Last year, a 90-minute coding interview meant choosing carefully which bugs to fix and which features to skip. Now, with Claude Code running alongside, the problem is different: you run out of ideas before you run out of time.
That is the core argument from developer Jason Lernerman in a recent essay about how Claude Code has changed his relationship with programming. His central claim is specific and testable: the bottleneck in software development has moved from execution to ideation. Writing code is no longer the slow part. Knowing what to build is.
The Job Interview That Proved the Point
Lernerman describes a real interview scenario - a 90-minute live coding sprint watched by a VP of engineering and a staff engineer. He came prepared with a codebase containing planted bugs and a feature backlog. When a 1.5GB ZIP file transfer went sideways during setup, he used Claude Code to debug and resolve the issue in real time, in front of his evaluators.
The telling detail is not that the tool solved problems quickly. It is that Lernerman's emotional response shifted from "I wish I had more time" (the classic developer lament) to something closer to abundance. The constraint was no longer his typing speed or his ability to remember CSS syntax. It was how fast he could articulate what he wanted next.
He also built his portfolio website using Claude Code, reportedly mimicking Claude's own UI aesthetic - a small but revealing sign of how deeply the tool has shaped his workflow.
The $200/Month Staff Engineer
Lernerman frames Claude Code's pricing tiers ($20, $100, and $200 per month) as buying access to "staff software engineer level contributions." That is a bold claim, and it overstates what the tool actually delivers - Claude Code is fast and capable, but it still needs a human who understands architecture, tradeoffs, and what "done" looks like. A junior developer with Claude Code does not become a staff engineer. A competent developer with Claude Code becomes significantly faster.
But the directional point holds. At $200/month, even the top tier costs less than a single hour of senior engineering consulting time. For freelancers, small teams, and solo founders, that math changes what is buildable on a given budget.
The essay is one developer's experience, not a benchmark study. But it captures something real about where AI coding tools are heading: the people who benefit most are not the fastest coders. They are the ones with the clearest picture of what they want to build.