The Prompt Is the New Commit -- And You're Already Being Monitored

There's a quiet infrastructure buildout happening in software engineering right now, and most developers have no idea it's pointed at them.

A new class of tools -- GuageAI, Codemetrics, Plandek, Waydev, Exceeds AI -- can now tell your engineering manager exactly how much of your code was written by AI, whether it shipped or got reverted, how your AI adoption compares to your peers, and what it's costing the company per pull request. Microsoft's Purview audit logs capture Copilot interactions automatically, on by default, retained for 180 days. OpenAI disclosed it monitors the full reasoning chains of its internal coding agents within 30 minutes of each session.

The prompt is the new commit. And most engineers don't know they're already being graded on it.

41% of Code Is Now AI-Generated. That Changed Everything.

According to Exceeds AI's March 2026 analysis, AI now generates 41% of code globally. 84% of developers use AI coding tools. Nearly half of companies report that 50%+ of their codebase is AI-authored.

CircleCI's 2026 State of Software Delivery report -- drawn from more than 28 million workflows -- found that AI-assisted development drove a 59% increase in average engineering throughput. The Pragmatic Engineer's survey of 18 companies including Google, Microsoft, and Dropbox found 85% of engineers now use AI coding tools at work.

When nearly half the code in a repository is machine-generated, the old question of "how productive is this engineer?" gets replaced by a harder one: "how effectively does this engineer use AI?"

And that question has spawned an entire measurement industry.

The Old Metrics Are Dead

Lines of code died as a metric decades ago. Story points were always a team-internal fiction. Even the DORA framework -- deployment frequency, lead time, change failure rate, mean time to recovery -- was designed for a world where humans wrote all the code.

The problem: AI broke the relationship between effort and output.

An engineer using Cursor can mass-produce pull requests that look impressive on a DORA dashboard but introduce 2x the rework risk and 30% higher change failure rates (Exceeds AI). Plandek's 2026 benchmark of 2,000+ engineering teams found that despite AI boosting throughput, bottom-quartile teams still take 35+ hours to merge pull requests. Waydev's analysis found feature branch throughput up 15.2% -- but main branch throughput down 6.8%.

More code is entering the pipeline. Less is reaching customers.

"AI is helping teams move faster, but it's also exposing where the delivery bottlenecks are," said Plandek CEO Charlie Ponsonby.

(remaining article unchanged)