Brewing cosmic thoughts...
lyra transforms AI from an answer machine into a learning partner.
Guided AI for schools, built to develop reasoning, not replace it.
Already have an account?
The button above works for login too.

A silent crisis is unfolding in classrooms worldwide.
of high school students use generative AI for schoolwork
Generative AI adoption among students is accelerating rapidly. Tools like ChatGPT are now embedded in everyday schoolwork, from brainstorming essays to researching assignments.
of teachers worry AI weakens critical thinking
Educators worry students may outsource thinking itself. When AI produces answers instantly, students risk skipping the struggle that builds reasoning, curiosity, and problem-solving.
schools restrict student AI use
Schools respond with bans
Around 2 in 5 schools in US do not allow generative AI use. But history shows prohibition rarely works. Technology is not the problem — how we teach students to use it is.
Sources: College Board AI Research Brief (US-2025) • US Education Week teacher surveys
Not another AI model. A regulated interface that connects to models like GPT, Gemini, and DeepSeek while guiding students through the learning process.
Pedagogical Guardrails
AI Models
AI models operate safely inside lyra's pedagogical guardrails.
lyra asks guiding questions instead of giving answers.
Uses school materials through RAG to stay contextually relevant.
Educators control prompts, boundaries, and difficulty levels.
Students access GPT, Gemini, DeepSeek through one governed platform.
A student types their homework question into lyra.
Instead of answering, lyra prompts the student to think deeper.
The student articulates their thought process step by step.
Through guided dialogue, the student reaches the solution independently.
Customize system prompts
Define exactly how lyra responds to student queries
Upload course materials
RAG-powered context from your own curriculum
Create few-shot examples
Train lyra with ideal conversation patterns
Adjust difficulty per student
Personalized scaffolding for every learner
M57 monitors out-of-policy traffic, triggers an alarm-grade intercept, then routes the student back into lyra's guided interface.
Heuristics and endpoint signatures catch suspicious traffic patterns immediately.
Policy-aware controls stop responses before prohibited content reaches the learner.
Students are reintroduced into a pedagogical path inside lyra with guided prompts.
lyra uses enterprise-grade Android device policies to prevent unauthorized removal. Only authorized school administrators can uninstall the application.
Kernel live
Live simulation
M57 flags a request to an external AI endpoint in under 50ms.
The browser flow is blocked before response content is rendered.
The learner is guided back to lyra for scaffolded support.
Ready: trigger the simulation to observe the full enforcement flow.
M57 kernel stream
[ M57 ] Initializing pedagogical enforcement kernel
[ M57 ] Scanning environment for AI systems▋
AI regulated competitions where creativity is tested, not copy-paste skills.
Instructor-controlled AI assistance that adapts to each course's pedagogy.
System-level enforcement ensuring academic integrity at scale.
Lines of Code
Total Commits
Last updated -. These repository metrics update automatically using GitHub Actions.