Talks

• CopenhagenJS: “Agentic whatt??” (15 October 2026)

  Telling quality apart from plausible in AI-generated code: why plausibility wins by default (sycophancy, instruction decay, gaming the oracle), the ten-gate setup that fights back (deterministic rules, behavioural oracle, mutation testing, an anti-gaming charter judged by a second model) and the AI-FOMO trap. Hosted at Danske Bank. Slides, charter and ten-point checklist.