Every product update we ship starts with the same question: does this make it measurably easier for learners to close their skill gaps faster? If the answer is no, we don't build it. Q2 2025 was a quarter where almost every item on our roadmap passed that test, which made it both exciting to execute and challenging to fit into 13 weeks.
Team Analytics Dashboard (General Availability)
After three months in beta with 12 enterprise customers, our Team Analytics Dashboard is now generally available to all Team and Enterprise plan subscribers. The dashboard gives L&D managers and people leaders a real-time view of their team's learning activity, competency score trajectories, and skill gap heat maps — all in a single interface that doesn't require any data export or manual reporting.
Key capabilities in the initial release include: team-level competency score distribution by skill domain, individual learner progress against assigned learning paths, completion rate trends over rolling 30/60/90-day windows, and a skills gap heatmap that shows which competencies are most underdeveloped relative to role benchmarks. We built the dashboard to answer the question an L&D manager faces in a Monday morning meeting: "Is our team's learning program actually working?"
In beta, managers who had access to the analytics dashboard had a 34% higher rate of proactive learner intervention — reaching out to struggling learners before they disengaged — compared to managers without the dashboard. That intervention rate correlated with a 22% improvement in completion rates for at-risk learners. Data-informed management of learning turns out to work.
Spaced Repetition Engine v2.0
Our original spaced repetition engine used a modified version of the SM-2 algorithm, which has been the foundation of spaced repetition software since Piotr Wozniak's work in the late 1980s. SM-2 is excellent, but it was designed for uniform flashcard-style review and does not account for the structural relationships between concepts in a knowledge domain or the varying difficulty of different recall question types.
Our v2.0 engine replaces SM-2 with a deep knowledge tracing model that maintains a probabilistic knowledge graph for each learner. Review scheduling now accounts for concept dependencies — if you haven't solidified a foundational concept, the system won't schedule advanced applications of it for review until the foundation is stable. It also accounts for question type difficulty: a scenario-based application question requires higher knowledge confidence before scheduling than a definition recall question. Initial A/B testing shows that v2.0 produces 28% better 90-day retention compared to our previous engine, with no increase in review session frequency.
Generative AI Content Module (Early Access)
Our generative AI content module — the most technically ambitious thing we have built at Learpy — is now in early access for select Pro and Enterprise subscribers. This module uses large language models to generate personalized practice exercises, alternative explanations, and on-demand Q&A interactions calibrated to a learner's current knowledge state.
Concretely: when a learner encounters a concept they don't understand, they can now ask the system to "explain this differently" and receive a fresh explanation using a different analogy or structural framing. When they complete a module, the system can generate novel practice problems in their industry context that test the same underlying skill. This is not a chatbot bolted onto the platform — it is a context-aware content generation system that has full access to the learner's knowledge graph and learning history.
Early access feedback has been strongly positive, with one beta user describing it as "the first time an online learning tool actually answered my specific question rather than sending me to a FAQ." We will graduate it to general availability for Pro users in Q3 2025 after completing safety evaluations and addressing hallucination edge cases we identified in beta.
Other Improvements
We shipped 11 additional improvements in Q2, including: offline mode for iOS and Android (download up to 5 hours of content for offline learning), improved skill assessment diagnostic accuracy following a retraining on our largest dataset to date, Workday integration for automatic learner provisioning and competency score sync, a new cohort-based learning mode that lets enterprise customers run cohort learning programs on top of individual adaptive paths, and significant accessibility improvements including full WCAG 2.1 AA compliance across the web application.
What's Coming in Q3 2025
Q3 will see the general availability of the generative AI content module, the launch of our first non-English language track curriculum (Portuguese, followed by Spanish), the release of our public API for enterprise integrations, and the beginning of our beta program for AI-powered manager coaching — a feature that uses learning and performance data to suggest specific development conversations for managers to have with their direct reports. We are also running a significant infrastructure upgrade to support the next phase of scale as our enterprise business grows.
If you have feedback on any of our Q2 releases or want early access to any Q3 features, reach out via our product feedback channel or contact our customer success team directly.