Open Source Power Struggles in Tech
When Amazon Web Services forked ElasticSearch in 2021 and rebranded it OpenSearch, the move was perfectly legal. The code was licensed under Apache 2.0 — one of the most permissive open source licenses in existence. Anyone could fork it, modify it, and build a commercial product on top of it. AWS did exactly that.
Elastic, the company behind Elasticsearch, cried foul. For years, AWS had sold Elasticsearch as a managed cloud service, generating enormous revenue while contributing little back to the codebase. Elastic's response was a license change — restricting cloud providers from commercializing the software without a separate agreement. AWS's counter was the fork.
Both sides had a point. Both sides were right in their own framing. And that's precisely what made the dispute so instructive — not as a story about two companies fighting over a search engine, but as a preview of the deeper fractures now running through the entire technology industry.
The Fork as Symptom
The Elasticsearch saga illustrates what happens when the foundational assumptions of open source collide with commercial reality. The original open source movement — rooted in Richard Stallman's Free Software philosophy of the 1980s — rested on a set of convictions: software is knowledge, knowledge should be shared, and proprietary software creates dangerous dependency. The practical corollary, validated by the success of Linux and Apache, was that collaborative development produces better software. Contributions and benefits would circulate within the community.
That premise held reasonably well for decades. It assumed a rough reciprocity: those who benefit from open source would, in some form, give back. But it never had teeth. The "moral contract" was always just that — moral.
The Elastic-AWS dispute made the absence of teeth visible. The community got to watch a large company extract commercial value from years of contributed labor without meaningful reciprocation, then watched the original creator respond with a license change that critics argued betrayed open source's own principles. Neither party was entirely wrong. The system simply wasn't designed to handle power asymmetry at this scale.
Today, the two projects have diverged significantly. Elasticsearch has moved ahead on ML and vector search features — ESQL, ELSER — while OpenSearch has differentiated on security defaults and an alternative query language. Both are viable. The fork wasn't a disaster. But the underlying tension it exposed has since grown into something far more consequential.
When AI Ate the Social Contract
If the AWS-Elastic conflict represented open source's stress test under commercial cloud capitalism, AI represents something more fundamental: an attack on the premise that the people who build the commons will continue to be needed.
The traditional open source model had an implicit safety valve. Even when companies profited from open source without contributing back, the developers who wrote the code retained value. Someone had to understand and maintain it. The knowledge lived in people, not just in repositories. Elastic could change its license and still sell support contracts, consulting, and managed services, because the human expertise underlying the software remained scarce.
AI eliminates that safety valve. When large language models train on open source codebases, they don't just redistribute value — they absorb and replicate the craft itself. The developer who spent years mastering a complex system now finds that system increasingly approximated by a model trained, in part, on their own public contributions. They have, in effect, provided free training data for a tool that competes with their own labor.
This is the paradox that makes the current moment structurally different from past open source conflicts: contributors are subsidizing their own replacement.
Existing legal frameworks offer little recourse. Copyleft licenses like the GPL — which require that derivative works also be open source — were built to address code reuse, not model training. When contributions dissolve into neural network weights, the notion of a "derivative work" becomes legally incoherent. Current copyright law simply wasn't written with this scenario in mind.
The industry is improvising responses. Commons Clause, which allows open code while restricting commercial service deployment, has gained adoption. Some projects are experimenting with explicit AI training exclusions in their licenses, though enforceability remains untested. There's a broader drift toward "source-available" models — code you can read but not freely commercialize. Elastic's relicensing, Meta's Llama (commercially restricted despite being called "open"), Mistral's proprietary license for its leading models: these are all expressions of the same defensive instinct.
But they're individual adaptations, not collective solutions. Which raises an obvious question: why hasn't the community organizing?
The Profession That Forgot It Was a Profession
The IT industry has, by most measures, an unusually well-compensated workforce. In many markets, software engineers earn more than doctors or lawyers. Their work is intellectually demanding, irreplaceable in the short term, and central to nearly every sector of the economy.
They also have almost no collective institutional infrastructure.
There are exceptions — the Alphabet Workers Union, Amazon Labor Union, and pockets of organizing at other tech companies. But these remain outliers. The dominant culture of the industry runs strongly against collective action, and understanding why requires tracing an ideology that was baked in from the beginning.
The tech industry mythologized meritocracy before the word became controversial. The belief that skill is reliably rewarded — that the best engineers advance, that anyone can start a company, that market outcomes reflect genuine contribution — was not merely an ambient cultural assumption. It was a founding narrative, reinforced by highly visible success stories and, for a long time, by the actual experience of workers in a field where demand consistently exceeded supply.
In this worldview, unions were for factory workers. White-collar professionalism meant not needing collective bargaining — your negotiating leverage came from your skills, not your solidarity. The "I" in the tech meritocracy was always singular.
Structural factors compounded the ideology. High turnover rates meant workers rarely accumulated enough tenure at a single company to build organizing relationships. Stock options — the promise of becoming, eventually, a shareholder rather than merely an employee — functioned as a continuous psychological bridge between worker and capital, making class solidarity feel vaguely incongruous. Remote work and distributed teams dissolved the physical proximity that has historically been a prerequisite for labor organizing. And the lavish on-site perks offered by major tech employers — the free meals, the gyms, the dry cleaning — weren't just generous benefits. They raised the cost of leaving, effectively substituting corporate welfare for collective bargaining.
Perhaps most consequential was the self-image that tech workers developed about automation. Through the 2000s and 2010s, software ate the world. Developers watched automation eliminate jobs in manufacturing, retail, and clerical work — and they were the ones building the automation. They were architects of disruption, not its victims. The instinct for collective self-defense that develops when workers perceive existential threat simply never had reason to activate.
That self-image has proven slow to update. GitHub Copilot was a "productivity tool." ChatGPT was "fine for junior tasks." Current AI coding assistants that can handle substantial engineering work are, in the current rationalization, only a threat to those who aren't adapting. The goalposts keep moving, each movement more compressed than the last.
The industry is now encountering an AI displacement shock without the institutional infrastructure that other professions built precisely to absorb shocks. Doctors, lawyers, and financial workers have licensing boards, professional associations, and lobbying infrastructure developed over generations of navigating adversarial environments. Tech workers have mostly their own individual exit options: switching employers, going freelance, starting companies. Those options will serve individuals well enough, for a while. They are not a systemic response.
The Missing Voice
The absence of organized labor is part of a broader absence: the IT industry, for all its economic and cultural dominance, has essentially no political representation on behalf of working engineers.
This is often obscured by the media visibility of tech billionaires. When AI regulation, data privacy law, or platform liability comes up for debate, the voices that carry are those of Zuckerberg, Bezos, Musk — executives whose primary identity is capital, not craft. They speak for the industry in the sense that they command its largest platforms and deepest lobbying budgets. But their interests as major shareholders are not the interests of the engineers, moderators, and technical workers who actually build and maintain these systems. Conflating the two is a category error that policy debates routinely commit.
The deeper problem is structural. Political influence, in mature industries, is accumulated through time and institutional memory. Medicine has centuries of guild tradition underpinning its licensing boards, professional associations, and legislative relationships, maintained by retired physicians who become regulators and politicians. Law built the judicial system itself. Finance has grown alongside state power for so long that the two are effectively co-constitutive.
IT as a mainstream commercial industry dates to the 1980s and 1990s, with genuine cultural ubiquity arriving only in the 2000s. That's not enough time to build the generational institutional depth that translates sector interests into durable political influence. The partners at a law firm who become senators were associates when the lobbying infrastructure that represents their profession was already mature. Tech doesn't have a comparable pipeline yet.
Worse, the timing of the industry's rapid growth coincided with a period of minimal regulation. Other sectors developed political muscle by fighting regulatory battles — learning to engage with legislators, staffing revolving doors, building coalitions. Tech spent its formative decades operating largely outside that adversarial dynamic, behind a cultural consensus that regulation stifles innovation. Political capability grows through resistance. Tech is now encountering serious regulatory pressure without having developed the capacity to navigate it constructively.
The fragmentation of technical specializations makes common cause elusive. Frontend developers, infrastructure engineers, AI researchers, security specialists, and embedded systems programmers don't naturally organize around shared interests the way that, say, unionized autoworkers do. There's no single professional body that speaks for all of them, and the velocity of technological change makes it hard to develop shared norms before the relevant context shifts.
The result is that on the questions that most directly affect working engineers — AI labor displacement, algorithmic management, data rights, non-compete enforcement — the loudest voices in policy debates represent either large corporations or abstract consumer advocacy, with the workers themselves largely absent.
A Structural Account
These three stories — open source's collapsing social contract, the absence of labor organizing, and the void in political representation — are not separate phenomena. They share a common root.
The IT industry grew extraordinarily fast in an environment that required almost no collective defense. The open source model worked because power asymmetry was manageable. The labor market rewarded individual mobility over collective bargaining. The regulatory environment demanded little political engagement. In each domain, the industry developed individual strategies and informal norms instead of institutional structures, because institutional structures seemed unnecessary.
AI is simultaneously invalidating the assumptions that made each of those individual strategies adequate. Open source licensing wasn't designed for a technology that trains on code without executing it. Individual job mobility doesn't protect workers from sector-wide displacement driven by a new technology paradigm. Informal political influence through CEO relationships doesn't translate into worker-protective policy.
The industry is, to borrow a phrase, finding out.
What comes next is genuinely open. There are signs of movement: renewed labor organizing at some tech companies, growing interest in alternative licensing models, a nascent conversation about whether professional certification in software engineering might be warranted. None of these initiatives are close to mature.
The more pressing question is whether the industry's culture — individualist, technically focused, skeptical of institutional solutions — can adapt quickly enough to generate the collective infrastructure that the current moment demands. The challenges ahead: adequate open source governance for the AI era, worker protections against automated displacement, and genuine political representation for engineers rather than executives. All of them require exactly the kinds of collective action that the IT sector spent its formative decades avoiding.
The industry built extraordinarily powerful tools. Whether it can build the institutions needed to govern them — and protect the people who build them — is a different kind of engineering problem, and not one that individual brilliance alone can solve.
