For folks who are adjacent to the US software industry, the past few years have been marked by waves of layoffs.
Some people with very keen eyes have been saying, in my opinion quite accurately, that layoffs like these are a failure of leadership. They are not strictly necessary: companies like Meta and Google continue to be extremely profitable. They often seem to be a masked correction of the pandemic-era overhiring that almost no one is willing to admit. The pandemic was an extraordinary event that generated atypical and transient usage patterns, and companies hired as if those patterns would last. And the ways these mass layoffs were carried out—often with a large chunk of the workforce guessing for weeks and then being cut off by the faceless corporate exec team without so much as a courtesy call—were devoid of empathy and dignity.
The cruelty arguably started with the Twitter takeover and subsequent mass layoff, which has been considered a travesty for anybody who isn’t a fan of Musk. But what’s sad about it is that many other (though not all) companies, instead of taking it as an example of what not to do, decided to treat it as a permission slip and a template to borrow from, especially in the last year or two.
And we are far from done. The entire software industry is now grappling with the new world where source code, the currency of the past 50 years, has been commoditized. Does that mean software lost its intrinsic value? Nobody seems to know for sure. “Talk is cheap. Show me the code.” sounds like the slogan of a bygone era. So if code is now also cheap, what does that mean for the skilled labor force whose ostensible job responsibility is to create code? A good chunk of industry leadership seem to believe the people are now less valuable too. As a case in point, the official reason for layoffs increasingly cites AI adoption, or preparation for it.
This reason, in my opinion, is deeply flawed if not wrong. It’s a poor attempt to mask the same leadership failure that has been the root of most layoffs, and in the long term, it is not helping the companies that are wielding it. Here is why.
“AI will do your job.” is a popular rationale behind AI-induced layoffs. It assumes a well-defined “job” to be done. But that is not really the case. As an industry, we have never quite articulated the nature of software creation. Software is open-ended in many ways. Yes, we all have the notion of “search” as a software feature. But exactly what that feature provides, or how it should be carried out, are very much open for interpretation. This is why Google used to be excellent at Web search but still sucks at it in Google Drive. The process of implementing software is equally fuzzy. Code gets written, and thrown away. Code is borrowed, and then modified, and sometimes merged or revived. No two engineers ever give the same estimate of how long something should take, nor can they back their numbers with certainty. The job is defined largely through the process of being done. And it can be redefined differently if it were to be performed again, including by someone who, doing it today, would likely reach for AI. Giving this reason effectively implies the speaker is holding onto a caricatured notion of the job.
“We will need to hire talent that is AI-proficient instead.” This reminds me of a joke when Node.js first became popular, that soon job postings would ask for 8 years of Node.js experience when the project was only 6 years old. This turns out to be a durable meme, as every tool that became trendy essentially evoked the same joke. How does anybody become proficient? By using the tool! Preferably for a real goal with meaning! Isn’t that what a job is supposed to be? But instead of giving people the space and support to help them acquire the new skill, leadership is all too happy to outsource that training to an unspecified “someone else”. Instead of cultivating an environment that encourages learning, this kind of thinking imposes a culture of fear, and fear stifles creativity.
But the creation of software is neither mechanical nor replicable. It is the tip of the iceberg of the social structure and collaborative process manifested by a community. In other words, software embodies the brain trust and the collective care toward the problem space. Engineers routinely learn what they need to fulfill the tasks. Senior engineers tend to downplay the importance of the code itself, but instead focus more on process and organization, because they increasingly see this hidden reality behind the veil of bits and bytes. Unfortunately, this is not a universally accepted truism. The industry, even in the age of code-on-tap, has yet to move past coding as the primary job function. (Otherwise we wouldn’t have so many companies, frontier AI labs included, still insisting on the whiteboard-style coding interviews.) By hollowing out the social structure and instilling fear in the people who work on problems, these companies harm the organic driving forces that produced the artifacts. But the artifacts were never the most valuable part to begin with, and they are further de-valued by the very thesis that AI is trying to make. When the tree of care and support rots, the fruit will turn bitter too.
Ironically, the places where AI has been the most constructive are places that didn’t have a great deal of competence or confidence in software. Places where software continues to be seen as a means to an end are unbothered by the exact flavor of the tools that comes to their aid, because their eyes are on the real prize—the problem they’re trying to solve, which remains unchanged. The software industry, of which I am a part, is somewhat blinded by the solutions we are able to create, instead of the problems we try to solve. But the real value is always in solving the problems. And empowering people is the exact thing that allows us to solve bigger problems with more efficiency. By walking away from that optimism, companies are undercutting their long-term potential.
I’m not sure what I can personally do in this environment that has become somewhat cold and cruel. But I have enough experience to say the doom-and-gloom is neither good nor right. The faith in our ability to rise to the challenge is not merely a byproduct of the boom era of Silicon Valley and tech industry in general, it is in fact a necessary condition to chart a path through unknown territories. Those who consider themselves leaders, the spiritual descendants of inventors who changed how things were done by cobbling together prototypes in rented garages, could probably unleash the promised productivity boost by lending a bit more grace to the very people who will be doing the actual work.