Summary
Stephanie Jarmak, a research scientist at Sourcegraph, delivers a conference talk reframing the classic developer advocacy role for an era where AI agents are now first-class users of developer tools. Drawing on her own background as an astronomer-turned-open-source-maintainer, she traces DevRel from 1980s software evangelism through the 2010s feedback-loop model and argues that the role must now evolve into what she calls “agent advocacy.”
The core of the talk covers two dimensions of this shift. First, agents read documentation, call APIs, and encounter errors just as human developers do — meaning DevRel teams need to instrument and measure how agents interact with their products, not just humans. Second, Jarmak introduces generative engine optimization (GEO) as the agent-era analogue to SEO: she ran experiments testing how often various AI chatbots recommended Sourcegraph’s code-intelligence product, finding a 65% mention rate when prompts mimicked active comparison shopping but a 0% rate when prompts described a specific pain point that the product directly solves.
The talk concludes that this gap reveals a messaging problem — agent training data may not associate certain customer pain points with relevant tooling — and Jarmak outlines a content-campaign approach to test whether changes to Sourcegraph’s public-facing copy produce measurable lift in agent recommendations. Practical GitHub repos with toy projects are shared for attendees to explore.
📺 Source: AI Engineer · Published August 26, 2026
🏷️ Format: Keynote Launch
