The university co-creation space "Transformer" at TU Wien serves as an experimental field for extracurricular STEM and climate education. Within the framework of the course "Soft Skills for Engineers - Science Communication and Co-Creation", students conducted field observations across 26 workshop formats involving 250 children and adolescents. The central research question was: What didactic success factors and challenges emerge in communicating climate change and sustainability topics in informal learning environments? The workshops encompassed a broad spectrum from upcycling and CO₂ experiments to ecological construction. Particular attention was paid to how technology-supported maker approaches (3D printing, digital fabrication) can contribute to climate education. The study further examines how university third-mission activities contribute to transforming climate knowledge into action competence and what role co-creation spaces can play as bridges between science and society. Methodology Data collection was conducted through structured participant observation using a standardised observation protocol. Eleven students systematically documented framework conditions, interaction patterns, learning processes and challenges between March and June 2025. Qualitative analysis followed a mixed-methods approach: observation reports were initially analysed independently by four different AI systems (M365 Copilot, Gemini 2.5 Pro, ChatGPT-5, Claude Opus 4.1) to obtain diverse analytical perspectives. The divergent and convergent analysis results were subsequently triangulated and condensed into overarching design principles. This innovative methodology of AI-supported multi-perspective analysis enabled particularly robust identification of recurring patterns. Divergences and redundancies across the four LLM syntheses were comparatively examined (consensus/dissent matrix) to separate robust patterns from model-induced artefacts. Methodological rigour was addressed through peer debriefings within the course, independent coding of selected reports, and negative case analysis. Limitations: heterogeneous workshop topics and durations, observer effects, absence of standardised outcome measurement; the LLM meta-level offers breadth but carries risks (style/priority drift per model). Results Four LLMs analyzed identical prompts but showed distinct approaches: All identified hands-on learning, age differentiation, scaffolded freedom, and technology motivation as key themes. However the LLMs prioritize differently: theoretical-scientific (ChatGPT, Gemini) to pragmatic-applied (Claude, M365 Copilot). The analysis identified six central design principles for successful climate education in makerspaces: 1. Material Literacy for Climate Understanding: Abstract climate concepts become tangible through direct material engagement. Children understood CO₂ emissions significantly better through shell experiments on ocean acidification than through theoretical explanations. Upcycling projects led to measurable awareness of circular economy principles. 2. Productive Failure as Resilience Training: Iterative design processes in 3D printing or construction tasks promote problem-solving competence - a key qualification for climate challenges. Children whose initial attempts failed demonstrated higher persistence and more creative solution approaches. 3. Scaffolded Freedom: Open-ended tasks require clear microstructures. Successful workshops combined creative freedom with visual instructions, temporal structures and exemplar artefacts. Completely free formats led to cognitive overload, whilst overly rigid specifications diminished motivation. 4. Technology as Bridge to Sustainability: 3D printing and digital fabrication acted as powerful motivators but required didactic orchestration. Productive utilisation of waiting times for material knowledge workshops or sustainability reflections. The combination of digital and analogue proved particularly conducive to learning. 5. Multi-level Differentiation: Mixed-age groups (6-14 years) necessitated adaptive task design. Successful formats operated across three levels: phenomenological (experiencing), model-based (understanding), systemic (abstracting). For instance, younger children created objects from waste materials (experiencing), whilst older ones analysed material lifecycles and circular economy principles (understanding/abstracting).Community Learning: Parental involvement had ambivalent effects - supportive yet potentially inequality-reinforcing. Peer learning between different age groups facilitated knowledge transfer. The Transformer evolved into a social learning space for families and communities. Climate Relevance: The workshops demonstrated that makerspaces constitute ideal venues for transformative climate education. Through the integration of creativity, technology and sustainability, action competence emerges rather than climate anxiety. Particularly successful were projects with local relevance (urban gardening and greening, vacant space transformation) and visible impact. The findings underscore the significance of university co-creation spaces as catalysts for societal transformation. They provide low-threshold access to climate topics and foster intergenerational learning. For scaling, the development of modular workshop concepts and systematic training of multipliers is recommended.