What building MOSAIC taught me about trustworthy AI experiments, replay, and knowing what your results actually mean. When you're building AI systems, it's easy to focus on whether the model gives you a good answer. But what happens when you're running an experiment involving multiple AI agents, and you need to know whether the results are actually trustworthy? That's the question I've been exploring while building MOSAIC , a research project focused on studying the behaviour of popul...