Two data points that should be in every CTO’s briefing packet right now:
- Computer Science enrollment dropped 8.1% in 2025-2026, the steepest decline of any field. CS-specific enrollment plummeted 11.2%.
- Hiring of new graduates at the 15 largest tech companies has fallen 55% since 2019.
Students are leaving the field because they believe AI will eliminate programming jobs. Companies are cutting junior roles because they believe AI can fill the gap. Both groups are making the same mistake: confusing code output with engineering capability.
The pipeline problem
Here’s how software engineering actually works: junior developers become mid-level developers become senior developers become architects and technical leaders. It takes roughly 7-10 years to develop the kind of judgment that lets someone design systems that scale, fail gracefully, and evolve over time.
When companies stop hiring juniors, they’re not saving money. They’re borrowing against a future where there aren’t enough senior engineers to go around. One industry observer called it “sawing off the branch they’re sitting on,” predicting a senior developer shortage in 5-7 years that will be “genuinely catastrophic.”
I watched a version of this play out at RF-SMART. When I took over as Director of R&D, I inherited a team where institutional knowledge was concentrated in a handful of people. We fixed it by investing in mentorship, agile practices, and test-driven development, building a pipeline of engineers who could eventually replace us. It wasn’t glamorous work, but when I left nine years later as VP of Products, the team was deep and resilient. That doesn’t happen by accident.
The AI augmentation misconception
The counterargument is obvious: “AI will do the junior work, so we don’t need junior developers.” This misunderstands what junior developers actually learn during those early years.
They’re not just learning to write code. They’re learning:
- How to read a codebase they didn’t write and figure out how it actually works (not how the docs say it works)
- What happens when your assumptions about a system are wrong and production goes down at 2am
- How to communicate technical constraints to non-technical stakeholders without either lying or lecturing
- The judgment to know when “good enough” is actually good enough, and when it isn’t
AI can generate code. It cannot generate the scar tissue that comes from shipping something broken, getting paged, fixing it under pressure, and doing the post-mortem the next morning. That scar tissue is what makes a senior engineer.
What this means for companies right now
If you’re a startup or growing company, this trend is actually an opportunity:
Hire juniors now while your competitors aren’t. The developers entering the field today, the ones who chose CS despite the AI narrative, are self-selected for genuine interest in the craft. They’re also learning AI tools natively. In five years, they’ll be the mid-level engineers that everyone is desperate for.
Invest in your existing team’s AI fluency. The data is clear: developers who are fluent with AI tools are commanding 50%+ wage premiums and getting promoted faster. Training your current team is cheaper than replacing them later.
Consider fractional technical leadership. Not every company needs a full-time CTO, but every company building software needs someone with the experience to make architectural decisions that will hold up over time. That’s a big part of what we do at Londonberry: bringing senior-level judgment to teams that aren’t ready for a full-time executive hire.
The Shopify signal
Shopify CEO Tobi Lutke’s leaked memo declaring “reflexive AI usage is now a baseline expectation” got a lot of attention. But the most important detail was overlooked: he didn’t fire his engineers and replace them with AI. He told existing engineers to use AI as a force multiplier. Managers requesting new headcount must first prove AI cannot do the job.
That’s the right frame. AI is a tool that makes experienced people more productive. It’s not a replacement for the experience itself.
The companies that understand this distinction will be the ones with functioning engineering teams in 2031. The rest will be fighting over the same small pool of senior developers who spent their early careers learning the hard way, because there’s no shortcut.