The Workflow Shift

• Engineering throughput is now driven by engineers running five to 10 concurrent agent sessions, abandoning the "one mouse, one screen" constraint for parallel agent coordination.1:53
• Traditional IDEs are being relegated to legacy or enterprise roles as companies build custom internal harnesses—platforms integrating Slack, private data, and agentic workflows—to route work efficiently.4:22

Metrics and Quality

• Platform data is stark: on Linear, agent-created tickets now outnumber human ones for the first time, while GitHub has seen a 10-fold increase in agent-generated pull requests since January.8:13
• Traditional code review is breaking down under this volume, leading to "zombie reviews" where teams selectively bypass scrutiny for non-critical code, creating a measurable strain on software reliability.17:01

Org and Infrastructure

• Migrations that historically took years—such as OpenAI’s Python-to-Rust API shift or Airbnb’s Enzyme-to-React transition—are now being executed in months or even weeks.6:27
• Leadership roles are under unprecedented pressure, with high churn among managers as teams flatten; many leaders are moving to fractional roles or taking career breaks due to increased founder-mode demands and startup instability.20:03
• Infrastructure bottlenecks have expanded beyond GPUs: CPU and memory lead times for cloud capacity have grown from weeks to as long as six months, complicating the strategy to move all development to the cloud.18:33