Field noteOct 13, 20262 min read1 source

This concern shows up in developer surveys too.

In one minute

The Stack Overflow 2025 Developer Survey reports that 66% of respondents to its AI-frustrations question encountered answers that were “almost right.” This is a signal about users’ reported experience, not a measured error rate for AI-generated code.

Thesis

The survey points to verification work as a practical friction in AI-assisted development, without quantifying the underlying code-error rate.

Observed facts

Stack Overflow reports that 84% of respondents use or plan to use AI tools in development.
1
The survey says 51% of professional developers use AI tools daily.
1
66% of respondents to the AI-frustrations question reported dealing with answers that were almost right.
1

Mechanism

An answer that appears plausible but is subtly wrong can shift effort from producing a first draft to checking assumptions, edge cases, and integration.

Why now

Reported daily use among professional developers makes the cost of reviewing AI output relevant to ordinary workflow design, not just experimentation.

Implications

Teams should evaluate AI assistance alongside the review and testing work needed to trust its output.
Perceived friction is not equivalent to an observed code-error rate.

Evidence boundary

These are survey responses, not an independent code-quality benchmark, and the excerpt provides no error-rate measurement.

Watch conditions

Watch for future evaluations that pair adoption data with task-level correctness and review effort.

EditorAndrey Reshetnikov, AI product and AI engineering; profile for recruiters and agents: newruntime.com/owner-profile.json

Evidence / sources

1survey.stackoverflow.co/2025/ai

All Field Notes