A 10x engineer with AI does not become 11x.
They become 100x.
Some companies are getting that math badly wrong right now.
The Multiplication Math
AI multiplies. It does not add.
If AI turns a baseline engineer into a 10x contributor, it turns your senior engineers into 100x contributors. The ratio stays the same. The absolute gap in output explodes.
Before AI, the difference between your best engineer and your average engineer was 9 units of work. After AI, it is 90.
And you are thinking about laying off the 100x people?
What 10x Actually Means
The 10x engineer was never just fast at coding. They were fast at everything around the code.
They have been in the industry long enough to know which trends matter and which ones are noise. They know which architectural decisions look fine today and cost you eighteen months of pain later. They navigate company politics well enough to get products shipped.
AI does not replace any of that. AI makes all of it count for more.
I Watched It Happen
I watched a company cut its senior engineers last year. The bet was that Claude Code and Cursor would close the gap. The remaining team shipped faster than ever. They also built an absolute mess.
The velocity was real. The progress was not.
Without senior engineers in the loop, nobody asked the hard questions. Nobody said "we have seen this pattern before, and it does not scale." Nobody pushed back on the architecture that looked clean in a pull request and would collapse under production load.
The Junior Dev with Superpowers Problem
AI behaves like a junior developer who will dutifully take on any task you give it. It often takes unusual approaches when what you need is clarity, observability, efficiency, and security.
One engineer described watching a junior developer build a small web app with Claude. When it was done, every form submission fired twice. The AI insisted the user was pressing the button twice. It chased its own tail for two hours. The engineer spent two minutes troubleshooting and found the cause: the AI had wired up two event listeners for the same button.
AI is still bad at troubleshooting the problems it creates. That is not a bug that gets patched next quarter. It is a gap that only experienced engineers can fill.
The Real ROI Is Nonlinear
Here is the part most engineering leaders miss. The real return on AI shows up for engineers who work in parallel. It is shocking how many engineers write a prompt and then sit there waiting for it to finish.
A 10x engineer should be running five to ten AI agents at any given time. It gets better when those agents delegate to more specialized agents. I wrote about what that looks like day to day in Claude Remote Agents: Running 10 AI Agents While You're Not at Your Desk.
The gain is not faster typing. It is orchestrating parallel streams of work that a less experienced engineer would not know how to decompose.
The senior engineer's superpower was never speed. It was knowing how to break a problem into the right pieces. AI lets them work on ten of those pieces at once.
Who Reviews the 100x Output?
One counterargument deserves attention. Who reviews the output of a 100x engineer? There is a real tension between velocity and understanding.
Remember the old triangle: good, fast, cheap. Pick two. The worry is that AI pushes you toward fast and cheap with the illusion of good. You get code that compiles and passes tests but lacks the guardrails only experience provides.
That is exactly why you need senior engineers in the loop. They do not just produce 100x output. They are the ones who can judge whether that output should exist at all.
Remove them and you do not just lose a coder. You lose the only person who could tell you the code was wrong. Russinovich and Hanselman made a version of this argument in the ACM, and I broke it down in Russinovich and Hanselman Just Redefined "Senior Engineer".
The Institutional Knowledge You Cannot Rebuild
Something costs more than rehiring a 10x engineer: losing the years of institutional knowledge they carry.
Claude can boost output. It knows nothing about your product, your domain, or the decisions your team made three years ago and why.
Every senior engineer on your team is a walking decision log. They remember the migration that failed, the vendor that oversold, and the architecture that almost sank a launch. That context does not live in a wiki. It lives in their heads.
When they walk out the door, it walks out with them.
They Are the Multiplier
The engineers who survived the last three evolutions of the web know when to say no. They know when to push back. They recruit teams, build culture, and drive outcomes, not just output.
The smartest companies do not treat their 10x engineers as individual contributors who happen to type faster. They treat them as the future managers of AI agents. These engineers orchestrate teams of automated workers and pass decades of judgment to systems that can finally apply it at scale.
The engineers who see that will hit 100x. Maybe 1,000x.
Those are the people AI amplifies most. Removing them does not save money. It removes the multiplier.
The Counterpoint Worth Hearing
Not everyone agrees with the 100x math, and it is worth being honest about that.
Some argue that AI compresses the gap instead of widening it. When every task already takes an hour with AI, the difference between your best and average engineer shrinks.
Others point out that engineers who built a 10x reputation on raw code generation, cranking out DAOs, DTOs, and CRUD scaffolding, might actually drop from 10x to 5x. AI commoditized their exact superpower.
Both are fair points. Both prove my point. The engineers who matter most were never the ones who were fast at the mechanical parts. They are the ones whose judgment, architectural instincts, and leadership AI cannot compress.
Do the Right Math
Before you cut headcount, do the math.
Not the salary math. The multiplication math.
While you are at it, flip the question. Instead of doing the same work with fewer people, ask why you are not doing ten times the work with the same people. Serve ten customers instead of one. Ship ten products instead of one.
The company that keeps its 100x engineers while competitors lay theirs off is the company that wins the next five years.