ASSET AI
About

Autonomous pedagogical agents for English

ASSET pairs an LLM agent pipeline with cognitive learning loops, so a teacher's prompt becomes reviewable course content and a learner gets a tutor that remembers where they struggled.

What the platform does

Teachers describe the outcome they want — “a B1 unit on negotiation for my Tuesday class” — and the agent plans, generates and self-checks a module before it reaches anyone. Nothing is published without a human approving it.

Learners work through those modules and through open conversation with an AI Companion that adapts to their proficiency level, tracks mastery per skill, and schedules what to revisit based on what they actually got wrong.

How the agent improves

Every attempt and conversation becomes a memory the agent can draw on: what helped, what did not, which skills are due for review. Those memories are reference material for the agent — they shape the next reply, never the code it runs on.

Changes to the agent itself arrive as reviewed code, not as self-modification. The platform is designed around that boundary.

Why it exists

The project began as a research study into how agentic AI changes English teachers’ AI competency, professional identity and entrepreneurial practice. It is built by a two-person team — one working on the backend and deployment, one on content and the research design.

The mission is access: lowering the cost of individual attention so a learner anywhere can practise speaking, writing and thinking in English with something that responds to them.

Availability

Teachers can publish a module and share it as a link that anyone with the link can open — no account required for the recipient. Classes, assignments and analytics stay private to the teacher and the learners enrolled in them.