If AI brings 90% productivity gains, do you fire devs or build better products?
TL;DR
A developer reports massive productivity gains from AI coding tools, especially for boilerplate, libraries, build tools, and refactoring.
Key Points
- Starting out skeptical, the hands-on experience was convincing: Java projects with Checkstyle and Maven Verify run almost automatically, no manual linter battles needed.
- The workflow shifts: keep IDE open, feed back 'analyze code' output, grab a coffee – instead of spending hours dismissing warnings.
- The thesis: what remains is only logic and real ideas. Whether that holds for complex legacy enterprise systems is still an open question.
- The real question behind this is strategic: if 90% of routine work disappears, does the team shrink – or do better products get built?
Nauti's Take
The 90-percent figure is obviously marketing – but it's getting hard to argue that something fundamental hasn't shifted. What's interesting is the psychological arc in the report: first checking every diff with suspicion, then realizing it mostly works, then letting go.
That's not a tool update, that's a new mental model. The real risk isn't the hype – it's companies giving the wrong strategic answer.
Anyone who reflexively cuts headcount now instead of raising product ambitions will find themselves outpaced in two years by competitors who did the opposite. Legacy enterprise systems remain the grey zone, but for anything greenfield, the bar just dropped dramatically.
Sources