2026-10-05 –, Bridge 2
AI coding assistants are easy to overhype. In practice, the interesting question is much narrower: where do agentic coding tools actually help when building and maintaining a real QGIS plugin?
This talk presents lessons from developing qfit, a QGIS plugin for importing, analysing, and publishing personal fitness activity data, using OpenClaw as an agentic coding environment. Rather than treating AI as a magic code generator, I will focus on practical workflows: breaking work into small reviewable changes, using agents to explore code, draft implementations, refactor modules, and strengthen tests, while keeping a human firmly responsible for architecture, validation, and release quality.
I will share what worked well, where the approach failed or created extra review burden, and which guardrails mattered most, especially around testing, CI, code review, and incremental design in a PyQGIS codebase. Attendees will leave with concrete patterns for using agentic coding in QGIS plugin development without lowering engineering standards.
This session is aimed at developers and technical QGIS users who are curious about AI-assisted plugin development but want concrete practices rather than hype.
Emmanuel Belo is a geospatial software and business leader at Camptocamp, where he leads the Geospatial Solutions Business Unit and serves on the management board. With more than 20 years of experience in open source geospatial technologies, he works at the intersection of GIS, custom software development, cloud infrastructure, and applied AI. He is also the creator of qfit, a QGIS plugin for importing, visualizing, and publishing fitness activity data.