Descriptions:
At The Pragmatic Summit, Gergely Orosz presents fresh industry benchmark data covering 121,000 developers at more than 450 companies, collected from November 2025 through February 2026. The headline numbers: 92.6% of developers use an AI coding assistant at least once a month, 75% at least weekly, and self-reported time savings hover around 4.08 hours per developer per week — a figure largely stable across recent quarters. Tools most commonly cited include Cursor, GitHub Copilot, and Claude.
The talk frames these findings against a broader argument about balancing AI enthusiasm with organizational pragmatism, drawing a parallel to the space race. Just as space exploration produced lasting improvements to life on Earth without requiring everyone to live on the moon, Orosz argues that the value of AI experimentation compounds even when individual organizations don’t adopt the most cutting-edge patterns. He pushes back on both the most optimistic productivity projections and the most dismissive skeptics, citing studies — including Google’s approximately 10% productivity increase figure — as useful but incomplete signals.
The second half turns to agentic workflows, arguing that agents expand what can be built and who can build it. This section is especially relevant for engineering leaders evaluating whether to invest in agent infrastructure versus more incremental AI tooling. The data-grounded framing makes this a useful counterweight to hype-heavy AI coverage, particularly for teams making evidence-based decisions about tooling and organizational strategy.
📺 Source: The Pragmatic Engineer · Published February 24, 2026
🏷️ Format: Deep Dive
