The 10x engineer is the most expensive slide in enterprise AI. Vendors keep selling it; my data has yet to show it. That gap isn't a rollout failure — it's a measurement error.
The slide says 10x. Production says 10%.
DX's longitudinal study tracked a sample of 400 companies from November 2024 through February 2026. AI usage rose 65%. Pull-request throughput rose 7.76% — after filtering out teams gaming PR counts. As one senior developer put it: "A four-day task might take three. But that doesn't mean I'm shipping 3x more PRs."
Most engineering leaders I talk to land in the 5–15% range. That's a real gain. It is not 2–3x, and it's nowhere near 10x.
Where the 10x number comes from
CMU researchers compared 806 GitHub projects that adopted Cursor against a matched control group. In month one, lines of code added jumped 281%. That's the screenshot on the vendor's slide.
By month two it collapsed to 48% — because the extra code arrived with logic errors and security holes that made the codebase harder to maintain. The speed didn't stick; the rework did.
Writing code was never the bottleneck. Writing code you don't have to rewrite is.
Speed is the wrong metric
"10x" mistakes typing speed for shipping. Code is a small slice of the SDLC. Planning, review, handoffs, and debugging AI-generated logic eat the rest. In a METR controlled trial, experienced open-source developers were 19% slower with AI tools — while convinced they'd been faster.
The takeaway
Stop measuring lines of code and PRs per head. Measure time-to-fix and rework rate after AI adoption. If AI had actually made your teams 10x, you'd have already cut engineering in half. You haven't. The data says you shouldn't.