Launching today
CartesiumAI
Learn by experimenting with a real-time AI companion
1 follower
Learn by experimenting with a real-time AI companion
1 follower
CartesiumAI turns learning into a live, evidence-driven process. Its context-aware 3D companion explains selected concepts, reads the learner’s journal and simulation state, and performs safe real-time demonstrations. Ten deterministic physics labs let learners predict, experiment, collect evidence, revise explanations, and build auditable reports. This can be extended to any other discipline, leading to a high-quality learning experience. Built end to end with Codex and designed for GPT-5.6.
Hi Product Hunt! I’m excited to share CartesiumAI 👋
I built CartesiumAI because most AI learning tools answer questions beside static content. I wanted to explore a different role: an AI companion that shares the learner’s context, safely acts inside an interactive environment, and helps turn predictions into evidence.
This creates an immersive learning journey in which users do more than find answers: they understand, internalize, and apply concepts, building durable knowledge and greater awareness of how they learn.
Physics is the first proving ground:
• 10 deterministic 2D and 3D laboratories
• a context-aware 3D coach designed for GPT-5.6
• safe, schema-validated simulation actions
• observations, snapshots, conceptual-change reflection, and evidence reports
• a visible audit trail connecting message → action → deterministic state change
The numerical engine, not the language model, remains the source of truth. When live model access is unavailable, the public demo clearly identifies and uses a deterministic learning-method engine, so the complete workflow remains testable.
CartesiumAI was engineered end to end through a Codex-led workflow. I’d especially value feedback on whether the predict → experiment → compare → explain → transfer loop feels clear to a first-time learner, and where the companion is most helpful.
During the review period, live GPT‑5.6 access is limited to authenticated judges, while public sessions use a clearly identified deterministic learning engine. In both modes, simulations and AI-requested state changes remain bounded, deterministic, and auditable.
Try it at https://cartesium.me. Thanks for taking a look!