Uploaded June 2026 | Updated September 2026, 2 weeks ago
Can AI completely eliminate engineering toil, or is it just generating massive PRs that nobody wants to review? In this InfoQ video, Meta Software Engineer Ian Thomas reveals exactly how Meta’s Reality Labs scaled an organic AI-for-productivity (AI4P) community from 0 to 400+ engineers, drove tool adoption to 80%, and successfully navigated the pitfalls of AI code "slop."
Senior software developers, architects, and engineering leaders will learn a battle-tested framework for moving teams from mere tool adoption to a truly AI-native workflow without sacrificing software craftsmanship.
Discover how one team used an unsupervised agentic workflow to skyrocket code coverage to 90% in just 3 hours of manual effort, and how Meta mitigates the architectural risks of massive AI-generated diffs.
⏱️ Video Timestamps (For Navigation)
0:00 – The Vision: Shifting from Builders to Innovators
1:15 – Launching the AI4P (AI for Productivity) Community at Meta
2:31 – The "Pit of Despair": Why Ad-Hoc AI Tooling Fails
3:47 – Building a DORA-Based AI Maturity Framework
5:10 – The 6 Dimensions of AI Integration (From 'SIT' to 'LEAP')
7:02 – AI Across the SDLC: Beyond Simple Autocomplete
8:15 – The Incident Risk Tool (DRS) & Custom Agents (Confucius)
9:44 – Case Study 1: 90% Test Coverage via Unsupervised Agents
11:15 – Case Study 2: Performance-Sensitive Unity Code Migrations
12:20 – Case Study 3: Querying Immersive World State via MCP
13:58 – The Crisis of Code Review: Handling Massive 4,000-Line Diffs
15:11 – Fighting Code "Slop" with Automated AI Test Reviewers
16:45 – Redefining ROI: Moving Past Weekly Active Users (WAU)
18:22 – 5 Leadership Playbooks for AI-Native Engineering
20:10 – Q&A: Handling Top-Down Executive Pressure & Code Accountability
🔗 Transcript available on InfoQ: bit.ly/4uRqTYq
#AINativeEngineering #SoftwareArchitecture #EngineeringLeadership #MetaRealityLabs #DevEx
Can AI completely eliminate engineering toil, or is it just generating massive PRs that nobody wants to review? In this InfoQ video, Meta Software Engineer Ian Thomas reveals exactly how Meta’s Reality Labs scaled an organic AI-for-productivity (AI4P) community from 0 to 400+ engineers, drove tool adoption to 80%, and successfully navigated the pitfalls of AI code "slop."
Senior software developers, architects, and engineering leaders will learn a battle-tested framework for moving teams from mere tool adoption to a truly AI-native workflow without sacrificing software craftsmanship.
Discover how one team used an unsupervised agentic workflow to skyrocket code coverage to 90% in just 3 hours of manual effort, and how Meta mitigates the architectural risks of massive AI-generated diffs.
⏱️ Video Timestamps (For Navigation)
0:00 – The Vision: Shifting from Builders to Innovators
1:15 – Launching the AI4P (AI for Productivity) Community at Meta
2:31 – The "Pit of Despair": Why Ad-Hoc AI Tooling Fails
3:47 – Building a DORA-Based AI Maturity Framework
5:10 – The 6 Dimensions of AI Integration (From 'SIT' to 'LEAP')
7:02 – AI Across the SDLC: Beyond Simple Autocomplete
8:15 – The Incident Risk Tool (DRS) & Custom Agents (Confucius)
9:44 – Case Study 1: 90% Test Coverage via Unsupervised Agents
11:15 – Case Study 2: Performance-Sensitive Unity Code Migrations
12:20 – Case Study 3: Querying Immersive World State via MCP
13:58 – The Crisis of Code Review: Handling Massive 4,000-Line Diffs
15:11 – Fighting Code "Slop" with Automated AI Test Reviewers
16:45 – Redefining ROI: Moving Past Weekly Active Users (WAU)
18:22 – 5 Leadership Playbooks for AI-Native Engineering
20:10 – Q&A: Handling Top-Down Executive Pressure & Code Accountability
🔗 Transcript available on InfoQ: bit.ly/4uRqTYq
#AINativeEngineering #SoftwareArchitecture #EngineeringLeadership #MetaRealityLabs #DevEx
