DevEx 2025 - Year in Review

2025 - in summary

2025 was a year of assessing the landscape and laying foundations for the future of DevEx.

In 2025, DevEx underwent significant transformation: we welcomed 15 new team members, established the API Platform group, and evolved from siloed teams into stream-aligned squads working collaboratively toward shared outcomes. Through two major strategy pivots, we found our footing within the Platforms department and established a unified approach to improving developer experience at GitLab.

Highlights ⭐

Compiled by Claude from our weekly Grand Review issue and quarterly kickoffs - DevEx Kickoff FY26Q3, DevEx Kickoff FY26Q4

Q1: (Feb-Apr)

  1. E2E test infrastructure savings: ~95 minutes per MR pipeline - Reduced pipeline time from 140 minutes to 45 minutes while saving approximately $30,000/month through faster, more reliable pipelines.
  2. Massive CI compute savings: 40,000 hours/month - Achieved roughly 3% reduction in RSpec usage, representing significant infrastructure cost savings and environmental impact.
  3. GDK performance improvements - Delivered tangible developer experience wins with gdk update optimized by ~12 seconds and gdk reconfigure reduced from ~30s to ~9s (67% improvement).
  4. Reviewer Roulette optimization - Improved the reviewer recommendation system by excluding previously reviewed MRs, making code review assignments more effective and reducing bottlenecks.
  5. Test coverage analysis improvement - Fixed rspec:coverage duration with P50 decreasing from 5 minutes to 3 minutes 30 seconds (30% improvement), making test feedback loops faster for developers.

Q2: (May - Jul)

  1. GDK Performance Transformation - Delivered massive productivity gains: 65% faster Docker builds (~80 min → ~28 min), database setup reduced from 15 min to 2 min (>80%), and predictive testing enabled locally cutting test suites to 15 minutes. Reduced stability errors by 25%.
  2. Flaky Test Management System - Built comprehensive infrastructure with dashboards and automated workflows that identified and fixed critical flaky tests, preventing 50-100 failed MR pipelines daily. Fixed one test alone that caused ~19 daily MR timeouts (90 min each).
  3. MCP Server for GitLab API - Delivered Experimental remote MCP server in 18.3, enabling AI assistants to interact with GitLab's API. Required cross-functional sprint on extremely tight timeline with dynamic OAuth and framework development.
  4. Quality Insights Dashboard Launch - Successfully deployed and launched to all Engineering teams on Runway, integrating customer impact tracking, bug trends, and actionable insights. Pilot showed strong positive feedback.
  5. Backport Pipeline Optimization - Reduced backport runtime by 39% (~113 min → ~69 min), directly supporting Delivery's extended Maintenance Policy and reducing release delays from broken stable branches. Bonus: Saved 40,000+ hours/month in CI compute (~3% of RSpec usage) and completed comprehensive Test Strategy documentation enabling the end of Stable-counterpart model.

Q3: (Aug - Oct)

  1. MCP Server Launch - Successfully delivered and released the remote MCP server to GitLab.com and Self-Managed (Experimental in 18.3), enabling AI assistants to interact with GitLab's API. Now referenced in Cursor documentation. Completed rapid cross-functional development and handed over to AI Engineering team.
  2. GDK Stability & Performance Gains - Reduced weekly GDK update errors from 50% (Q1) to 17% of team members affected. Achieved 93% reduction in Ruby installation time (~14 min → <1 min) with precompiled binaries. Added AI-powered troubleshooting and macOS validation testing covering 83% of users.
  3. Test Observability & Data Infrastructure - Established "golden path" for DevEx data with real-time test results pushed to Clickhouse, enabling comprehensive test observability across all test levels. Built initial Grafana dashboards for Engineering-wide visibility into test performance and coverage.
  4. Flaky Test Intervention Program - Implemented automated weekly top 10 flaky test issue creation and launched intervention process requiring teams to fix or quarantine ~500 of 2000 flaky tests. Reduced deployment-blocking test failures and improved engineer confidence in test results.
  5. Performance Testing Evolution - Enabled shift-left performance testing with Duo-integrated reports on backend MRs showing potential performance degradation. Added Git SSH testing capability uncovering Gitaly bottlenecks. Completed incident analysis identifying key performance challenge areas (Gitaly expensive calls, Rails performance issues). Bonus: Launched OpenAPI 3.0/3.1 implementation with first MR merged enabling parallelized development, completed Feature Gates blueprint for cross-department review, and took ownership of LabKit as central DevEx platform for development best practices.

Q4: (Nov & Dec)

  1. Flaky Test Quarantine System - Achieved 33% reduction in pipelines blocked by flaky tests (from ~1,200 to ~800 per week). Making progress towards a fully automated flaky test detection and issue creation with documented process expectations.
  2. CI Pipeline Cost Reduction - Delivered massive savings: disabled gitlab-foss scheduled pipelines saving 24,000 CI hours/month, extended freshness window from 8h to 16h reducing redundant runs, and implemented spec-only MR pipelines for test-only changes. Cost-cutting work will continue through Q4.
  3. AI Development Experience Transformation - Streamlined AI development setup for GDK, eliminating manual configuration steps: automated Duo Agent Platform runner setup, implemented AI Gateway development mode switching, created self-provisioning Duo Enterprise licenses, and published comprehensive AI development guide responding directly to DX survey feedback.
  4. LabKit Standardization & Logging - Announced field logging standards across Engineering (presented at All Hands), implemented runtime linters for immediate developer feedback, and began Go slog migration.
  5. Performance Testing & Test Environment Strategy - Extended Git SSH performance testing identifying Gitaly bottlenecks, fixed CPT tool early termination enabling complete MR performance analysis, and launched comprehensive test environments catalog initiative identifying key gaps and opportunities for unified strategy. Created Duo-enabled debugging guides reducing support toil.

What's next? 🚀

2025 was a year of stabilization and improvements. Flaky tests, high costs, and general scaling challenges created friction for all teams. We've resolved many of the immediate blockers and now have a clear roadmap of work to transform the developer experience into a frictionless, enjoyable experience that works with GitLab's scale.

Our next GetDX survey will go out on 2026-01-12. Results will shape the DevEx Q1 plans. In addition, we'll be:

  1. Strengthening Testing, API, and Quality Systems
  2. Enhancing development tools to support Other Company Initiatives
Edited by Amy Phillips