Real data from real engineering teams
Case studies, productivity data, and engineering insights — all from measuring our own team with the tools we build.
GitHub Copilot vs Claude Code: How to Track Adoption and ROI for Engineering Teams
Neither GitHub Copilot nor Claude Code tells you whether it's working across your team. Here's the framework engineering managers use to measure adoption, spot ghost seats, and calculate real ROI for both tools.
Ghost Seats: How to Find and Reclaim Unused AI Coding Licences
24% of our Claude Code seats were never activated. Here's a practical guide to detecting ghost seats in your AI coding tool rollout, understanding why they happen, and getting your licence budget back — without losing the seat.
Your team's AI coding prompts are inconsistent. Here's what that's costing you — and how a shared skills library fixes it.
We audited 668 AI sessions across 22 developers and found 14 different prompts doing the same job. The best one produced 3× better output than the worst. Nobody knew which was which.
We Ran GitHub Copilot and Claude Code Side-by-Side for 30 Days. Here's What the Data Said.
Same team. Same codebase. Same sprints. We split our engineering org across two AI coding tools and measured everything — sessions, commits, PR cycle time, and multipliers by work type. The winner depends on what you're building.
What 4 Weeks of AI-Assisted Coding Did to Our Sprint Velocity, Bug Rate, and Bus Factor
Velocity +28%. Commitment hit rate +18 pts. Bug escape rate −26%. PR cycle time −42%. Bus factor improved from 31% to 19%. The delivery-leader's case for AI tool ROI.
We Tracked Every AI Coding Session Across Our Team for 30 Days. Here's What We Found.
After deploying Claude Code to 29 paid seats across our 22-developer team, we measured every session, prompt, and commit. The results challenged everything we thought we knew about AI tool adoption.