Thursday, November 19, 2026

2:30 PM - 3:30 PM

Swan Room

---

Presentation Outline

Allied EW doctrine has called for distributed, mesh-coordinated electronic support systems since the counter-IED after-action reviews of the late 2000s, but the engineering has lagged. The gap is in systems below the $10,000 per node tier that are fast to build, scalable in numbers, and cheap enough to put on every vehicle in a company group rather than a privileged few. As a swarm an EW bubble is no longer a specialist asset attached to specialist platforms; it becomes an organic capability of the deployed force, with every vehicle contributing to and benefiting from a shared electromagnetic picture. This commodity tier, where "quantity has a quality of its own", becomes operationally meaningful, has remained unpopulated. This presentation is a case study in how a small Australian SME is trying to fill that gap, using AI-assisted engineering as a force multiplier, through designing an EW swarm built from commodity AD9361-class software-defined radios, edge compute, GPS-disciplined timing, and an encrypted mesh stack. What was learned trying to build it: the documentation-as-code problem in AI-assisted codebases, the discipline of separating planning context from execution context, the points where AI tooling shines and the points where it fails dangerously. Lessons include phase-coherency challenges in achieving direction-finding, the reference-clock trap, and the slop audit, which found that the AI's worst output wasn't in the code but in the documentation layer, shaping every subsequent session. The meta-lesson: AI-assisted engineering does not make Defence R&D easy. It makes possible what previously wasn't, at a speed of relevance.

---

Speaker