Tech companies think they benefit from employees who can't quit

I've been moving through tech orgs long enough to recognize a pattern that many of you will recognize as well. Some teams are passive, they ship on time, hit targets, run clean processes, and yet nobody ever kills a bad idea in the meeting. Nobody says this is the wrong thing to build. The roadmap has a thing, or six, on it that three people privately talk about how it won't work, but it moves through planning without a word, with smiles even. You watch the lead engineers nod at something they've already agreed among themselves is wrong, and you know what that nod means. It means they've done the math and speaking up isn't worth it.

After you've seen this in a few different orgs, you start to wonder what they have in common. It often is that a large share of the engineers are on H-1B visas, many of them years into the green-card process and still far from the end of it. Their ability to stay in the country depends on staying employed with a sponsoring employer, which means getting fired isn't just a job loss, but rather a life changing event that could get you deported to another continent.

To understand why that matters, even to the org itself, it helps to understand what the constraints look like in practice. An H-1B worker who loses their job has sixty days to find a new sponsoring employer, change status, or leave.¹ The harder problem is what happens to workers who are deep in the employment-based green-card queue. If you switch employers mid-process, you risk losing your priority date, the position you've been holding in line. For Indian nationals, that line is long in a way that's hard to convey without looking at the numbers. Current government data shows EB-2 and EB-3 India final action dates running about fourteen years behind today's filing date.² The projected wait for someone entering that queue today runs to several decades, policy analyses have modeled the backlog clearing in timescales that often exceed a full working career.³ Workers who have built up enough time in the process can transfer to a comparable role at a new employer without losing their place, but only once a specific adjustment-of-status application has been pending for 180 days.⁴ Until that threshold, the queue position is tied to the employer. L-1B workers are even more constrained: the visa is specific to the sponsoring company, there is no portability provision, and if the employment ends, the visa ends with it.⁵ When the cost of getting fired includes blowing up years of immigration progress, your relationship to workplace risk changes in ways that aren't about courage or culture.

The first thing that changes is how people handle disagreement. Most employees already think twice before pushing back on a manager publicly. Add visa dependency and that hesitation gets much harder to overcome, not because the worker is passive or doesn't have a view, but because the downside of being wrong about when to speak up is much larger than it would be for someone with an easier path out. The engineers who keep their heads down tend to get better reviews, better projects, and more internal sponsorship. The ones who cause friction, even accurate and useful friction, get noted in ways that follow them. Organizations don't have to plan this outcome. It just happens through normal performance processes running over enough time.

What it looks like on the inside is a lot of side conversations that never make it into the room. A private discussion on why the architecture decision is wrong, never raised at the design review. The post-mortem that identifies the incident as preventable in private but produces a clean official write-up. The engineer who sends a note to a small Slack channel instead of the all-hands. Individually none of these look like much. Accumulated across an org and a few years, they add up to a culture where managers stop hearing bad news until it's very late.

The second thing that changes is how people approach innovation, and this is where I think the long-term cost is most underestimated by the companies themselves, because it's backfiring on their own bottom line.

Real innovation requires someone willing to be visibly wrong, and requires a degree of personal safety to do it well. You have to propose the thing that might not pan out, defend an unpopular position through the skepticism of a design review, spend weeks on something and then say it didn't work out. For someone whose immigration status depends on holding their job, the asymmetry of that bet is far more severe than for someone who, at most, loses their job. If the bet fails badly enough to cost them their position, the consequences run well past a performance note. In the worst case it's a managed exit, which starts a sixty-day clock, puts years of green-card progress at risk, and means uprooting a family and returning to a country they may not have lived in for a decade. If the bet succeeds, they get a good review cycle and maybe a raise. That's not a risk-reward ratio that makes people want to take swings. It produces the opposite: safe, predictable execution on work that's already been decided, from a team that ships reliably and invents almost nothing. The execution metrics look good quarter after quarter while the ideas that were never proposed and the bad product calls that nobody stopped accumulate quietly. The damage doesn't show up on a dashboard. It shows up later, when the org needs to do something new and discovers it doesn't have the muscle for it.

There's a related pattern that's easy to miss because it looks like productivity. When keeping the job is the primary goal rather than the secondary one, the work tends to drift toward whatever is most visible and easiest to defend in a performance review. Towards getting things done, but never completed. That often means overengineering: building something more complex than it needs to be, adding layers that demonstrate effort and technical depth, choosing the approach that produces the most impressive artifacts rather than the simplest one that would actually work. It also means promotion-driven development, where the roadmap gets shaped less by what the product needs and more by what generates the kind of output that advances careers, or at least protects them. Neither of these patterns is unique to visa-dependent workers. They show up anywhere that job security becomes the primary thing people are optimizing for. But the incentive is significantly stronger when getting it wrong doesn't just cost you a bad review cycle, but your legal status and everything you've built toward over years.

The third dynamic is that the system tends to reproduce itself through the people it promotes.

Managers who built their careers inside this system learn, through experience, that certain kinds of workers are easier to manage. They don't push back on scope creep. They don't threaten to leave when conditions get bad. They absorb overwork without making it a negotiation. When those managers start hiring and promoting, they tend to draw from the same networks and favor the same profile, often without thinking of it in those terms. Referral hiring concentrates along existing demographic lines. Low-friction engineers get promoted ahead of high-friction ones. Over time the org develops a kind of institutional preference for workers who are less likely to cause problems, and the structure of immigration dependency is one of the main things that creates that preference in the first place.

The immigration architecture that produced this outcome wasn't designed to produce it, and understanding where it went wrong helps explain why it's so pervasive. The per-country cap on employment-based green cards was a deliberate choice to prevent any single nationality from monopolizing permanent residency slots. The idea was reasonable: a healthy immigration system should draw from a broad range of countries, and no one national cohort should dominate the permanent workforce in a critical sector. The problem is that this logic was applied at the end of the pipeline, getting the green , and never extended to the front of it: getting the H-1B or L-1B visa.

H-1B visas have no equivalent per-country diversity cap. Neither do L-1B visas.⁶ Employers can sponsor as many workers from a single country as they want, and the market has done exactly that at scale. Indian nationals now account for roughly 70 percent of H-1B petitions in a typical year.⁷ The green-card system correctly tries to distribute permanent residency broadly, but by the time workers reach that stage, the intake has already been concentrated so heavily from one country that the cap produces an enormous backlog for Indian nationals specifically, while workers from most other countries move through quickly. The per-country diversity rule was applied at the wrong stage. You can't fix a pipeline concentration problem at the output when you've allowed unrestricted concentration at the input.

The case for fixing it doesn't require being against immigration. Tech, and our economy, has genuinely benefited from global talent and will keep doing so. The argument is that when one country provides most of the workforce in one critical industry, and those workers are structurally constrained by a backlog that compounds over decades, you've created a labor-control dynamic that nobody planned but that companies benefit from regardless of intent. The straightforward policy response is to extend the diversity logic that already applies to green cards to H-1B and L-1B approvals as well, so that no single country accounts for a large majority of work visa approvals in a given year. Canada's Express Entry system, which draws from a broad national distribution without creating per-country intake bottlenecks, is one example of what a better-designed front end looks like. That doesn't close the door on any individual engineer. It distributes the intake more evenly, the backlog concentration eases over time, and fewer workers end up in the decades-long queue that produces the dependency dynamic in the first place.

The team I started with is still shipping. The dashboard still looks good. The managers are pleased with the velocity numbers and the clean on-call rotation.

What the numbers don't show is the design review where the bad call went through without a real challenge, or the roadmap item that several engineers think is wrong but nobody has said so in a room where it would matter. There's probably an engineer on that team who had a better approach to something, thought through whether raising it was worth the risk, and decided it wasn't. That's not a character flaw. Given the constraints, it was the right call. The org just paid for it in a way that won't show up until later, when the thing they built doesn't work the way it needed to, and the people who saw it coming have no particular reason to say so now either.

It's worth being clear about something before leaving this topic. If you were in the same position, you'd probably do the same things. You'd keep your head down during the bad meeting. You'd write the private doc instead of raising the issue. You'd build the thing that gets you promoted rather than the thing that's right. Not because you're weak or cynical, but because you're rational and you have a family and years of your life invested in staying. The engineers navigating this system aren't doing anything that most people wouldn't do given the same constraints. That's precisely the point. The behavior that looks like passivity or careerism from the outside is just a reasonable response to a situation that was never designed with their interests in mind.

The problem is the system that creates those incentives, not the people who respond to them sensibly. The H-1B backlog, the per-country caps, the decades-long wait for workers from high-demand countries, the employer dependency that follows from all of it, those are policy choices, and they can be changed. The engineers caught inside the current system didn't ask to be put there, and most of them would tell you exactly what's wrong with it if they thought they could afford to. Fixing it doesn't require resenting the people it affects. It just requires being clear-eyed about what it produces and willing to say that a better design is possible.