I sat in on a debrief last Tuesday for a candidate we’ll call Alex. Two years ago, Alex would have been an immediate "yes." He had three years of React experience, knew the core hooks inside out, and could build a generic CRUD interface with his eyes closed. But as we reviewed his take-home and system design responses, the room was awkwardly quiet. Our lead engineer finally broke the silence: "Everything he built is exactly what Copilot generates on the first pass. He didn't catch the edge cases, and he didn't design for scale. We don't need someone to just write the boilerplate anymore." We passed on Alex. He’s not a bad engineer—he’s just a completely average one. And that’s the problem.
Being a competent, average engineer used to be enough to build a solid career. It no longer is.
The popular narrative is that AI is indiscriminately eating software jobs, shrinking the entire market. But that's fundamentally untrue. The market didn't shrink uniformly—it split. Senior and specialized engineering hiring remains highly stable, while entry-level and generalist hiring has collapsed.
The data paints a clear picture. Engineering hiring at the largest tech companies is down only about 11% from 2019, while total company hiring fell about 25%. Engineers are actually taking a bigger share of hires than before—making up 55% of new hires in 2025, up from 46% in 2019, according to SignalFire's 2026 State of Talent report.
Meanwhile, entry-level hiring at big tech is down roughly 65% from 2019, and a staggering 76% at early-stage startups. Average isn't disappearing everywhere. It's disappearing at the bottom of the ladder, and the ladder itself just got significantly shorter.
Don't just take my word for it. Look at the numbers:
| Era | What Got You Hired | What Didn't Matter Much | | :--- | :--- | :--- | | ~2015 | CS fundamentals, data structures/algorithms, "can you code" | Deployment experience, system design (rarely asked below senior) | | 2020–2022 (COVID Boom) | Warm body + can pass a LeetCode round. Remote unlocked huge pools. | Depth. Many were hired fast and shallow because headcount mattered more. | | 2026 | System design even at mid-level, AI-integration skills, cloud infrastructure, ability to direct AI tools and catch their mistakes, shipped projects. | Pure LeetCode grinding; generalist frontend/CRUD skills with nothing distinguishing them. |
Engineers with LLM integration or cloud infrastructure experience report 3–5x higher callback rates than generalist applicants in Q1 2026. Furthermore, workers with multiple AI skills earn a salary premium estimated near 43% above non-AI counterparts.
Boilerplate backend work—CRUD APIs, admin tooling, basic data transformation—is increasingly being handled by AI. Crucially, this is specifically what used to be junior training work. This explains why the junior pipeline narrowed so aggressively.
QA provides the clearest case study: engineering teams that used to run three to five QA engineers are now operating with just one human overseeing AI-generated test coverage. The baseline of what constitutes a "hirable skill" has fundamentally shifted up the stack.
Junior engineers were traditionally trained on small, well-scoped, low-ambiguity tickets. Unfortunately, this is exactly the category of work that AI tools now do reasonably well. The training ground got automated before the humans who needed that training ground got hired. This isn't a matter of "AI is smarter than juniors"—it's simply that the specific low-ambiguity tasks juniors cut their teeth on are now the cheapest thing to automate.
However, not every company is blindly following this playbook. IBM, for example, tripled its entry-level hiring in 2026 on the logic that AI still needs extensive human oversight. But for the vast majority of the industry, the economics of automating low-level ticket work are too compelling to ignore.
If you're reading this and feeling the pressure, here is how you need to adapt:
My honest read of the data:
The people most panicked about AI replacing engineers are, disproportionately, the ones who were coasting on "I can write working code" as their entire value proposition. That was always a fragile foundation—it just took this long for something to expose it. The data backs this up directly: senior and specialized hiring is stable or growing, while entry-level generalist hiring has collapsed. That's not AI replacing engineering. That's AI removing the floor that let mediocre engineering coast.
The genuinely uncomfortable part for the industry is that the traditional junior-to-senior pipeline is structurally broken right now. The entry-level work that used to train future senior engineers is the exact work getting automated first. That's a profound problem nobody has really solved yet. It's not just a matter of telling juniors to "get good"—it's asking how the next generation of senior engineers even gets made if the on-ramp is gone.
Two years ago, Alex would have easily landed that mid-level role. Today, his inability to see past the boilerplate cost him the job. The engineering landscape hasn't necessarily shrunk, but the requirements for participation have irreversibly changed. The days of getting by on baseline competence are behind us.
The market didn't get harder to survive in, it got harder to hide in.
(Sources: Indeed Hiring Lab, SignalFire State of Talent Report 2026, Ravio 2025/2026 Tech Job Market Report, Layoffs.fyi, Stanford HAI 2026 AI Index Report)