Escola Lendária

Escola Lendária is a production education platform created in Mozambique to make structured practical learning accessible from an ordinary phone or computer. It organizes authored lessons into progressive learning paths: learners study or practise one lesson, complete the required activity, and unlock the next lesson. Each course offers five free practice lessons before course-specific payment is requested.

Its core educational approach is the PSF learning method. Instead of beginning with a finished definition, formula, or technical name, each lesson starts with something the learner can observe, manipulate, compare, test, or reconstruct. The learner moves through a fluid sequence: observe, attempt, compare, discover, name, practise, and validate. Meaning is built before terminology. In technical electricity, for example, a learner follows phase and neutral, connects conductors, closes a circuit, and discovers why a lamp turns on before receiving the formal explanation.

The platform contains more than one thousand authored lessons across mathematics, physics, chemistry, binary, programming, technical electricity, typing, and related practical areas, plus a motherboard laboratory simulator. The curriculum is authored and organized according to the PSF learning method; Gemini does not automatically generate the curriculum.

How AI operates in the business

Gemini is used as an operational layer around the authored school rather than as a generic chatbot. A learner can ask a contextual question inside the lesson currently open. The application passes only the relevant lesson context through a Supabase Edge Function to a server-side Gemini Operations service, which calls Gemini 2.5 Flash Lite through Google Cloud Vertex AI Express mode. The response is grounded in the lesson title, concept, and content so that AI supports the learning path instead of replacing it.

The same isolated service supports administrative triage. It can classify an incoming support message by category and priority, summarize it, and recommend the next human action. This is useful as the number of learners grows because the first organizational pass can be automated while consequential decisions remain controlled by a person.

Humans remain responsible for curriculum authorship, content approval, payment confirmation, course unlocking, and other consequential actions. AI is responsible for contextual learner assistance and the initial organization of support information. The architecture intentionally separates these roles: the Gemini API key remains server-side, the browser never receives it, and the payment/unlocking workflow is not delegated to the model.

During the hackathon we built and tested the Gemini path, deployed the server-side AI service, created the authenticated application bridge, and integrated the learner-facing support workflow. Evidence includes a successful authenticated Vertex AI Express call to gemini-2.5-flash-lite, the deployed AI service and public health endpoint, an active Supabase psf-gemini-api Edge Function, application screenshots, and CI checks for the Gemini service and application bridge.

Business model and early traction

Escola Lendária is currently a direct-to-consumer education business. A learner can enter a course and complete the first five practical lessons without paying. Continuing that course costs 100 MZN as a one-time course access payment, with progress still unlocked sequentially rather than opening every lesson at once. Payment confirmation is performed by a human administrator.

The production records used for the hackathon P&L show two paid course orders of 100 MZN each, for 200 MZN total revenue from one distinct paying account. We deliberately exclude test accounts and unverified registrations from traction claims. The documented cash expense in the working P&L is the project domain/operating payment; no paid marketing campaign or directly attributable COGS was recorded in that statement.

The model is designed to keep marginal delivery costs low. One authored course can serve multiple learners, while Gemini can provide contextual assistance and help organize support without requiring a human to answer every first-line question. Revenue can grow through additional learners and additional paid courses while the same platform, payment workflow, lesson engine, and AI operations layer are reused.

Impact and economic opportunity

The immediate impact target is access to structured practice and feedback for learners who may not have continuous access to an individual tutor. We measure progress through verified users, lesson completion, practice success, course continuation, free-to-paid conversion, paying users, and revenue. Over time, we also want to measure whether contextual assistance helps learners continue independently instead of abandoning a lesson when they become stuck.

The broader opportunity is to turn locally authored knowledge into a scalable educational service for Mozambique and other Portuguese-speaking markets. As the school grows, it can create or enable work around curriculum authorship, lesson review, learner support, administration, technical maintenance, and local educational content. AI is intended to increase the amount of learning and support that a small team can deliver, not to remove human responsibility for education.

The project was built with a simple principle: use AI where it can multiply access and operational capacity, and keep humans accountable where judgment, money, curriculum, and learner consequences are involved.

Built With

• cloudflare
• docker
• gemini-2.5-flash-lite
• github-actions
• google-cloud-vertex-ai-express
• next.js
• node.js
• supabase
• typescript