Over the past sprint, we tightened our engineering habits, deepened our campus partnerships, and turned hackathon sparks into production-ready features – while threading the needle on privacy and deployment inside UVic’s ecosystem. Same crew, same energy, just a little more grown-up about process. And like we said from the start, we kept the human side front and centre with quick stand-ups, clear communications, and shared priorities so nothing fell through the cracks.

A big unlock was nailing our deployment cadence with UVic Systems, which turned “when can we release?” into a predictable rhythm. Regular touchpoints gave us a faster path for approvals and clearer expectations around observability, so our biweekly drops land safely on campus infrastructure instead of as unpredictable big-bangs. Just as important, we aligned on analytics instruments and workflows – using data to encourage course-level adoption and to actually learn what students prefer, not just guess.

We also waded into the fun stuff: AI, privacy, and licensing. Auto Quiz Generator feature brushed up against textbook licenses and privacy impact assessments, so we looped in the UVic Library and mapped the governance route across Brightspace, ethics, and privacy. On the build side, with guidance from our AI mentor Saman Rahbar, we mapped a stepwise implementation plan for features like quiz generation: starting with LangChain-based RAG, layering in evaluation metrics, and preferring open-source models where feasible. In short, responsible capability, one measured step at a time.

A highlight that reframed our thinking came from Self-Regulated Learning (SRL) expert Professor
Allyson Hadwin. The core lesson: students make the biggest gains when they author the plan, not when the tool does it for them. Instead of breaking tasks down for them, Mathiné should help students break their own work into bite-sized, one-hour chunks that fit real schedules. She also underscored
confidence (self-efficacy) as the hinge – when students feel even “slightly confident” about planning, time management, and foundational academic behaviours (showing up, getting work done, participating), engagement follows. With that in mind, our immediate next steps are about learning fast and staying honest. We’re heading back into classrooms to run a short survey on confidence, challenges, and SRL practices, then a follow-up at the end of November to see what moved. In parallel, we’ll keep the privacy-approval track moving for AI features and use usage analytics to understand which tools students actually reach for. Fewer assumptions, more evidence – and a tighter loop between what we build and what students
need.

From Reflection 1’s “head first” energy to Reflection 2’s “full speed ahead,” this chapter is our third act: bridging – between prototypes and production, between innovation and policy, and between student needs and institutional requirements. That bridge is how Mathiné scales responsibly, and it’s the kind of work we’re excited to keep doing – together!