Growing & Thriving in a Multi-model World • Alberto Brandolini • GOTO 2025

Alberto Brandolini delivers a career-spanning reflection on why single, unified domain models inevitably fail and how multi-model approaches built around bounded contexts are the antidote—but one that requires unnatural discipline to sustain. The core argument: software naturally trends toward a 'big ball of mud' because our brains default to verbal, data-centric thinking that ignores purpose. Teams merge unrelated concepts (e.g., invoices and forecasts), creating models that serve too many stakeholders, accumulate fear of change, and eventually produce the classic plateau where more investment yields diminishing returns. Brandolini distills a decade of experience from his own company into practical countermeasures: never modify a working bounded context; never make design decisions without a visual map; always generate three alternatives before choosing a solution; and use context maps to detect the real friction points—whether business value, aesthetic ugliness, or fear. The talk extends into multi-market scenarios, dissecting why the 'second country' implementation always fails when it inherits a brownfield codebase with a junior team. It also challenges the authority of domain experts, regulators, and business timelines, arguing that software teams must own their models rather than replicating external narratives. The most transferable insight: teams should split when signals are clear and decisions are cheap, not when scaling forces their hand. The session is rich with psychological observations about modeling—visual anchoring outperforms verbal reasoning, and addiction-replacement patterns can reshape team habits. Missing are empirical benchmarks or controlled comparisons; the evidence is auto-ethnographic, drawn from consulting and a single long-lived internal project. That doesn't invalidate the lessons, but it limits generalizability. Engineering managers will find the most value in the practical decision-making heuristics and the framing of fear as a legitimate architectural driver. The final minutes rush through multi-country domain expertise, leaving the most novel ideas underdeveloped.