~/talk

How to Scale AI Coding for Real Business Outcomes | Jellyfish Use Case

on YouTube 27m
Talkat Sonar Summit 2026 · on Tuesday, 2026-03-03· [ai-assisted-development]

speaker

name:NICHOLAS ARCOLANO, PHD
links:none on file

resources

video:https://youtu.be/9Lsi5vEzTb0
slides:none on file

summary

Jellyfish aggregates 20 million pull requests from roughly 200,000 engineers at 1,000 companies and finds concrete effects: interactive tools like Copilot Cursor and Claude Code correlate with 2× PR throughput and 24% faster cycle time as adoption rises, autonomous agents (Codex and others) reach ~1% median of PRs with top adopters near 10%, PR size grows ~18% and revert rates tick up 7–11% while bug reports stay flat. Centralized repos show ~4× gains; highly distributed architectures show no uplift due to context gaps.

AI-written summary · about

chapters (22)

  1. 0:00Introduction
  2. 0:17Jellyfish: AI Observability for Engineering Teams
  3. 0:45Key Questions About AI Transformation in Software Development
  4. 2:05The Jellyfish AI Engineering Trends Research Site
  5. 2:34Data Sources: Developer Tools, Source Control, and Task Systems
  6. 4:12Measuring AI Adoption: Weekly Active Usage of AI Coding Tools
  7. 5:11Adoption Depth: Identifying Frequent AI Coding Users
  8. 7:18Trend: Growth of Autonomous AI Coding Agents
  9. 8:29Autonomous Agent Pull Requests: Median Performance vs Outliers
  10. 9:33Defining Developer Productivity in the AI Coding Era
  11. 10:05Key Productivity Metrics: Pull Request Throughput and Cycle Time
  12. 11:06Correlation Between Pull Request Throughput and AI Adoption
  13. 12:38Faster Delivery: Cycle Time Drops as AI Adoption Increases
  14. 13:53Side Effect of AI Coding: Pull Requests Getting Larger
  15. 15:23Code Quality Signals: Bugs and Pull Request Reverts
  16. 16:13Results: Slight Increase in Reverts, Bugs Stay Flat
  17. 17:57When AI Productivity Expectations Don’t Match Reality
  18. 18:51Architecture Lens: Active Repositories per Engineer
  19. 20:21Software Architecture Distribution Across Four Quartiles
  20. 21:42Why AI Productivity Gains Differ by Architecture Type
  21. 23:13Root Cause: Context Gaps Across Multiple Repositories
  22. 25:07Recap: AI Adoption, Productivity Metrics, and Context Challenges

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.