AI Engineering Trends

Jellyfish measures Access Percentage as the fraction of engineers at a company who have a license to an AI coding tool. As a baseline adoption metric, Access Percentage establishes the foundation for deeper analysis, including how frequently and deeply AI tools are integrated into developer workflows.

Jellyfish measures Weekly Active Users (WAU) percentage as the proportion of engineers at a company who actively use an AI coding tool in a given week. This metric captures the frequency of adoption, showing whether engineers are actually integrating AI tools into their regular workflows.

Jellyfish defines a “frequent AI user” as an engineer who is using AI coding tools three or more days a week. Frequent AI Users Percentage measures the fraction of engineers that have achieved this level of sustained adoption as a separate metric beyond basic WAU.

Jellyfish measures AI Code Percentage as the fraction of a company’s shipped that is AI-assisted. It is calculated as the fraction of merged code additions that were AI-assisted, relative to all code additions in merged pull requests for each company.

This metric moves beyond tool usage frequency to capture the actual depth of AI’s impact on codebases and coding work.

*These figures reflect an updated methodology that more precisely measures AI-assisted code by incorporating validated code acceptance rates, rather than relying solely on AI-assisted coding activity. This provides a more conservative and accurate picture of AI’s true contribution to shipped code.

Jellyfish measures Autonomous Agent Activity as the percentage of pull requests that are autonomously generated by AI agents. A PR qualifies as agent-generated if its commits or the pull request itself were authored independently by an AI coding agent. This metric tracks the emerging frontier of AI adoption, where AI isn’t just assisting engineers, but independently producing shippable work.

Jellyfish measures productivity impact in terms of PR throughput, specifically through differences in the average PRs per engineer merged (by company). The data above represent weekly PRs per engineer aggregated by month (with a trailing three-month average). Companies are grouped by adoption level, defined by the percentage of “frequent” AI users (i.e. engineers using AI coding tools 3+ days per week).

The first plot represents the absolute level of PR throughput for each cohort, while the second plot represents the relative level with respect to the low-adoption cohort as a baseline.

The average PR throughput increases with the adoption level tier, with approximately 2x the throughput at the highest tier relative to the lowest.

One measure of quality is the fraction of “reverted” PRs – code that was deployed but needed to be rolled back. The data above represent the percent of reverted PRs relative to total PRs for each company aggregated by month (with a trailing three-month average). Companies are grouped by adoption level, defined by the percentage of “frequent” AI users (i.e. engineers using AI coding tools 3+ days per week).

The first plot represents the absolute level of revert rate for each cohort, while the second plot represents the relative level with respect to the low-adoption cohort as a baseline.

The average revert percentage increases slightly with higher adoption level tiers, representing a 5-11% increase over the baseline (0.03–0.07 percentage points).