Descriptions:
Paul Everitt, developer advocate at JetBrains — a 25-year-old, privately held, profitable European company — delivers a keynote at the AI Dev 26 x SF conference making the case for what he calls “agentic engineering.” His central argument challenges the dominant 10x productivity narrative: citing DX Research data and 2024 Nobel Prize-winning economist Daron Acemoglu’s analysis of the productivity paradox, Everitt argues that code generation speed was never the bottleneck in software delivery. Speeding up code output alone yields roughly 10% gains, not the transformational results many vendors promise.
The talk synthesizes thinking from several influential voices in the field — Simon Willison’s taxonomy of agentic patterns, Addy Osmani’s writing on the discipline, and OpenAI’s internal “harness engineering” framework — to define what agentic engineering actually means in practice. Rather than vibe coding (Andrej Karpathy’s term for prompt-and-hope development), agentic engineering asks engineers to shift identity: from people who build things to people who build the systems that build things.
Everitt argues that organizations chasing layoffs and raw code output are asking the wrong questions, and that engineering teams need to offer leadership a new frame — one centered on system design, scaffolding, and human augmentation rather than headcount reduction. The talk is particularly valuable for senior engineers and engineering leaders trying to articulate why AI coding tools haven’t delivered organizational-level impact despite individual productivity gains.
📺 Source: DeepLearningAI · Published May 22, 2026
🏷️ Format: Deep Dive
