Why Software Engineering Is Becoming a Poor Choice for Most Americans in 2026
I wrote this piece for American students considering a career in software engineering, especially those who might benefit from a more honest picture before committing significant time and money to this path.
I’ll say upfront: much of what follows is subjective, shaped by personal experience and observation. I’m not asking you to take my word for anything. Think critically, do your own research, and treat this as one data point among many. That said, I’ll ground the argument in real evidence where I can, so you have something concrete to weigh. Consider every point raised here a starting point for your own deeper investigation, not a final verdict.
Finally, and this matters, everyone’s situation is different. Your financial circumstances, location, interests and aptitude, and risk tolerance all affect whether this advice applies to you.
This blog is fairly long, so I have broken it down into the following sections:
College Admissions Competition — or Shall I Call It the Hunger Games?
Internship Competition & the First Job Search — Welcome to the Hunger Games, Round Two
Ghost Jobs, Resume Harvesting, and the Games Companies Play
The Global Competition That You Didn’t Know About
Offshoring and the Infinite Labor Supply
Quant/FAANG+/Unicorns or Bust
Cost Cutting: The CFO and CEO’s Endless Efficiency Theater
Layoffs and Instability, PIPs: The Ugly Stuff
The Million Strong Shadow Work Force
The Specialist vs Generalist Duality: Both Come With a Price
The Treadmill That Never Stops: Upskilling, Bar Raisers, and Ageism
No Ladder, Just a Glass Ceiling: Stagnation and an Ambiguous Promo Path
Profits Up, Paychecks Flat: Five Years of Going Nowhere
The Cost of Living Trap: High Salaries, Higher Rent
The Unknown Looming Dangers and Possible Benefits of AI
Stuck: The Surprising Lack of Exit Optionality for Software Engineers
Final Word
1) College Admissions Competition — or shall I call it the Hunger Games?
For many people, the journey toward becoming a software engineer begins well before college, often as early as middle or high school. It might start with an AP Computer Science class, a course at a local college, a personal project born out of curiosity, a family member’s influence, watching a classmate build something cool, or a YouTube rabbit hole that made you want to try it yourself. Others simply got swept up in movements like “learn to code” or chose to major in computer science in order to gain practical, hard skills.
For those of us who love building things, are naturally curious, and are drawn to the promise of a strong income, the career naturally seems appealing. And honestly, the job has, or at least had, a lot going for it:
Intellectually stimulating work
Limited exposure to difficult people
No physical hazard
Strong compensation
Potential for early retirement
Remote work flexibility
Good work life balance
Respectable social status
No requirement for significantly further education before entering the workforce
And last but not least, the opportunity to change the world
This led to a massive influx of students pursuing computer science and related degrees over the last decade. It is now consistently one of the most competitive majors at universities nationwide:
UC Berkeley: EECS acceptance rate of 7.6% and CS at 1.9% in 2024, compared to the school’s overall acceptance rate of 11.4%
UCLA: Computer science acceptance rate of 7.6% in 2025, versus an overall school acceptance rate of 11%, with an interquartile GPA range of 4.64 -- 4.92 (the highest of any major at the school).
USC: CS acceptance rate of 7.5%, compared to the school’s overall rate of 10%.
While comprehensive data across all top private and public universities is difficult to obtain, the trend is consistent enough to draw a clear conclusion: just to gain admission to a computer science program at a respected school, you already need to be well above average, academically stronger, in many cases, than the typical admitted student at some of the most selective universities in the country. And even then, you are competing against a self selecting pool of some of the most driven and well prepared students in the nation.
Now let’s say you are one of them. You work hard, maintain a strong GPA, take AP classes, score well on the SAT or ACT, stack your extracurriculars from your school’s STEM club to national science fairs and international Olympiads and everything in between, and somehow manage to clear every hurdle. Congratulations, you have earned a spot in a competitive computer science program.
But this is just the beginning.
2) Internship Competition & the First Job Search — Welcome to the Hunger Games, Round Two
So you made it in. You are enrolled in a reputable computer science program, happy that a bright future seems to lie ahead and confident that you chose a major that actually leads to a job. You hear from peers, online communities, clubs, and upperclassmen that the key to landing a full time software engineering role is internships. But to get an internship, you need prior experience, or at the very least, you need to demonstrate that you are worth investing in.
So you get to work. You join computer science clubs and organizations, participate in hackathons, do research, build side projects, become a teaching assistant, open a LeetCode account, and begin grinding problems in your spare time. You survive the brutal CS courses: operating systems, algorithms, computer architecture. Maybe you even sacrifice your grades for these side quests in order to secure a job.
And then you start applying. Regardless of how prepared you are, your response rate is under 10 percent and, for most students, it is probably 3 to 5 percent or less. But you do not give up. You keep grinding, keep applying, keep tweaking your résumé. You keep interviewing, and eventually you land something. Maybe it is a big tech company, maybe a large enterprise, maybe a scrappy startup. But something is better than nothing, so you take it. You know you need to stack your résumé.
If you are lucky, they extend a return offer. But that offer is often subject to more randomness than anyone will admit, ranging from the mood of the CEO or VP to whether the company is struggling and a hiring freeze is quietly underway.
If you are among the lucky ones and receive that return offer, you can finally breathe. You can pause the grind for a while or use the offer as leverage to pursue better options. But if you are not lucky, you have to repeat the entire process. Only this time, the market is worse, and somehow having more experience still is not enough.
If it is your first job search in this industry, you are quickly surprised by the sheer variety of interview experiences you encounter. Startups might reject you over “culture fit,” even though the CEO sometimes has barely more experience than you. Mid size and smaller companies ask a random mix of questions that can feel impossible to prepare for. You complete a stream of online assessments with no follow up, despite passing them.
Then there is big tech. Some companies have reasonable interview processes, while others do not. Apple is notorious for grueling multi round processes that can stretch on for weeks, and that is only if you manage to secure an interview in the first place. Every company has its own distinct hiring process. Since most large tech companies lean heavily on LeetCode style algorithmic problems, you might get lucky and receive something similar to what you have practiced. If you do not, well, better luck next time.
The only advice anyone offers is to grind more, apply more, and keep those LeetCode numbers climbing.
What really catches you off guard, though, is how low your response rate is, even when you meet the job requirements. At the end of the day, it is an entry level role. It seems unreasonable to expect you to already know all of X, Y, and Z technologies. You start to wonder whether companies are even serious about hiring. This is your first real encounter with how the corporate world operates.
Now, with your graduation date looming, if you have gone above and beyond, or if luck happens to be on your side, you might land something. But if you graduate without an offer, you may soon find yourself forced into the gig economy, IT help desk work, or something else entirely just to keep the lights on, because bills do not pay themselves. The cruel irony is that the longer you remain outside the tech workforce, the less attractive you become to employers.
That said, most people do eventually make it or at least that is what government data shows. Whether you believe it or not I leave that up to you. 😉
3) Ghost Jobs, Resume Harvesting, and the Games Companies Play
As pointed out above, you might have been wondering why you don’t get a response for many of the jobs you apply to despite being qualified, or why you quickly receive a rejection after applying in less than a day or two.
Welcome to the ghost job epidemic.
A ghost job is a posting that a company has no real intention of filling anytime soon. Companies post them to create the impression of growth, build a passive pipeline of candidates for future roles, or simply collect résumés, gather data on the job market and pay ranges. Approximately 30% of the jobs posted online are fake postings, but there is not much you can do about this, so you eventually just have to accept it and move on. There is also the added layer of automated résumé screeners, commonly called ATS. To be honest, even today most people do not really know how they work. The best you can do is try to match the job description as closely as possible, but most applications will still get rejected before a human ever sees them. A referral might help you get past this stage, but it is highly case dependent.
4) The Global competition that you didn’t know about
If you are a 16 or 17 year old on the path to become a software engineer you probably have never thought of the global competition you face, you might have never even stepped outside of the U.S., but the reality is half of the competition you face is not even inside of the U.S.
H-1B: Perhaps the most controversial visa in America.
As of 2026, there are approximately 700,000 to 730,000 active H-1B visa holders residing in the United States. Including their dependents — spouses and children — the total number of people in the U.S. under this program is estimated at nearly 1.3 million. The majority of H-1B holders work in the tech industry. According to the U.S. Bureau of Labor Statistics, the United States employs roughly 5 million workers in computer and mathematical occupations, the category that covers most technical jobs in the tech sector. With approximately 700,000 active H-1B workers in the United States, visa holders represent a meaningful and visible share of that technical.
H-1B workers broadly fall into two groups. The first consists of genuinely exceptional talent whose skills are scarce enough that companies actively recruit them from around the world. These are the kinds of engineers and researchers who can meaningfully move the needle in highly specialized areas.
The second group is more controversial: workers brought in to fill standard technical roles across the industry. Some are employed through large consulting and outsourcing firms, often referred to in industry discussions as “WITCH” companies, such as Wipro, Infosys, TCS, Cognizant, and HCL. These firms frequently sponsor large numbers of H-1B applicants and place them with client companies across the United States. In these arrangements, the consulting firm employs the worker and contracts their services to other businesses, keeping the margin between what the client pays and the employee’s salary.
But this category is not limited to outsourcing firms. Large companies also hire many H-1B workers directly into ordinary software engineering roles alongside domestic employees.
It is difficult to know the exact balance between these groups. What is clear is that the system has generated significant debate. Critics argue that it expands the supply of labor and puts downward pressure on wages, while supporters argue that it helps companies access global talent and remain competitive.
Given the scale of tech layoffs over the past few years, however, the claim that the United States faces a broad shortage of tech talent has become increasingly contested. That question deserves its own deep dive, but for now, let’s move on.
OPT: Optional Practical Training
Optional Practical Training, or OPT, is a work authorization program that allows international students on F-1 visas to work in the United States for up to 12 months after graduation in a role related to their degree. If you studied a STEM field — which virtually all CS graduates qualify for — that window extends to a full 36 months. In Fiscal Year 2024, just under 110,000 student visa holders were actively on U.S. company payrolls through this program, with that number likely growing year over year. There are currently 1.1 million international students in the U.S. with over 200,000 eligible for OTP.
Here is why this matters: the vast majority of these workers are competing for the same entry-level roles you are. And it is genuinely difficult to argue that most of them cannot be filled by qualified domestic graduates. The uncomfortable reality is that OPT has evolved far beyond its original intent as a practical training program. For many international students, it is explicitly and openly used as a structured multi-year pathway to U.S. permanent residency, functioning as a bridge to H-1B sponsorship and eventually a green card.
This has given rise to a massive, largely unspoken industry. Hundreds of U.S. universities — including many second and third-tier programs that few would attend purely for educational reasons — now aggressively market one and two-year master’s programs almost exclusively to international students. The pitch is straightforward and everyone involved understands it: pay $40,000–$75,000 in tuition, earn a STEM-designated master’s degree, and unlock 36 months of OPT work authorization plus multiple shots at the H-1B lottery. The degree is, in many cases, secondary to the visa access it provides. Universities benefit from the revenue and employers benefit from a pool of candidates who, because their visa status depends entirely on staying employed, are willing to accept lower compensation, longer hours, and less favorable conditions than domestic candidates would reasonably tolerate.
Other Visas: There exists other types of visa like L1. The L-1 is an intracompany transfer visa with no annual cap and no lottery. Any qualifying multinational can move employees from their foreign offices into the U.S. indefinitely, as long as they meet the basic eligibility criteria. The numbers tell the story of how aggressively it has been used: in 1980, the U.S. issued 26,535 L-1 visas. By the late 1980s that figure averaged 60,000–70,000 entries per year. By 2017, the total L-1 worker population had grown to 374,234. The dominant users are not American companies, but major IT outsourcing firms — particularly Tata Consultancy Services (TCS) and Infosys.
The broader picture is this: the L-1, with its no-cap structure, no wage floor, and minimal oversight relative to its scale, has become one of the most quietly effective tools for deploying large volumes of foreign labor into the U.S. tech market — with domestic workers having little visibility into how extensively it is being used.
PERM: The Final Piece of the Puzzle
PERM — Program Electronic Review Management — is the Department of Labor’s mandatory labor certification process that serves as the first step toward an employment-based green card. Before sponsoring a worker for permanent residency, a company must formally demonstrate that no qualified American was available for the role. On paper this sounds like a protection for domestic workers. In practice it is largely a compliance exercise — the approval rate sits above 95%. As of early 2026.
Once approved, the worker joins a queue determined by their country of birth. For most countries the wait is short. For Indian nationals — who represent the overwhelming majority of the pipeline — it is staggering. The current EB-2 priority date for India sits around 2013–2014, meaning the government is processing applications filed over a decade ago. With roughly 395,000 approved petitions waiting and only 2,800–3,000 green cards available to India annually due to the 7% per-country cap, someone filing today could wait 15 years or more.
What one must realize is that you and your colleague on a visa are playing completely different games. For them, securing residency is the primary objective. Lower wages, longer hours, and less favorable conditions are not just tolerable, they are a rational trade-off when the alternative is losing your immigration status and everything you have worked toward.
5) Offshoring and the Infinite Labor Supply
Many companies are simply moving the work offshore entirely, building what the corporate world has quietly rebranded as Global Capability Centers, or GCCs.
A GCC is a company’s own captive offshore operation — not an outsourced vendor, but an owned subsidiary that handles product development, engineering, AI, data, and increasingly, core business functions. The old framing was “back office support.” The new reality is that these are full engineering teams building the same products that were once built in California, Austin, or Seattle. India alone is currently home to over 1,700 GCCs employing 1.9 million professionals and generating $64.6 billion in annual revenue. More than 65% of those centers are owned by U.S.-headquartered companies. By 2030, that market is projected to reach $100 billion, with nearly 50% of Fortune 500 companies expected to run GCCs in India.
Meta, Google, Apple, Amazon, Microsoft, and Netflix collectively added over 32,000 jobs in India in 2025 alone — an 18% year-over-year jump and the strongest hiring stretch since 2022. Google plans to open new offices in India too with up 20,000 staff.
Albertsons is perhaps the starkest example of how far this has spread. This is not a tech company — it is a grocery chain, operating 2,270 stores across 34 U.S. states under banners like Safeway, Vons, and Jewel-Osco. They barely hire full time tech employees in the U.S. anymore. While they are rapidly expanding their GCC, expecting to hire 1000 technologists in 18 months. Albertsons GCC Career page.
A similar story holds for most of the American companies rapidly expanding headcount elsewhere with a shrinking employee base in the U.S. Uber is an example of a top technology company rapidly doing it which can be tracked via their website and Linkedin.
What makes this wave different from every previous offshoring cycle is that the conditions that caused it to fail before have significantly improved. Global talent quality has genuinely gone up. Engineers abroad have better education and more experience now while requiring less compensation, are more fluent in English, have access to the same books, the same lectures and online courses, the same AI coding tools as anyone in the US, and there are far more tools for building stable and scalable software with guardrails than there were two decades ago. Georgia Tech’s OMSCS gives anyone in the world top education for cheap where 36% are international students. The knowledge gap and advantages that once justified paying an American engineer three to four times more have narrowed significantly. Management has become better at running distributed teams, and communication tools have matured. Software itself has also matured. Cloud infrastructure, managed services, modern frameworks, extensive documentation, Stack Overflow, GitHub, and AI coding assistants have made building applications more predictable and accessible than ever before.
There is no top, mid, or low level software engineering job that you can reliably hold long term. The industry is facing pressure on all fronts. Companies now effectively have access to an almost unlimited global labor pool, and frankly I am surprised wages have not collapsed yet.
What I mean is that countries such as India, Brazil, Mexico, Bangladesh, the Philippines, Poland, and Pakistan together have more than 2 billion people. Only a minority of their populations currently hold a bachelor’s degree, but even that minority is already enormous, and the talent pool will likely continue to grow as more people gain access to higher education.
Many of these countries also have very young populations. India’s median age is about 26, compared with roughly 39 in the United States. The millions of engineers in the U.S. barely make up 1% of the combined population of those countries.
Given how modular software engineering, and many other engineering fields, have become, and the fact that human labor is the primary input, it is relatively easy to move work across locations. Once wages rise in a given country or uncertainty increases, companies often shift work to the next lower cost location.
Africa represents a relatively untapped labor pool, with a population of about 1.5 billion and a median age under 20. It is unlikely to remain untapped forever.
If you want a preview of how this plays out, look at manufacturing. When China entered the WTO in 2001, the US employed about 17 million in manufacturing jobs, it dropped to 13 million by 2008 and now stands at 12.5 million. If anything, manufacturing was harder to offshore, it had physical infrastructure, expensive machinery, and complex supply chains. And what happened politically? Not much. A few retraining programs that mostly didn’t work and some tariffs that came twenty years too late. The entire thing has been really painful. I am not sure if there is a reason to believe the story ends differently for techies.
6) Quant/FAANG+/Unicorns or Bust
Everything in this article points to an industry that is volatile, globally commoditized, and increasingly indifferent to the domestic engineer who does not sit at the very top of it. Which leads to a conclusion that is uncomfortable but follows directly from the evidence: the FAANG+/quant obsession that looks like status-chasing from the outside may actually be the only coherent strategy from the inside.
The core problem is whether your employer views engineering as a cost center or a profit center. At most companies — the Albertsons, the regional banks, the mid-market SaaS firms, the Fortune 500 building its GCC — engineers are overhead. When the board asks for fiscal discipline, engineering headcount is one of the first levers pulled. Engineering headcount is reduced because they were expensive, and replaceable at lower cost somewhere else.
While you are not immune to problems at any of these places, the pay is materially higher which gives you a bigger safety cushion, and more resume value that can be cashed in faster when things go sideways, but unfortunately getting a foot in the door in these companies is getting continuously harder.
Ironically, quant seems to be holding up the best. Firms like Jane Street are crushing it, with revenue per head approaching $10 million. But to land and keep a job at these firms you need to be a different beast entirely, and if you are, you will likely be fine.
What separates quant firms from most of the tech industry is that engineering is not a support function. At firms like Jane Street, Citadel, Two Sigma, and Hudson River Trading, the engineers, researchers, and traders are directly tied to the firm’s core profit engine. Every improvement in latency, every optimization in a trading model, every better data pipeline can translate almost immediately into trading profit. That creates an economic structure that is very different from most corporate technology departments.
7) Cost Cutting: The CFO and CEO's Endless Efficiency Theater
For most of the 2010s, the dominant religion in tech was growth. Not profit, growth. Revenue multiples, monthly active users, and year over year expansion were what mattered. Profitability was treated as a problem for later, something that could be addressed once scale had been achieved. Investors rewarded this mindset, public markets rewarded it, and companies hired aggressively because headcount itself became a signal of ambition and momentum. Large engineering teams suggested velocity, product expansion, and the ability to capture market share before competitors did. The sentiment has changed since then, and focus has gone toward operational efficiency.
Quarterly earnings reports amplify this pressure. Public companies operate under a relentless cycle where every ninety days the market re-evaluates the entire business. A single disappointing quarter can erase billions in market value overnight. A good example is Bumble. Despite generating over $1 billion in revenue in 2024, its market capitalization has fallen below that revenue level largely because growth has slowed and investors no longer believe the company has a strong expansion story. The stock has dropped roughly 80 percent from its peak. Similar dynamics are unfolding across large parts of the SaaS sector.
Salaries are the largest operating expense at most companies, and labor costs can reach roughly 60 to 70 percent of total operating expenses in many organizations. Within that payroll line, software engineers are often the most expensive employees in the building. A senior engineer in a major tech hub may cost $150,000 to $250,000 in base salary alone, before equity, benefits, and employer payroll taxes are included. Once those additional costs are factored in, the total cost of employing a single engineer can be substantially higher.
When a CFO is instructed by the board to reduce expenses, engineering headcount quickly becomes one of the most visible levers. A small number of high paid roles can represent millions of dollars in annual cost. And increasingly, executives believe they have alternatives available to fill the gap, whether through outsourcing, contractors, or lower cost global teams. At many companies, engineering is not viewed as a direct source of revenue. It is viewed as infrastructure, a cost that must be controlled rather than a profit center that must be expanded.
8) Layoffs and Instability, PIPs: The Ugly Stuff
The median tenure for wage and salary workers fell to 3.9 years in January 2024, according to the Bureau of Labor Statistics — the lowest since January 2002. The tech industry runs well below even that deteriorating average. For software engineers specifically, average tenure sits around three years and even less for many big tech companies.
Amazon and Meta: Institutionalized Attrition
Amazon formalized workforce churn into a policy called “unregretted attrition.” The target: managers are required to push out at least 6% of their staff every year. The mechanism is a process called Pivot — their internal PIP. Employees are evaluated on a forced curve, with 10 to 15% of any organization required to be rated Least Effective, regardless of absolute performance. Former Amazon VP, Ethan Evans straight up encourages one to accept his fate and leave the company as rarely if anyone survives it. By mid-2025, Meta had expanded its “below expectations” bucket to 15–20% of staff
Amazon and Meta are among the most widely documented examples, but the practice is not unique to them. The broader issue with these approaches is that they create a culture of instability and uncertainty, making it difficult for employees to confidently plan their lives around their current employer.
When workers know that there may be a meaningful risk of layoffs or performance based cuts in any given year, long term personal decisions become harder to make. Something as basic as buying a home near the office or starting a family becomes a risk calculation. If there is a significant chance that your role could disappear within a year or two, committing to geographic stability or long term financial obligations starts to feel much less secure.
The unending Tsunami of layoffs:
Tech layoffs surged in 2022, with an estimated 244,000 tech workers laid off. That figure doubled in 2023, reaching around 430,000. In 2024, companies laid off nearly 243,000 employees. In 2025, another 246,000 were let go. The four-year cumulative total from 2022 through 2025 approaches over a million tech workers. It is unclear where we are headed as of now, but the recent Block and Amazon layoffs are not promising news at all.
Block had a massive unprecedented layoff at 40% of the entire staff. AI productivity was cited as the primary reason, but I am not sure that is the whole story. The company’s headcount had grown very rapidly in the years prior, and there may be deeper structural issues at play. It warrants its own deep dive.
9) The Million Strong Shadow Work Force
When a company announces its headcount, it is not telling you how many people actually work there. It is telling you how many people it has chosen to legally employ. Those are very different numbers.
Beneath the official count sits a parallel workforce — temps, vendors, and contractors who do essential work, sit in the same buildings, use the same tools, often report to the same managers, and yet appear nowhere on the company’s books. They receive no stock, minimal or no benefits, can be terminated without severance at any time, and are bound by NDAs that prevent them from discussing their conditions. The mechanism that makes this possible is simple: they are formally employed by staffing agencies such as Adecco, Cognizant, Randstad, Accenture, and dozens of smaller firms, not by the company itself.
Google’s shadow majority
Google calls them TVCs: temps, vendors, and contractors. By 2018, Bloomberg confirmed that TVCs had crossed a threshold that should have been front-page news: contractors had officially outnumbered full-time employees for the first time in the company’s history. As of early 2019, Google had roughly 102,000 direct employees — and approximately 121,000 TVCs. The shadow workforce was the majority workforce. By some estimates, the TVC count had reached 130,000 to 150,000 before the pandemic.
Microsoft’s orange badges
Microsoft runs the same architecture under different terminology. Full-time employees are “blue badges.” Contractors are “orange badges,” subdivided into “a-dash” temporaries and “v-dash” vendors. Around 2014, when Microsoft employed about 100,000 direct staff, it simultaneously had more than 71,000 contractors — a shadow workforce exceeding two-thirds the size of its disclosed headcount. These contractors were not peripheral; they were deeply embedded in building, testing, and marketing Microsoft’s core products.
Apple’s Contractor Machine
Based on industry estimates, Apple employs roughly 65,000 contractors, maintaining close to a 1:1 ratio between contractors and full time employees in parts of its U.S. workforce. While the company does not publicly disclose exact figures for its contingent workforce, contractors are widely used across functions such as IT operations, software testing, content review, HR support, mapping, and technical support.
Apple has roughly 130,000 full time employees globally in 2019, and according to staffing industry data compiled by OnContracting, the company works with several dozen staffing firms to supply contract labor. These contractors support a wide range of internal and customer facing systems. Some work on infrastructure supporting services such as iTunes and backend platforms, while others handle customer support operations or curate content for Apple News.
One example is Apex Systems, a major division of ASGN Incorporated. Apex has provided Apple with mapping technicians responsible for validating the accuracy of Apple Maps data, such as verifying that roads are correctly represented or responding to reports of mapping errors submitted by users.
Tech giants collectively employed over 1.2 million contractors. The practice is standard across every major company now, and many of them also happen to be engineering, QA, SRE, Project management roles, while most are in support roles. The practice has become so common that companies like Goldman Sachs ask if you are a contractor in their job application. It is hard to find the exact figure of contractors in the U.S.. The number is certainly in the hundreds of thousands at the very least. Also, when layoffs are reported, the numbers typically only include full time employees. What often goes unreported is what happens to contractors, and how many of them quietly come and go over time.
10) The Specialist vs Generalist Duality: Both Come With a Price
The specialist goes deep. They become genuinely expert in areas like distributed systems, compilers, embedded systems, robotics, or ML infrastructure, narrow domains that can command a premium when demand is strong. The trade off is that specialization narrows your options quickly. You may become unqualified for the majority of job postings outside that niche. While your response rate from relevant companies may improve, the total number of opportunities shrinks. Those roles tend to exist at a limited set of companies, often clustered in specific geographies, and when that segment of the market slows down there may be very few places left to go. Deep specialization also usually requires years of focused investment, often including advanced degrees and continuous effort to stay current as the field evolves.
A lack of optionality can also weaken negotiating leverage. If your skills are only valued by a small number of employers, those employers face less competitive pressure to bid aggressively for talent. This dynamic may help explain something that often seems counterintuitive: hardware engineers have historically earned less than software engineers despite the work frequently requiring more specialized knowledge. Part of this difference likely comes from software’s scalability. Software can be deployed globally at near zero marginal cost, which allows companies to justify higher compensation for engineers who build it. But another factor is labor market structure. When your skills are transferable across thousands of companies, your bargaining power increases. When they are relevant to only a handful of employers, that leverage declines regardless of how difficult those skills were to acquire.
There are, of course, exceptions. Timing and domain choice matter enormously. Engineers who developed deep expertise in GPUs and joined a company like Nvidia, or researchers who entered advanced AI work at organizations such as OpenAI or Anthropic, have benefited from enormous demand and compensation growth. But these outcomes often depend on aligning specialization with the right technological wave. For every person who specialized in an area that exploded in value, there are many others who invested deeply in domains that later stagnated or became commoditized.
The generalist faces the opposite challenge. They are adaptable, capable of learning new technologies quickly, and able to contribute across multiple teams or problem areas. That flexibility is valuable in many organizations. However, in a hiring market that filters heavily for specific signals, breadth without depth can become a disadvantage. Specialized companies often design interview processes to identify candidates who have gone extremely deep in one area. Generalists may struggle to pass those screens even if they are highly capable engineers. At the same time, their broad and transferable skill sets make them more exposed to global competition, since adaptable software work is often the easiest to move across geographies.
11) The Treadmill That Never Stops: Upskilling, Bar Raisers, and Ageism
Software engineering is one of the few professions where your technical knowledge can have a visible expiration date. Consider the history of mobile development alone. Early mobile apps were written using Java ME for feature phones. When the iPhone launched in 2007, development centered on Objective C. Then Android entered the picture and developers had to learn Java for a different ecosystem. In 2014, Apple introduced Swift, effectively replacing Objective C and forcing iOS developers to learn a new language and toolchain. Soon after, the industry began pushing cross platform frameworks such as React Native, Xamarin, and Flutter. Each transition brought new tooling, new abstractions, and new expectations in interviews. An engineer who was deeply experienced in Objective C around 2013 often had to reinvent their skill set within just a few years. This is only one narrow corner of the field, yet it illustrates how quickly paradigms can shift.
The upskilling treadmill is relentless. Every few years a new technological wave arrives, and the expectation is that engineers will absorb it largely on their own time while continuing to perform in demanding roles. Not long ago the focus was on cloud infrastructure and distributed systems. Today the conversation revolves around retrieval augmented generation pipelines, large language model integration, and AI assisted development. The industry often frames this constant reinvention as professional growth. In reality it is simply the ongoing cost of remaining employable in a field that evolves unusually quickly.
At the same time, the interview process has become an arms race of its own. Because software engineering lacks the formal licensing structures found in professions like law or medicine, hiring processes have become the primary mechanism for filtering candidates. Companies continuously raise the bar, adding layers of algorithmic interviews, system design evaluations, and technical screens. What began with large tech firms has spread throughout the industry. Coding challenges popularized by platforms like LeetCode are now common even at mid sized companies. Preparing for these interviews has become a substantial commitment, sometimes resembling a part time job in itself.
Age dynamics add another layer of pressure. In technology, the pace of change can make it harder for some engineers to stay aligned with new tools, languages, and interview expectations over time. While many experienced engineers successfully adapt and continue thriving, others find the constant reset more difficult. The industry also tends to reward perceived “cutting edge” experience more than tenure alone, which weakens the traditional experience premium seen in other professions. In places like Silicon Valley, engineers in their forties sometimes encounter skepticism that would be unusual in other fields.
These forces reinforce one another. Engineers are expected to continually acquire new skills, prepare for increasingly demanding hiring processes, and do so within a labor market that may value their profile most strongly during a relatively narrow window of their careers. The result is a profession where stability often requires constant reinvention.
12) No Ladder, Just a Glass Ceiling: Stagnation and an Ambiguous Promo Path
One of the least discussed realities of a software engineering career is that the path upward is not really a ladder. It is a pyramid that narrows quickly, and most people stop climbing much earlier than they expect.
Every large technology company runs on a leveling system. At Google engineers progress from L3 to L4 to L5 and upward. L5 corresponds to Senior Software Engineer, followed by L6 Staff, L7 Senior Staff, L8 Principal, L9 Distinguished Engineer, and L10 Fellow. At Amazon the equivalent path is SDE I (L4), SDE II (L5), Senior SDE (L6), Principal Engineer (L7), and Senior Principal (L8). At Meta engineers move through levels E3 to E10, while at Microsoft the progression typically runs from Level 59 or 60 for entry level engineers, to 63 or 64 for Senior Software Engineers, then 65 or 66 for Principal Engineers, followed by Distinguished Engineer and Technical Fellow at the very top. These levels determine compensation, influence within the organization, and how the company evaluates an engineer’s long term trajectory.
What is less obvious to people entering the industry is that companies only promote when the organization needs a larger scope filled. If there is no increase in scope or organizational need, there is no reason to move someone up the ladder. In practice, many companies are perfectly comfortable keeping engineers at the same level for long periods as long as the work continues to get done.
The distribution of levels makes this dynamic very clear. At Meta, internal estimates suggest that roughly 15 percent of engineers reach E6, which corresponds to Staff level. Only around 3 percent reach E7, and fewer than 1 percent reach E8, the Principal level responsible for initiatives that span entire product lines such as major parts of Instagram or Facebook’s marketplace infrastructure. E9, Distinguished Engineer, represents a fraction of a percent of the engineering population and carries influence comparable to senior executive leadership.
At Microsoft, Level 63 or 64 Senior Software Engineer is often described as the practical ceiling for many engineers. Principal Engineers and higher levels exist but represent a sharp drop off in headcount. Roles such as Distinguished Engineer or Technical Fellow are so uncommon that many engineers will never encounter one during their entire career. Microsoft has fewer than 60 Distinguished Engineers with a headcount of over 200,000.
The promotion process itself is also famously opaque. Advancement usually requires more than simply performing well. An engineer typically needs a manager willing to sponsor the promotion, strong performance reviews, peer feedback that supports the case, and a written promotion packet that must pass a committee review. Engineers often operate at the scope of the next level for a year or longer before receiving formal recognition. Some never do.
The bar for promotion also shifts quietly with the broader economic environment. During hiring freezes or cost cutting cycles, promotion rates often drop without formal announcements. Engineers who believed they were on track for advancement sometimes discover that expectations have changed or timelines have been extended.
Promotion criteria themselves are written in deliberately broad language. Concepts such as impact, influence, scope, and autonomy appear in nearly every promotion rubric, yet none of them are objectively measurable. Two engineers doing work of similar quality can have very different outcomes depending on their team, their manager’s ability to advocate for them, and whether their projects happened to be visible to the right leadership at the right moment.
For many engineers the result is a long plateau. They reach the Senior level, continue doing solid work, and see compensation grow only gradually. The next level begins to feel permanently out of reach, not necessarily because they are underperforming, but because the system was never designed to promote most people past that point. The ceiling exists, even if it is presented through leveling systems and promotion frameworks that give the impression of a pure meritocracy.
In stronger hiring markets, engineers had at least one practical way around this ceiling. If internal promotion timelines stalled, they could interview elsewhere, secure a higher title or compensation band, and effectively “reset” their level by switching companies. Job hopping functioned as a parallel promotion system across the industry. In the current climate, that escape valve has weakened considerably.
13) Profits Up, Paychecks Flat: Five Years of Going Nowhere
The story of tech worker compensation between 2018 and 2025 is not one of dramatic collapse. It is quieter and in some ways more disturbing: a seven-year period in which wages nominally appeared to grow while quietly going nowhere in real terms.
The Nominal Illusion
Start with the raw numbers from the Bureau of Labor Statistics Occupational Employment and Wage Statistics (OEWS) survey, which tracks median annual wages for software developers nationally:
Start with the Bureau of Labor Statistics OEWS median wage series for software developers. In May 2019, the national median was $107,510. In May 2020 it was $110,140. In May 2021 it jumped to $120,730. In May 2022 it reached $127,260. By May 2024, the Occupational Outlook Handbook summary lists the median software developer wage at $133,080.
Taken in isolation, that trajectory looks respectable: from roughly $104K in 2018 to $133K in 2024, a nominal gain of around 28 percent over six years. But the story ends when you hold those numbers against inflation.
The Inflation Reality
Compounded across those seven years, the cumulative price increase from the end of 2017 to the end of 2024 amounts to approximately 28 percent. The median tech worker’s nominal salary grew by almost exactly the same amount. In real purchasing power terms, the median software developer in 2024 earned essentially the same as their counterpart in 2018, despite the massive bull market over this period.
Phase One: The Pre-Pandemic Plateau (2018–2020)
The period from 2018 to 2020 was relatively placid. Nominal wages grew from $103,620 to $110,140, a 6.3 percent increase across two years, while inflation over the same period was approximately 3.7 percent compounded. That was modest real growth, around 2.6 percent over two years in total, which is roughly in line with long-run historical norms. Tech workers were not being cheated yet; they were simply not getting rich from their salaries. The real compensation story was in equity grants at high-growth companies, which inflated total compensation far above what salary figures suggested.
Phase Two: The Pandemic Mirage (2020–2021)
The pandemic created the appearance of a breakthrough. The median annual wage for software developers jumped to $120,730 in May 2021 from $110,140 in May 2020, a 9.6 percent nominal increase in a single year. Consumer spending on digital products surged, companies scrambled to hire, sign-on bonuses inflated total compensation packages, and tech workers appeared to be entering a golden era. This was the period that generated the exuberant discourse about software engineering as the ultimate career.
The reality was more complicated even at the time. Inflation remained subdued at 1.4 percent in 2020 before surging to 7.0 percent in 2021, meaning the apparent salary boom of 2021 was almost entirely consumed by the very inflation that was beginning to accelerate. The gap between salary announcement and salary purchasing power was already narrowing.
Phase Three: The Inflation Erasure (2021–2023)
This is the cruelest chapter. The 2022 CPI came in at 6.5 percent year-over-year by December. This was the highest sustained inflation the United States had seen in four decades. For tech workers, this arrived at the exact moment when the hiring surge reversed. Layoffs began at scale in late 2022, hiring freezes became common, and leverage shifted decisively back to employers. Workers who had expected further salary increases found themselves navigating a market where firms were actively conducting reductions in force and implementing performance management systems designed to reduce headcount.
The direct data from compensation platforms confirmed what aggregate statistics implied. According to Levels.fyi’s 2022 compensation report, software engineer median total compensation fell 2.2 percent in nominal terms in 2022 against an inflation backdrop of more than 8 percent for the calendar year on average. That is a double-digit real wage decline in a single year. Market data from WeAreDevelopers found that U.S. tech salaries moved from $111,348 in 2022 to $111,193 in 2023, essentially flat nominally at a time when the average annual inflation rate for 2023 was 4.1 percent.
The ADP Research Verdict
The most damning summary comes not from a single year but from the full arc. Between January 2018 and January 2024, the median base pay for developers grew by 24 percent while pay growth for total U.S. workers grew 30 percent. ADP This is a finding with significant implications. It means that not only did tech workers fail to preserve their purchasing power in real terms; they also fell behind the broader American workforce during a period when their industry generated some of the most spectacular profit concentrations in economic history. The programmers and engineers who built the infrastructure of the modern digital economy were outpaced on compensation by the average U.S. worker.
In January 2024, the U.S. employed fewer software developers than it did six years ago.Falling employment combined with below-market wage growth is not the portrait of a high-value profession in command of its own destiny. It is the portrait of a labor market that has been systematically rebalanced in favor of the buyer.
The Shareholder Contrast
The numbers above describe only base wages, and base wages do not capture total compensation at large public companies where equity is a significant component. That framing is sometimes used to suggest that tech workers did better than salary data indicates. The equity argument has merit for a narrow cohort of workers at large-cap companies who held stock from the trough of 2022 to the subsequent recovery. It has little merit for the majority of the workforce.
What the equity comparison does do is illuminate where the returns actually went. The same companies that were holding salary budgets flat and reducing headcount were simultaneously returning capital to shareholders at historic rates. Apple conducted $104.2 billion in share buybacks in 2024 alone, part of a ten-year total exceeding $663 billion. Alphabet’s cumulative buybacks over the same decade exceeded $255 billion. Microsoft returned over $192 billion to shareholders over ten years while simultaneously presiding over rounds of layoffs and a wage trajectory that, in real terms, gave workers nothing.
The question this data forces into the open is not whether tech workers are well-paid in absolute terms. They are. The question is whether a workforce that generated unprecedented returns for shareholders over a seven-year period was compensated commensurately with its contribution. The BLS data, the ADP Research data, the Levels.fyi data, and the inflation record all give the same answer: it was not. The gains were real. Their allocation was not neutral. And the median software developer in 2024 bought no more with their paycheck than their predecessor in 2018, in an economy that became materially more expensive across every dimension that matters to a person trying to build a life.
The FAANG Exception: Real Gains, But Not When You Look Level by Level
The one genuine counterargument to the wage stagnation thesis is the compensation trajectory at the largest consumer tech companies. To evaluate it properly, the comparison must be done the way you would evaluate any wage claim: like for like, the same level at the same point in the career ladder, with inflation applied.
At the entry level, the data is almost startling in its flatness. In June 2019, according to levels.fyi data reported by CNBC, a Google L3 engineer — the standard entry-level hire fresh out of college — received approximately $189,000 in total compensation, with roughly $124,000 in base salary and $43,000 in stock. A Facebook E3 at the same point received around $166,000. By 2024, the median total compensation for a Google L3 in the United States was $188,805, and Facebook’s entry-level package had moved to roughly $195,000. In nominal terms, Google’s entry-level package had gone nowhere in five years, and Meta’s had risen by approximately 17 percent. Cumulative inflation between 2019 and 2024 was approximately 23 percent. In real purchasing power terms, the Google entry-level engineer of 2024 was earning meaningfully less than their 2019 counterpart. The Meta entry-level engineer barely broke even in real terms, and only because Meta made a deliberate decision to increase starting packages after the talent war of 2021.
At the senior level, the picture is somewhat better, but not as different as advocates of FAANG compensation often suggest. In 2019, a Google L7 engineer — the practical ceiling for the vast majority of engineers over a full career — received around $608,000 in total compensation according to levels.fyi. By 2024, the levels.fyi annual pay report recorded Google L7 at approximately $728,000. That is a nominal gain of roughly 20 percent over five years against inflation of 23 percent. Senior FAANG engineers also largely went backwards in real terms, though by a narrower margin than their entry-level colleagues. Google’s L5 senior software engineer median total compensation currently sits at approximately $420,000 according to levels.fyi — a figure that sounds impressive in isolation but is almost identical in nominal terms to where it stood years earlier. Levels.fyi’s own 2018 annual report placed Google L5 at $407,000. A nominal gain of roughly three percent over six years, against cumulative inflation exceeding twenty percent, means the senior cohort has gone backwards in real purchasing power terms by nearly as much as their entry-level colleagues — the dollar figures are just larger.
The conclusion from a level-by-level reading is not that FAANG workers did well while everyone else stagnated. It is that essentially the entire profession, from entry-level hires at the most prestigious companies on earth to the median developer at an ordinary employer, failed to keep pace with the cost of living over a seven-year period that generated extraordinary returns for the shareholders of the companies they worked for.
14) The Cost of Living Trap: High Salaries, Higher Rent
The compensation discussion that dominates tech career culture operates almost entirely in nominal figures. A software engineer making $189,000 in San Francisco and a software engineer making $110,000 in Columbus, Ohio are compared as if the dollar amounts speak for themselves. They do not. The geography of the American tech industry is not incidental to the wage question; it is central to it. The cities where the overwhelming majority of software engineering jobs are concentrated happen to be the same cities where the cost of a basic life has grown so expensive that even salaries most Americans would consider extraordinary are being systematically consumed by housing, taxation, and routine living expenses. The high salary is in many cases not a reward. It is a partial offset for an artificially inflated cost environment that the employer helped create.
The rent figures alone arrest the argument before it can get started. According to Zumper’s National Rent Report, New York, San Francisco, and Boston are the three most expensive one-bedroom rental markets in the country, at $4,250, $3,630, and $2,930 respectively. Against these figures, the national median one-bedroom rent stands at $1,499. A software engineer living in New York is paying 2.84 times the national median rent before they have bought a single grocery item, filled a gas tank, or paid a utility bill. In San Francisco, the multiplier against the national median is 2.42. In Boston it is 1.95. Even Los Angeles, the next major tech hub on the list, extracts 1.53 times the national median for a single-bedroom unit. These are not abstract cost-of-living statistics. They are the fixed monthly obligations that come due before any discretionary spending begins, and they fall on exactly the workers whose compensation the FAANG premium is supposed to explain.
The homeownership picture is, if anything, grimmer. The national median home price as of January 2026 sits at approximately $423,000, according to Redfin. Against that baseline, the divergence in tech-hub markets is severe. San Francisco’s median home sale price stands at $1.3 million, up 2.8% year over year. Los Angeles comes in at a median of $975,000 for the city proper. Boston’s median home sale price is $825,000 as of January 2026. These are not luxury properties. These are median figures, meaning half the available housing stock in these markets costs more. The consequence is not merely that homeownership becomes difficult. It is that it becomes a different financial category entirely. Buying a median-priced home in San Francisco requires an annual income of $321,463, pushing monthly mortgage costs above $7,500. In New York City, buyers need $200,280 annually, virtually double pre-pandemic levels. In Boston, the income threshold has surged from $101,895 in late 2019 to $190,858 today. In Los Angeles, the required salary is $224,190. The national threshold, by contrast, is $106,731 — meaning that of the four cities under analysis, even Boston, the most accessible of the group, requires nearly twice the income a median American household needs to purchase a median home anywhere in the country.
The cost difference does not stop at housing. San Francisco’s cost of living overall is 71 percent above the national average, according to RentCafe data. Bankrate Numbeo’s mid-2024 global ranking placed San Francisco fourth in the world for cost of living, behind only Geneva, Zurich, and New York City. KRON4 Groceries, transportation, healthcare, and restaurant prices are all materially elevated relative to the median American city. The MIT living wage calculator estimates food costs for a single adult in California run approximately $4,500 per year, compared to $3,812 in Texas. These secondary costs accumulate into thousands of additional dollars per year that do not appear in any salary table but are extracted from every paycheck regardless.
Then comes taxation, the most consequential variable that virtually no salary comparison headline ever mentions. California’s top state income tax rate is 13.3 percent, the highest in the nation. New York’s top rate is 10.9 percent, with New York City residents paying an additional city income tax of approximately 3.9 percent on top of that. Massachusetts applies a 5 percent flat rate with a 4 percent surcharge on high incomes. TurboTax Against these figures, nine states including Texas and Florida levy no state income tax whatsoever. A high-income earner making $200,000 in California pays approximately $16,000 to $18,000 in state income tax, while paying zero in Texas or Florida. Mynetpay A software engineer earning $189,000 in San Francisco does not take home $189,000. After federal income tax, California state income tax, Social Security, and Medicare contributions, actual take-home pay is closer to $118,000 to $122,000 depending on deductions. That same $189,000 salary in Austin, Texas, with no state income tax, yields approximately $130,000 to $135,000 in annual take-home pay. The San Francisco engineer hands over an extra $12,000 to $15,000 per year to the state before they have paid a single dollar toward the rent that is itself $24,000 per year higher than the Austin median.
The composite effect of these forces on real, spendable purchasing power is what the raw salary number obscures. A software developer earning the national BLS median of $133,080 in Columbus, Ohio, pays roughly $1,200 per month in rent, faces a home purchase threshold of approximately $80,000 in required income, and pays Ohio’s flat state income tax of around 3.5 percent on that bracket. After federal and state taxes, their take-home pay is approximately $93,000. Their housing costs consume roughly $14,400 per year, leaving approximately $78,000 for all other expenses and savings. Their counterpart in San Francisco earning $189,000 takes home perhaps $120,000 after California taxes and pays $41,940 per year in rent for a one-bedroom apartment at the city median. That leaves approximately $78,000 for all other expenses and savings. The San Francisco engineer earns 42 percent more in nominal terms. They have approximately the same disposable income after housing and taxes. If that San Francisco engineer has a partner and a child and needs a two-bedroom apartment, their annual rent at the San Francisco median rises to approximately $61,440, wiping out every nominal advantage their salary holds over the national average entirely.
The worker who moves from Iowa to San Francisco for a $60,000 salary increase has in many cases accepted a pay cut measured in purchasing power, a longer commute, a smaller living space, and a housing market that has structurally locked them out of the wealth-building mechanism available to their parents. Nationally, median single-family home prices rose nearly 48 percent between 2019 and 2024, at more than twice the rate of median income growth of 22 percent. In San Francisco and Los Angeles, the divergence is more extreme still. The conversation about whether tech workers are fairly compensated cannot be conducted in nominal salary figures alone. It must account for where those salaries are paid, what the state retains, what the landlord extracts, and what a house actually costs in the city where the work exists. Measured against those realities, the premium the industry advertises turns out, for most workers, to be considerably thinner than the headline numbers suggest.
15) The Unknown Looming Dangers (and Possible Benefits) of AI
Let me be direct about something most analyses of this topic avoid: nobody knows. Not the CEOs announcing workforce reductions in AI’s name, not the academics publishing competing studies on productivity, not the venture capitalists who poured over $250 billion into AI infrastructure in a single year. The honest position is one of genuine uncertainty, and anyone who tells you otherwise is either selling something or has confused the volume of their conviction for the weight of evidence behind it.
What we can say with more confidence is something narrower: the prevailing narrative about AI is already reshaping corporate hiring decisions regardless of whether that narrative is accurate, and for most workers, the near term is what actually matters.
The productivity data should give pause to anyone who has confidently declared either that AI will solve everything or destroy everything. A National Bureau of Economic Research study surveying 6,000 CEOs and executives across the US, UK, Germany, and Australia found that the vast majority report little impact from AI on their actual operations. Apollo chief economist Torsten Slok observed that “AI is everywhere except in the incoming macroeconomic data,” noting no visible signs of AI in employment data, productivity data, or inflation data. Nobel laureate Daron Acemoglu, whose modelling suggests a 0.5 percent productivity increase over the next decade, said the result was “disappointing relative to the promises that people in the industry and in tech journalism are making.” This is the return of the productivity paradox: powerful technology widely adopted but not yet visible in aggregate output statistics. It happened with computers in the 1970s and 1980s.
Yet companies are acting on the narrative regardless. Of the 1.2 million job cuts U.S. companies announced in 2025, AI was cited as a direct reason for just 55,000, or 4.5 percent of them, according to Challenger, Gray & Christmas. Companies have discovered that framing layoffs as “restructuring for the AI era” plays better with investors than admitting they are cutting costs to boost margins. The effect on the terminated worker is identical in either case. The broader pattern is that of companies reporting record profits while citing AI efficiency as justification for not rehiring. Amazon’s revenue was growing at 11 percent year over year during its largest announced layoff round. Google’s advertising business was stronger than ever. These are not companies that cannot afford workers. They are companies choosing not to grow headcount, and AI gives them a credible public narrative for doing so.
The most concrete and measurable harm is landing on entry-level workers right now. A Stanford Digital Economy Lab study published in August 2025, analyzing millions of ADP payroll records, found that employment for workers aged 22 to 25 in the most AI-exposed professions, including software engineering and customer service, experienced a 16 percent relative decline since late 2022, while employment in less AI-exposed fields like healthcare and maintenance held steady or grew. Employment for software developers aged 22 to 25 specifically declined nearly 20 percent from its late 2022 peak. Big Tech companies reduced new graduate hiring by 25 percent year over year in 2024. Computer science graduates now carry a higher unemployment rate than liberal arts graduates.
A sentiment now common among hiring managers is that where they previously needed ten engineers, they now need “two skilled engineers and one LLM-based agent.” Amr Awadallah, CEO of AI startup Vectara, said: “We don’t need the junior developers anymore. The AI now can code better than the average junior developer that comes out of the best schools out there.” This view is contestable and probably overstated as a universal claim, but its prevalence in hiring decisions is what matters. A hiring manager who believes AI replaces a junior developer, even if wrong, will not post the opening. The belief shapes the market regardless of its accuracy.
Whether AI ultimately benefits workers through higher productivity, new categories of work, or reduced drudgery remains genuinely open. I was among those who believed innovations like LLMs would generate a significant wave of new engineering roles for the people building these tools, but that has not come to fruition at my expected scale.
16) Stuck: The Surprising Lack of Exit Optionality for Software Engineers
One of the more persistent myths attached to a software engineering career is that the technical foundation opens doors. The implicit promise is that unlike, say, an accountant or a paralegal, a software engineer possesses a generalist problem-solving skill set that translates across industries, functions, and career pivots. This turns out to be substantially less true than advertised, and the gap between the myth and reality has widened considerably in recent years.
The most commonly cited exits — product management, data science, and venture capital — share a common feature: they are all experiencing the same structural pressures as software engineering itself,
Even other functions like go-to-market, operations, and sales are often out of reach for most engineers. Companies hiring for these roles are not looking for former engineers as a default. The skills do not transfer on a resume in legible ways, the cultural fit assumptions cut against engineers in most hiring processes, and the compensation reset required to make the transition.
Within engineering itself, the most visible internal pivot is into engineering management. For those who make the transition, it can be a strong path with higher pay and broader organizational influence. But it suffers from the same structural constraint as the rest of the industry: there are far fewer openings. Teams typically need many engineers but only a small number of managers, which means only a minority of engineers can realistically move into those roles.
The final exit most engineers consider is entrepreneurship — starting a company. The structural case for this has weakened. Venture funding in 2024 and 2025 has concentrated dramatically at the top, with early-stage deals increasingly difficult to close without a network connection to investors or a prior exit on the resume. The proliferation of AI coding tools has lowered the barrier to building software prototypes, which is a double-edged dynamic: it has made it easier to build, and simultaneously easier for every other engineer to build, reducing the competitive advantage of technical skill as a founding differentiator.
The more structural problem is what Anthropic has done to the destination. The canonical software engineer’s startup is a vertical SaaS product: a focused tool for legal teams, a financial analysis dashboard, an HR workflow platform, a sales automation layer. These are exactly the categories that Anthropic systematically targeted with Claude Cowork plugin releases in early 2026 — shipping prebuilt agents for legal review, financial modeling, investment banking, equity research, HR operations, and sales, essentially releasing a general-purpose AI workforce configurable to any domain. The market understood the implication immediately: the plugin announcements triggered a $285 billion selloff across software, financial services, and asset management stocks, with Thomson Reuters falling 16%, LexisNexis parent RELX down 14%, and the Goldman Sachs basket of US software stocks recording its largest single-day decline since April’s tariff selloff. What investors were pricing in was not just disruption to incumbents — it was the compression of the entire layer of vertical software that engineers have historically founded companies to build.
17) Final Word
I am no expert, and I encourage you once again to do your own deep research, ask people in the industry, search companies on LinkedIn, and track the news. To me, unfortunately, the path looks bright for a minority of talented computer science graduates, and not the majority. I don’t believe we are experiencing a cynical shift, but a structural change in the underlying economics of the profession. For most, it is an uncertain path, and at worst a painful one. But ultimately, your fate is in your own hands. Take everything I wrote with a grain of salt — these are the byproduct of my own lived experiences and are obviously biased. Make the decision that is right for you.