ARDOR
We built this specifically for middle and high school students who grew up on interactive media, and for the history teachers who are exhausted trying to compete for their attention. Right now, history class feels like a chore. Kids are handed heavy, dry textbooks and told to memorize lists of dates and names just to pass a Friday quiz. The biggest failure is the complete lack of emotional connection—they are learning about their own heritage, but it feels like reading a phone book. Because the traditional education system forces every kid through the exact same text-heavy curriculum, visual and interactive learners simply tune out. They don't "feel" the past; they survive the 45-minute period and forget the facts as soon as the bell rings. We are treating students like passive readers in a subject that should be an epic, living story.
Preliminary score
- Value for the user13.5 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI11.5 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing7.5 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility6.0 / 15
Data requirements, running costs and a clear next step.
- Originality5.0 / 10
A meaningful difference from common solutions.
What the judges wrote
- GPT-6 AstraPreliminary49/100
- Strengths
- Your focus on Azerbaijani students and curriculum-backed history gives the product a clear audience and a relevant source base. Your testing write-up identifies specific failure modes—date mixing in long contexts and retrieval misses on historical terminology—rather than only claiming success. Your proposed historical-character dialogues and timeline exploration offer a sensible engagement layer beyond textbook search.
- To improve
- Your prototype evidence should show one complete, source-cited learning session and distinguish working features from planned ones; replace the generic repository README with implementation details and consistent setup instructions. Your testing should include a counted question set, expected answers, before-and-after error rates, and a comparison with textbook search or a learning-retention measure. Your feasibility plan should quantify inference costs under the 5₼ unlimited-dialogue tier, clarify textbook permissions, and define a teacher pilot. For the enterprise track, your team should identify the school buyer and a teacher workflow with a measurable adoption benefit.
- Claude Opus 5.5Preliminary38/100
- Strengths
- Your problem framing is vivid and grounded: middle and high school students cramming dates for a Friday quiz, and history teachers struggling to hold their attention. Grounding answers in Azerbaijani national curriculum textbooks through retrieval, instead of open web search, is a sensible way to keep the content verifiable. Your testing notes name two real failure modes: chronology mixing on long contexts and retrieval misses on Old Azerbaijani/Arabic-script terms. That is more honest than a generic 'it works well'. Dialogues with historical figures, plus the competitor comparison including Assassin's Creed Discovery Tour, show you have thought about what makes this engaging.
- To improve
- Your repository README is the unedited Vite template, and it contradicts your stated Next.js stack. Replace it with a description of the actual RAG pipeline, the data ingestion and the screens, so judges can verify the core scenario works. Make your testing measurable. For example, run 30–50 curriculum questions with known answers, report how many the system got right before and after your prompt and chunking fixes, and include a few concrete wrong outputs. Define one outcome metric, such as quiz score gain or minutes of engagement versus a textbook chapter, and try it with even a handful of students. Your market numbers need fixing (SOM of 20M exceeds SAM of 1M), and you should add per-user LLM and Whisper running costs against the 5₼ premium plan.
The team
Nihad JafarovLead
product
Elvin İsayev
developer