Back to browse
Visual Learning with LLMs

Visual Learning with LLMs

by aab99·May 10, 2026·1 point·0 comments

AI Analysis

MidEye CandyShip It

Visually impressive, but the agent still needs manual intervention to generate good visuals.

Strengths
  • Generates full interactive spaces including games, not just text.
  • Pre-built demos like ATC communications show clear potential use cases.
  • Clean, inviting UI that makes complex topics feel approachable.
Weaknesses
  • Author admits the agent requires human feedback to polish visuals reliably.
  • Unclear how this differs from existing interactive course builders beyond the AI generation.
Category
Target Audience

Students, self-learners, educators

Similar To

Khanmigo · Duolingo Max · Brilliant.org

Post Description

Hi HN, sharing a learning tool I built which pairs an LLM generated learning space with interactive visuals and small games. The part which I am still working on is getting an agent to reliably generate good visuals without intervention.

Here are three demos you can explore without logging in:

Math refresher for robotics https://forray.io/space/61e71354-049d-405a-9ef9-b5b64826630a

History of the Roman Empire https://forray.io/space/e2b2971c-fb19-4239-b8fe-00b16b5b13c6

ATC communications https://forray.io/space/2d2d7386-6080-4ef3-85a1-3d5b95a3f92c

These were generated by the same agent that runs in the app, but I intervened with feedback on the visuals via the chat feature, so they are a bit more polished. If you want to signup and try it out the code is HN-BETA, which gets you 2 free spaces

Happy to hear any feedback or thoughts.

Similar Projects