This isn't an anti-AI piece. We use AI coding tools every day - Claude Code, Cursor, Copilot. They're genuinely transformative. But there's a growing narrative that AI "levels the playing field" between junior and senior engineers, and the data tells a very different story.
AI is a multiplier. And multipliers widen gaps - they don't close them.
Seniors are measurably faster with AI
A January 2026 study published in Science tracked 160,000 developers across 30 million Python contributions on GitHub. The finding was stark: AI productivity gains are "driven exclusively by experienced users" - beginners "hardly benefit at all." Less experienced developers actually use AI in 37% of their code versus 27% for veterans, yet only the veterans see real output gains. AI is a skill amplifier, not a skill equalizer.
The Fastly developer survey confirms the pattern: 32% of seniors ship over half their code as AI-generated, versus just 13% of juniors - a 2.5x gap. A senior in the survey: "AI will bench test code and find errors much faster than a human." A junior: "It's always hard when AI assumes what I'm doing." Same tool. Completely different experience. The senior knows what "correct" looks like. The junior doesn't yet have the mental model to steer it.
The 70% problem
Addy Osmani at Google gave this a name: the 70% problem. AI gets you 70% of the way to a working solution, fast. The remaining 30% - error handling, edge cases, security, production readiness - is where products succeed or fail. As Osmani put it: "AI is like having a very eager junior developer. They can write code quickly, but they need constant supervision."
As ISHIR's 2026 analysis puts it: "AI amplifies whatever judgment it is given. With weak judgment, it produces fast, confident chaos; with strong judgment, it becomes a force multiplier." The real bottleneck has shifted from execution to judgment. And Andrew Ng: "Vibe coding might sound chill, but it's misleading... I'm frankly exhausted by the end of the day." Guiding AI is a deeply intellectual exercise. It only looks easy when the person doing it has decades of context.
The last 30% is where products succeed or fail. AI can't close that gap - only experience can.
What juniors miss
Anthropic's skill formation study tracked 52 junior engineers and found those using AI coding tools scored 17% lower on comprehension tests. The largest gap was on debugging - exactly the skill you need when AI gives you something that looks right but isn't. Developers who delegated code generation scored below 40%. Those who used AI for conceptual questions ("why does this pattern work?") scored above 65%.
The downstream effects confirm this. Veracode's Spring 2026 update found AI-generated code still fails security checks 45% of the time, stagnant despite testing the latest models - syntax looks clean, but the security gaps persist. And METR's developer productivity study found AI made experienced developers 19% slower on their own repos, but they believed it had sped them up by 20%. The tools feel productive even when they aren't.
The security gap
Apiiro found developers using AI produced 3-4x more code but introduced 10x the vulnerabilities. Privilege escalation paths up 322%. CodeRabbit: security issues 2.74x higher in AI-generated code. The code compiles, the tests pass, and there's an injection vulnerability a junior doesn't know to look for.
That's the core asymmetry. AI generates code that looks correct - proper variable names, clean patterns, passes linting. A senior spots the missing input validation because they've been burned by it before. A junior doesn't know what they don't know.
You can't catch what you don't know to look for.
AI is a multiplier, not a replacement
Option A: A junior with AI tools. Fast code generation, but needs senior review on every PR, misses edge cases, accepts suggestions that create security debt. Timeline: 3 months plus remediation. Option B: A senior with AI tools. Uses AI for boilerplate, tests, and migrations while applying judgment on architecture, security, and data modeling. Timeline: 6 weeks. Similar cost. Dramatically better outcome.
None of this means juniors shouldn't use AI - they absolutely should. But the way they use it matters. Using AI to understand why something works builds skill. Using AI to skip understanding builds dependency.
AI is the best tool we've ever had. But a tool is only as good as the hand that holds it.
We build this way.
AI tools make it more important to have senior engineers, not less. We've built every VectorLabs project with this model: senior engineers using AI as a power tool, not a crutch. If you're building a product and want a team where AI amplifies expertise instead of papering over its absence - we should talk.