The tech job market is going through one of the largest structural changes in history. For the youngest generation, employment is 20% below average and falling. But for the senior counterparts in the exact same roles, it's growing. Entry-level jobs are vanishing and the applications to the grad schools are exploding. Anthropic just published new labor market research showing that the occupations most exposed to AI are not low skill or entry-level service jobs. As most people thought, it's programmers, data analysts, financial analysts. In other words, often highly educated and highly paid knowledge workers in the economy. By the end of this video, you will understand exactly what is happening to the tech labor market for juniors, for seniors, and for everyone in between. Why this shift is happening, and whether AI is the only thing to blame. Let's dive in. Anthropic published their new research called labor market impacts of AI, a new measure and early evidence. Its main premise shifts the narrative around what models could potentially do to what models are actually being used to do in jobs. They use the term observed exposure. Now, what exactly does it mean? Observed exposure takes the theoretical potential, like what could an LLM theoretically do and smashes it together with what's actually happening in the real world. What are people actually using AI for when it comes to their jobs? Anthropic built a whole framework around this idea to understand how AI is going to hit the labor market. Anthropic's economic index shows that the highest productivity gains in AI, meaning the areas that are significantly enhanced by AI, are associated with tasks that require high human capital. But human capital, they're referring to the people with college level education or above. This is an inversion of earlier predictions that AI would first automate simple tasks. The most exposed occupations under this framework are software developers, data analysts and financial analysts. All of these are predominantly higher wage jobs and for almost all of them at least an undergraduate degree is required. So anthropics productivity data on the occupational level shows that the effect of AI is strongest at higher skill trades. But the labor market data shows that instead of getting rid of these roles, companies are not opening junior roles in those same occupations.

Junior market is a very hot topic right now. But we will talk about all three levels. Juniors, mid-level, and seniors. Starting with the blood bath, juniors. Every major labor market data set is pointing in the same direction when it comes to entry- level work. US entry- level job postings are down 35% since 2023. 50% drop in new job opportunities for people with less than one year of post-graduate experience across major public tech companies. 13% decline in entry- level hiring specifically for AI jobs. There's also a repeating pattern across data sets around the junior hiring freeze. Companies are not firing or laying off juniors. They're simply not hiring them. the logic from the employer's perspective that a junior costs between $60 to $90,000. I'm talking about the tech industry here. Plus benefits plus onboarding plus management plus the ramp up period. It's not cheap. If AI can handle basic research, first drafts, data entry, data cleaning, basic financial modeling, code debugging, why hire a junior? So naturally, companies that have large teams in the exposed occupations are substituting AI costs for junior labor costs. The junior hiring freeze is the most economically significant finding in the anthropic paper. And from purely capital allocation lens, it makes sense. But the problem with this logic is that the entry-level roles are not just the cheapest labor in the building. Juniors are the pipeline. the pipeline that you build for the next five to 10 years from now, the pipeline to which you transfer knowledge, where you find people for your next leadership roles, and where you find your next elite individual contributors. Sure, a 22-year-old kid starting a career in finance and doing basic ledger reconciliation is not providing a ton of value, but they are providing massive indirect value in six or seven or eight years from now. Because when they manage team they have to understand how the firm actually works from the inside. Now is this hiring freeze happening because of AI? AI is a contributing factor but only about 45%. The mass unemployment scenario has not materialized yet. But we are witnessing a major change in the org structure where five seniors with Claude replace what was previously done with seven people including two to three juniors. And yes, the tasks the juniors performed were often quite basic, often administrative in nature. But those tasks are essential for one to understand the implications of their work and of their mistakes. If you consider the consequences of this shift outside of the capital allocation and ignore the P&L statement, this is a talent crisis in the making. The cognitive development in humans requires constant and deliberate practice through real challenges. and entry-level positions is the practice environment where these challenges naturally occur and where young people develop professional judgment. And on one hand, you may say, "Well, we're not going to be doing things by hand. Nobody's going to be writing the ledger. Nobody's going to be writing code." Basically, all jobs will become supervising AI. Sure, except supervising AI requires a rockolid critical thinking and judgment on a system level. And the very people who will be supervising AI aren't developing the judgment. Let me give you an example from my own career. My first job in corporate tech was actually in product marketing at a pharmacy technology company. I was 19 years old at the time and one day we were launching a software update for drug distribution machines that package medications for patients in hospices. and I had to connect with staff at various hospices to send them the promotional materials and schedule training. So, one more time, the company's rolling out an update that will affect how medication gets dispensed into pouches for patients in hospices. This is a very big deal from the operational point of view. An update like this takes many months of preparation with on-site visits, lots of training materials for the hospice staff, for technicians, for managers who use the software that controls the machines that dispense the medication. So what happened was my boss asked me to send an email to hospital staff just to start planting the seed in their heads that we're working on something that this something will be available in 6 to 8 months from now and that we're slowly preparing them for a fact that this update it's something they should be looking forward to and something that they would want to invest in. So on one hand I was asked to draft an email, a very dumb, a very basic task that a kid could do if you explain well enough what that email needs to say. But my 19-year-old brain did not have the sense of scale at that time. What I heard from my boss is that he wants me to tell all hospices that we will be changing how the medications are going to be packaged and I am the messenger who delivers the news. Did I think about the potential implications of my email? No. Did I bother asking my boss a clarifying question, for example, like are they aware of this change or am I the first person telling them about it? No. Did it occur to me that delivering news of that kind or that scale is simply outside of my area of responsibility and influence and that there are much more important people in the building who would typically deliver those kind of news? No. So what did I do? I sent an email to about 400 hospices across Canada with a cheerful subject line. Hello, this is Daria from marketing and by the way, we're changing something your entire institution depends on. Have a nice day. Was I bombarded with calls 10 minutes after I sent that email? Yes. Did my boss get the same calls? Yes. Did support, implementation, project management, finance, accounting get the same calls? Yes. The whole company knew my name. But I learned my lesson. And that lesson was triggered by one email. The point I'm trying to make is that it's not the email that defines the job of a junior. If something this routine can be automated or drafted in seconds rather than making someone look for the right adjective, by all means, take it, use it, polish it, do whatever you want with it. But that email encapsulates the need to teach junior people what consequences mean. That email forces a junior to develop a high order thinking. A high order thinking that is the very skill that this labor market demands. If we want to grow good seniors, we have to remember that a good senior is not a product of a function of age. A good senior is an accumulation of thousands of solved problems, bug fixes, and near disaster experiences. If the current generation of juniors never grapples with those low-level problems because AI solves them automatically, they may never develop that intuition or tacit knowledge required for senior roles. It's almost like ending up with a generation of doctors who wouldn't be able to insert an IV. There are at least two direct consequences of this shift that I personally see and want to bring up. The first one being Gen Z may choose to go for longer education paths and it's not because they truly want those paths but because the sheer number of years at school lets them escape the job market situation. For example, law schools in the US are seeing the highest number of admissions since 2010. There is the documented surge in grad school applications. MBA applications went up 8%. Medical school enrollment cross 100,000 students for the first time ever in the United States in 2025. When the job market is tough, juniors head to grad schools. And the second consequence, rapid and drastic career switches. There is another interesting finding that I came across when preparing for the video. Stanford published a paper in August last year called Canaries in the Coal Mine. They used payroll data across millions of workers at thousands of US private companies and found that in jobs with the highest AI exposure, such as software development and data analytics, the employment for workers aged 22 to 25 fell 6%. While employment for workers aged 30 and older rose 13% in the same category. For the youngest software developers, employment in July 2025 was 20% below the peak. But for workers aged 35 and older in the same category, it grew. Which brings me to my next point. What's the deal with the seniors?

The mid-level and senior market is also weird. Anyone who has touched AI generated code will tell you that it is flawed in many ways and it does need thorough debugging and testing. So the assumption that we're running with is that AI would compensate for the mentoring hours and that even if we don't have the juniors and the work falls on seniors, they will still be around three to four times faster. The problem is AI handles speed sensitive tasks but things like design or stakeholder management or quality control. A lot of these things are parts of these things can also be delegated to juniors. But now that the junior is not there, this becomes the responsibility of a mid-level or a senior. So the consequence of the hiring freeze for juniors is seniors doing more work. Good engineers are staying put because the market is very unstable. And if you're a senior, chances are you're in your late 20s to mid-30s to late 30s, which makes it likely that you have family responsibilities or a mortgage or two dogs. And therefore, in this economy and in this market, you choose stability over adventures. So, solid mid-level and senior folks don't move around too much. They're stagnant. At the same time, juniors are suffering and bombard recruiters with lots of empty and meaningless applications. And this crisis gives the rise to career coaching. The career coaching that costs $3 to $5,000 a course that teach people how to reach out on LinkedIn efficiently so that you get responses. Except your LinkedIn invitation is just as efficient as a million other efficient invitations. And that is where the loop closes. The seniors are not moving. The mid level is not moving that much. And the junior level can't get in. The first two are sitting in their positions because the riskreward of moving has flipped. The market is unstable and a stable job is worth a lot more than chasing a marginal comp improvement. So the only way that hiring managers get access to senior or highquality candidates is through personal networks. And personal networks take years to build. So this brings us back to the cycle where juniors are at a disadvantage again. AI makes senior individual contributors more used. At the same time, tech is simultaneously desperate for senior engineers and eliminating senior managers. But senior individual contributors are running out of inventory because they are blocking the upward mobility. On top of it, consider the broader aspect. The boomer exit and the junior pipeline. The pipeline that would normally produce the next cohort of seniors is running thin. In a normal labor market, the boomer exit would create upward mobility because the seniors would retire, the mid-levels promote, and juniors are hired to fill the entry seats. But seniors are not retiring because economic anxiety extends careers and the junior slots beneath them are disappearing anyway. The senior market is tricky because on one hand, it's significantly safer than the junior market. The demand for truly experienced engineers, architects, data scientists, product managers hasn't collapsed the way entry- level demand has. On the other hand, there is a growing commentary from hiring managers about how difficult it has become to find a genuinely solid senior. But this paradox makes sense when you understand what's happening underneath. The same AI-driven efficiency gains that are eliminating junior roles are also raising the bar for what senior means. Because now a senior engineer or a senior PM is expected to manage AI outputs, create and work with systems that integrate AI tooling and do the work that used to require a team of two or three. So the pool of people who actually qualify as senior by 2026 standards is also smaller than it looks on paper.

When media contemplates the SAS apocalypse, it's tempting to blame everything entirely on AI. But the reality is a lot more complicated. After co the growth of most large-scale SAS businesses became plateauing and then declining because the private equity money started to decline as well. This matters because private equity is the valve so to say that releases capital back to venture funds who then reinvest in the next generations of startups. And when it comes to private equity and software as a service market, roughly 50% of all private equity investments were flowing into SAS and it was profitable. SAS subscription revenues were sticky and predictable. So investors were willing to bet on them. For enterprise software, you can pretty safely raise prices because enterprise customers can't easily switch between vendors. And the math works beautifully for you as an investor when you invest in software as a service market. And this math works as long as the revenue keeps growing. But this condition doesn't hold anymore. On top of that, for big tech tier one and tier 2 SAS, there is another factor that made them very attractive for years, but has become a structural burden, and it's the stockbased compensation. Companies often paid 10 to 20% of total compensation in equity, sometimes more. And it was a brilliant tool during hyperrowth because it let companies attract talent without burning cash. and the promise of a big equity event like an IPO or an acquisition made the employees accept the salaries. But stock compensation is only cheap when your stock is appreciating. When multiples compress, as they have, stock grants that a company owes to the employees don't disappear from the income statement just because the stock price fell. For established SAS companies in a low growth environment, equity comp is a heavy and structurally difficult burden on their financials. So what does this mean for mid-level and seniors? Will there be layoffs? Yes, most likely. But the situation for seniors is fundamentally different from the junior market. There will be layoffs going forward, but they won't primarily be the AI layoffs. Seniors will be affected by the financial reasons rather than technological. Fewer exits for founders means fewer liquidity events for venture capital. Fewer liquidity events for venture capital means that it has less capital to redeploy into new bets, which also means that fewer new SAS companies will get funded and scaled. The job market shrinks not because AI fired everyone, but because the capital formation engine that built the SAS industry over the years is running at a fraction of its previous speed.

Now, every time I talk about the job market analysis, I get blamed for not offering solutions. I am no career expert, but here is what I would do and here is what I think works in this market and this economy. But take it with a grain of salt. The definition of career entry is rapidly changing and I almost feel like this is one of the most under reportported shifts because even Altman talks about how OpenAI doesn't know what to do with this whole shift. Entry level today means something different. You got to have the experience. You got to be proficient in AI. You got to have a portfolio. You have to have freelance and gig experience before you start working. There are active debates about the suitability of university education to the requirements of tech jobs. Whether you agree if the university education is important or not, some companies already began launching programs for juniors that are essentially a new way to do education. Here's one year palunteer. They literally wrote in the description, "You've told us you can't sit in lectures while the biggest story of our lifetime is written around you. It's clearer every day and is being written faster than ever. Palunteer calls this role forward deploy software engineer, but in simple words, they combine three jobs in one. A software engineer, a consultant, and an architect of sorts. This is a very progressive program, and yes, a program like this is rare and specific to enterprises that can afford it, but I'm pretty sure we will be seeing more of these going forward. Now, a few words on portfolios. Building a personal portfolio is a fantastic tool if you're trying to bypass the traditional application process entirely. This is something that I would recommend to absolutely everyone. Is there a guarantee that it'll help? Of course, there isn't. But I personally have been complimented multiple times on my portfolio simply because it exists. Portfolio really helps people stand out from the crowd because it very quickly tells anyone who's interested in you who you are and what you've done. I've personally had people come up to me during network events asking very specific questions about my experience and offering to interview for a role at their company because they're looking for someone with my skill set. So, here's the takeaway for me. AI didn't cause a mass unemployment event and it probably won't, but it did accelerate a shift that was already underway since co and that is having a solid proof of competence. The barrier to entry is much higher. The junior pipeline is thinning, seniors are stretched, and the middle is absorbing pressure from both sides. Whether you're a junior, a mid-level, or a senior, the worst thing you can do right now is wait for the market to come back to normal because it won't. As always, we hope this was helpful. Till next time. Bye.