GitHub displays our final commits as Aug 9 because our team works in Pakistan (UTC+5). The IncludAI deadline of Aug 8, 11:45 PM PT corresponded to Aug 9, 11:45 AM PKT. Our final submission commit was completed before that cutoff. Log: Type2Learn Commit History 4d87959 — docs: present Type2Learn product and architecture Documented the implemented adaptive-learning architecture, learner safeguards, consent/data boundaries, model routing, understanding-check design and feature-gated capabilities.
73edc53 — feat: strengthen adaptive learning and model routing Added AI-use/learning telemetry integration, purpose-specific GPT-5.4 Nano/Mini routing, stronger adaptive support, deterministic assessment evaluation, assessment integration and automated tests.
5dac8b1 — feat: harden objective-based adaptive assessment — final eligible submission commit Added the new assessment-monitor.mjs layer and integrated it into the assessment service. The final pipeline became: reviewed/authored assessment bank → deterministic item ordering → deterministic evidence evaluation → optional scoped GPT-5.4 Mini evaluation → objective-evidence monitor → targeted supportive next step. Behavioural/interaction summaries may affect which question is asked first, but cannot determine whether a learner passes or needs review. The monitor identifies the exact curriculum objective/module needing another look, while the learner sees an “Understanding check” and “Next helpful step,” never a percentage or ranking. Raw answers, chosen MCQ options, scores and model rationale are deliberately excluded from learner-progress storage. The flow also retains no-provider fallback checks and avoids forced/unlimited retestin
Log in or sign up for Devpost to join the conversation.