aime
✦Research Institute

Research & Insights

Working notes, frameworks and architectural research from the aime team on the future of educational intelligence.

Access

aime is currently operating in stealth. Full working papers are shared with ministries, institutional partners and qualified investors under NDA. Abstracts are public; the underlying papers, methods and benchmarks are released on request.

Agentic Education Systems

Replayable Planning: an event-sourced architecture for the aime Planner Agent

Notes on aime's shift from plan-replacement to decision-event planning. The Planner Agent emits typed educational decisions; the current plan is a projection over a periodic snapshot and recent events. Extends the deterministic, replay-based posture already proven inside the Loom™ Workflow Engine to give planning incremental updates, full audit, explainable evolution and a foundation for human-in-the-loop governance. Event taxonomy, projection engine, snapshot strategy and conflict resolution remain aime IP, shared with partners under NDA.

AI-Native Education

Smarter on Smaller Models: architectural optimization for educational AI

Notes on aime's platform-wide shift to compact, education-tuned models — narrower agent responsibilities, deterministic software in place of language work, schemas shaped for small models, and validation as a first-class layer. Targeting a 60–80% reduction in prompt complexity and LLM cost with classroom quality held constant. Implementation specifics remain aime IP, shared with partners under NDA.

AI-Native Education

Toward a definition of educational intelligence

A framework for understanding what should be taught, what has been taught, what has been understood and what should happen next — at the level of the system.

AI-Native Education

Beyond AI tools: education systems built around intelligence

Why grafting AI applications onto existing educational platforms underdelivers — and what an AI-native architecture requires.

Offline AI

Intelligence at the edge of the network

Deploying reasoning and pedagogy to schools without continuous connectivity — design principles for offline-first educational intelligence.

Agentic Education Systems

Composing educational agents with Kern and Loom

Architectural notes on building reliable, observable educational workflows from a small set of well-defined agents.

Curriculum Intelligence

ThinkBook: a knowledge substrate for curriculum

Representing curricula, learning outcomes and pedagogical dependencies as a first-class knowledge architecture.

National Infrastructure

Sovereignty as a first principle for educational AI

Curriculum, AI, language and offline sovereignty — the operating constraints that shape national-scale deployments.