A common framing of AI coding tools: they make great engineers greater. Boost the best to new heights. Multiply productivity for top performers.
The pattern I’ve actually observed in working with engineers across many teams: the skill floor moved up substantially. The ceiling barely moved. The effect is in the bottom and middle of the distribution, not the top.
This has implications for hiring, mentoring, and career strategy.
What “skill floor” means
The skill floor is the threshold below which an engineer isn’t productive. Engineers below the floor produce more problems than they solve. Engineers above the floor add value.
Pre-AI, the floor was high. You needed to:
- Understand programming concepts deeply
- Be able to write code from blank screen
- Debug systems independently
- Read documentation and adapt
The floor excluded a lot of would-be engineers. They couldn’t reach productivity in reasonable time.
Post-AI, the floor moved down. Engineers who couldn’t reach productivity pre-AI can now:
- Generate working code through AI assistance
- Get explanations of unfamiliar concepts
- Debug with AI-suggested probes
- Adapt patterns the AI shows them
The bar to entry is lower. More people are above it.
What “skill ceiling” means
The skill ceiling is the maximum productivity an engineer can reach. Top performers operate near it.
Pre-AI, top performers had:
- Deep technical knowledge
- Strong judgment about design and tradeoffs
- High typing and editing speed
- Excellent debugging instincts
- Architectural insight from years of experience
These engineers shipped 5-10x more than median engineers, by various measures.
Post-AI, top performers still have all of those. They also use AI tools effectively. But the AI’s productivity gain is bounded — maybe 20-30% improvement on the work they’re already good at.
The ceiling moved up modestly. Top performers are slightly better. The 5-10x ratio between top and median didn’t change much. If anything, it shrunk because median engineers got more help than top engineers did.
Why this asymmetry
Several reasons:
Top engineers don’t need help on most tasks. They already know the answer. AI tools save typing; they don’t help with the thinking.
Top engineers do harder tasks. The hardest engineering work — system design, debugging novel issues, architectural decisions — is where AI is least helpful.
Top engineers are constrained by judgment, not implementation. Writing the code is fast; deciding what code to write is slow. AI helps with the fast part, not the slow part.
Median engineers benefit more from AI’s defaults. AI’s “reasonable but generic” defaults match what median engineers want. Top engineers want non-default solutions; AI’s defaults aren’t a fit.
The mathematical structure: AI tools have an upper bound on productivity gain. Engineers who were near that bound pre-AI gain less from AI than engineers who were further from it.
What this means for hiring
If the floor moved up, hiring quality from the same talent pool should improve. Median candidates are more productive than they used to be.
If the ceiling barely moved, the top of the funnel is similar. The very best engineers are about as productive as before.
The implication: the gap between “good hire” and “great hire” narrowed. The economic value of identifying truly great engineers may have decreased; the cost of hiring competent engineers may have decreased more.
For some companies, this means hiring more junior engineers and relying on AI to help them ramp. For others, it means hiring fewer engineers because each is more productive.
What this means for engineers
For engineers in the bottom half of the distribution (no offense; we’re all in the bottom half of some distribution):
The AI tools are good news. They make work more accessible. The skill of “use AI tools effectively” is learnable; the gains are real; the productivity boost is meaningful.
For engineers in the top quartile:
The AI tools help, but they’re not transformative. The 20-30% gain on routine work is welcome. The harder parts of your job — judgment, design, debugging — are less affected.
The honest truth: AI tools don’t make a top-quartile engineer dramatically more competitive. They reduce some grunt work. They don’t change the underlying judgment-based work that distinguishes top performers.
The career implication
For young engineers building careers:
The path “become competent with AI tools and ride them to productivity” is real. It works. But it has a ceiling. The path levels out at “median engineer enhanced with AI.”
The path “develop deep judgment and engineering intuition” is harder and slower. It’s also where the long-term value is. Engineers who go deep on fundamentals plus AI tools end up more productive than engineers who go wide on AI tools alone.
Both paths produce employed engineers. Only one produces top performers.
What I tell engineers
When asked about career strategy:
“Use AI tools well. The bar to do useful work is lower than it used to be. But don’t mistake AI fluency for engineering excellence. The deep skills that distinguish great engineers — system design, debugging, judgment, architectural insight — still require years of deliberate practice. AI doesn’t shortcut that.”
“The engineers who thrive in 2027 will be both AI-fluent and deeply skilled. AI fluency alone is increasingly common; deep skill is increasingly rare. The combination is where the value is.”
A counter-perspective
A reasonable counter: “the ceiling will keep moving as AI improves.” This is true. The ceiling in 2027 may be higher than in 2026.
But the same dynamics apply. The ceiling moves; the floor moves more. Top engineers don’t get the same proportional gain.
Whether this is good or bad depends on perspective. For the profession overall, more competent engineers is good. For top engineers specifically, the relative advantage shrinks.
What stays valuable
A few skills remain firmly valuable in any AI-assisted future:
System design. AI doesn’t design well-architected systems.
Debugging. Investigation, hypothesis-forming, root cause analysis. Mostly human.
Communication. With teammates, customers, stakeholders. AI doesn’t replace.
Domain expertise. Knowing the business, the user, the data. AI is generic; domains are specific.
Tradeoff judgment. Choosing among options requires understanding consequences. Human work.
Engineers who invest in these are investing in skills that compound. The investment in “be really good at AI tooling” has a lower ceiling.
Closing
The narrative around AI tools often emphasizes “everyone gets superhuman.” The reality is “the bar got lower.” Both are real. The difference is in the distribution: more people clear the bar; fewer rocketship past it.
For most engineers, this is welcome. The work is more accessible; the productivity is higher. For engineers aiming for the top, the implication is that AI is helpful but not the primary edge. The primary edge is still the deep skills that come from years of careful work.
Plan careers accordingly. AI tools are part of the toolkit; they’re not the foundation. The foundation is engineering judgment, which is built slowly through experience and reflection.