The Problem

Most OutSystems developers learn best practices by reading documentation or watching tutorials. That's fine, but there's a gap between knowing what best practices are and actually applying them under pressure on real code.

There's no structured, hands-on way to practice refactoring, security hardening, or architecture fixes within the OutSystems ecosystem. You either learn on production apps (risky) or you don't get to practice at all.

The Idea

Extend AI Mentor to include a "Practice Mode", a generator that deliberately creates imperfect OutSystems applications for developers to find and fix.

AI Mentor already understands OutSystems best practices deeply. The idea is to flip that knowledge: instead of pointing out what's wrong in your app, it builds a training app that already has the problems baked in, and challenges the developer to find and resolve them.

This is not about AI completing the work. It's about AI leaving the right kind of mess, on purpose.

How It Works — Three Stages

Stage 1 — Foundations (5 issues)

Easiest tier. Designed for developers new to OutSystems best practices.

Typical issues: unused variables, missing indexes on foreign keys, hardcoded values that should be site properties, public actions that should be internal, missing NullIdentifier() checks.

Stage 2 — Intermediate (10 issues)

Mid-tier. Developers need to think about architecture and maintainability.

Typical issues: logic inside Preparation that belongs in a Server Action, missing input validation, exposed sensitive data in REST APIs, poor naming conventions, N+1 query patterns, missing error handling in integrations.

Stage 3 — Advanced (15–20 issues)

Senior-level challenge. Security, performance, and architectural decisions.

Typical issues: SQL injection surface areas, missing role checks on screens and actions, exposed system entities, broken session handling, unoptimized aggregate fetching, tight coupling between modules, missing audit trail logic, improper use of client-side vs server-side logic (maybe bpt and tenant optimization and error handling too).

Goal is simple do that changes remove error, remove warning and publish the application.

Why This Works

The value isn't in the generated app itself — it's in the deliberate, structured imperfection. Developers get a realistic codebase with real OutSystems patterns, real module structures, and real antipatterns to hunt down.

AI Mentor can then serve as the referee: once the developer believes they've fixed the issues, AI Mentor reviews the app and confirms whether each problem has been correctly resolved — closing the feedback loop without giving away the answers upfront.

Expected Outcomes

- Developers build muscle memory for catching common OutSystems antipatterns

- Junior developers have a safe, structured environment to level up

- Teams can use Stage 3 as a pre-promotion or onboarding assessment

- OutSystems certifications could eventually reference this as a practical preparation tool

Summary

AI Mentor already knows everything that's wrong with your app. Let it build an app that's wrong on purpose — and teach developers by making them fix it.

Stage 1: find 5. Stage 2: find 10. Stage 3: find 15–20.

Learn by doing, not by reading.