a throttled capacity: a direct implication of how ads alone have unfailingly brought in more than 75% of company revenue over time.

A vicious cycle ensues: advertisers invest money in companies like my ex-employer, which then work to serve the advertisers’ interests, enabling them to continuously and increasingly farm and financialize any Internet wanderer’s attention. Advertisers, empowered by entire ecosystems designed to boost consumerism, generate more income to increasingly fund algorithmic operations. The phenomenon is spreading like wildfire: tech companies are transforming the advertising landscape so much that every plot of virtual real estate has been converted to billboards. Running out of space for ads? The companies will create space out of thin air and squeeze them in: in between stories, embedded in every minute of a video, overflowing in pop-up windows. The result is shocking. For Instagram, for example, the single-minded pursuit for growth through advertising has produced $39.7 billion of revenue in 2023, up from $6.18 billion in 2018, just five years ago⁸. Behind these impressive numbers: millions of users who harbor complaints about the plummeting quality of the app, all while still tormented by its addictive nature.

The crux of the issue goes way deeper than what an analysis of technical infrastructure or the attention economy can reveal. The broader question stands: where is the tech industry heading toward? And for us tech workers, whose labor is intertwined with a future that technology posits, what efforts are we contributing to? When I started learning Computer Science in 2014, I dreamed of working on projects that would tangibly and drastically improve people’s quality of life. It was a time when the industry was undergoing rapid transformations, and possibilities seemed abundant. Reversing the trend of climate change, alleviating poverty in rural regions, connecting loved ones who are oceans apart… these all struck me as real-world needs and crises that can be addressed programmatically. A decade down the road, technology has arguably only propelled us into a world of increasing frivolity, existential peril and magnifying inequality. The domains that receive the most capital, thought, and resources in tech remain the most morally and politically ambiguous, if not outright destructive: advertising strategies that excessively monetize our engagement, artificial intelligence breakthroughs that displace jobs and introduce bias, surveillance and monitoring tools for controlling citizens…. Most recently, ‘Lavender’, an AI machine trained to identify suspects for assassination, was repeatedly deployed to advance bombing sprees in Gaza. From the first AI-facilitated genocide to rampant ads on social media, the trends in tech emerge and grow to respond to capitalism, profit maximization, and reinforcing established power structures.

Still, many people clamor for a seat at the table. Last year, 42% of seniors at MIT graduated with a Computer Science degree, up from 23% just a decade ago, an upward tick reflective of the increases elsewhere in North America⁹. Beyond college, many workers dissatisfied with their careers have opted for vocational software engineering training through boot camps or online programs. Against the backdrop of capitalism and its repercussions—hyper-inflation, extended work hours, unprecedented atomization of individuals—a job in tech seems to most reliably deliver promises of respectability and financial security, without which we can barely survive, let alone thrive.

With the influx of skilled workers who are eager to contribute to the tech sector, the supply of job positions unfortunately hasn’t been stable in the least. After a pandemic-fueled hiring spree in 2021, many tech companies experienced dramatic reversals in hiring strategies, swinging rapidly to extensive layoffs. Between November 2022 and January 2023—in the span of three short months—at least 200,000 tech workers in the United States were laid off, often without sufficient prior notice and sometimes without even a severance package. The trend is nowhere near dying down: 3 months into 2024, some 222 tech companies have laid off more than 56,000 workers. A closer look at the layoffs shows that the decision to keep or fire a worker is often determined regardless of one’s performance or tenure. As lawyers debate the morality and legal implications of the unprecedented amount of terminations, one thing is clear: many tech companies—even the ones that have always boasted prioritizing wellbeing and work-life balance—eventually view employees as disposable, conveniently sacking them when they need to cut costs or re-organize internal structures.

The sense of looming precarity is intensifying: even if one miraculously manages to keep a job through endless rounds of layoffs in tech, it’s hard to evade the feeling that our humanity is abstracted, essentialized, and reduced to a single dimension: the amount of engineering productivity we can provide. That labor is in turn viewed through a pragmatic, purely economic lens: desirable if capable of generating ample revenue, expendable otherwise. For the hundreds of thousands of immigrants on work visas, the sense of dread is compounded by the fact that once terminated, we only have 60 days¹⁰ to scramble and find our next employer, lest we risk deportation and losing entire lives we’ve built for ourselves in this beautiful country.

While treated unfairly and oppressed, a common coping mechanism people develop is empathizing with their oppressors, constructing justifications and mental models “rational” enough to alleviate their cognitive dissonance. I’ve encountered tech workers who dismiss the potential harm of the software they develop because “if I’m not building it, someone else will anyway”. I’ve also seen workers from historically marginalized communities swallow the discrimination and injustice they face at work, because it’s “already better than what my parents went through”, and resort to keeping quiet lest they face retaliation.

We all deserve better: both as tech workers, and as individuals navigating a society where technology reigns omnipresent and indispensable. My plea is not for you to turn away or renounce the field altogether. Instead, my plea—if you ever find yourself mired in the vicissitudes of the tech industry—is to preserve your humanity. Don’t think of yourself just as a cog in the machine; don’t toil away just to reinforce systemic oppressions and exacerbate existing inequalities. Digital infrastructure predominates the developed world, and we’re equipping ourselves with incredibly valuable knowledge that could tangibly improve our collective material conditions: how, then, can we put it to good use? For whatever we are working on, ask ourselves: is this a meaningful avenue for my labor? If not, how can I imagine a more expansive future aligned with my values? What are some of the efforts I can undertake to improve working conditions and software accessibility? (joining a workers’ union, mentoring less experienced folks, and contributing to a long history of tech labor movements immediately come to mind)

The tech industry is gradually losing its prestige as the golden standard of white-collar labor. While we contend with the reality of unethical practices, casual worker mistreatment, and the devaluation of maintenance work under the behemoth that is the tech industry, the silver lining remains: together, we have the agency and power to contribute in meaningful ways, and through perseverance and interdependence, build a technological future that is just, equitable, and transcends capital-and-power accumulation.

Footnotes

1. An acknowledgement (ACK) is a signal passed between communicating processes, computers, or devices to signify acknowledgement as part of a communications protocol. After the term was coined in the context of data networks and telecommunications, it’s been widely adapted in software engineering to also indicate human acknowledgement of network and computer updates. ↩︎
2. For the ease of illustration, over 3.4 years of 24/7 Full HD video recording would be around 1PB in size ↩︎
3. https://sre.google/ ↩︎
4. A core principle behind the Agile software development methodology ↩︎
5. https://aeon.co/essays/innovation-is-overvalued-maintenance-often-matters-more ↩︎
6. https://www.are.na/editorial/miller-ukeles ↩︎
7. One of the simplest storage abstraction in software engineering: an append-only, totally-ordered sequence of records ordered by time.
   ↩︎
8. https://www.statista.com/statistics/1104447/instagram-ad-revenues-usa/ ↩︎
9. https://www.theatlantic.com/technology/archive/2024/03/computing-college-cs-majors/677792/ ↩︎
10. Workers on nonimmigrant work visas have 60 days of grace period after termination for their immigration status to remain valid. ↩︎