Pioneering Cognitive EdTech R&D since 2023Educational Pilot Cohorts Open
Moving beyond rote memorization. An adaptive acquisition engine grounded in Diane Larsen-Freeman's Form-Meaning-Use (FMU) framework and Conceptual Metaphor Theory.
Institutional pilots prioritized. No spam, ever.
Technical Architecture
Each learner turn is routed through specialized agents built on Anthropic's Claude family. Depth where diagnosis matters, speed where momentum matters, and a safety gate in between.
Claude 3.5 Sonnet
Deep multi-turn discourse analysis. Diagnoses why a learner's conceptual metaphor or register breaks pragmatically, rather than generating simplistic grammar corrections.
Claude 3.5 Haiku
Real-time, sub-second formative feedback and adaptive hints mapped dynamically to CEFR progression stages (A1 to C2).
Evaluation Pipeline
Context-bounded evaluation pipelines preventing hallucinated linguistic rules before any feedback reaches the learner.
The FMU Dimension Engine
Larsen-Freeman's framework treats grammar not as rules to memorize but as a dynamic system of form, meaning, and use. Select a dimension to see how the engine reasons about a single learner sentence.
Learner input
Dear Professor, I am agree with your idea, because it gave me a light. See you!
Meaning diagnosis
The learner correctly activates UNDERSTANDING IS SEEING, but realizes it with an L1 collocation. In English the metaphor surfaces as "shed light on", not "give a light" (which evokes lighting a cigarette).
Scaffolded revision
gave me a lightit shed light on the problem for me
Research & Early Cohorts
Our pilot cohorts pair the engine with practicing educators to validate outcomes against established applied linguistics benchmarks.
Cohort-level analytics and CEFR-aligned progression for instructor-led programs.
Research-grade discourse diagnostics for EAP, ESP, and applied linguistics departments.
Rigorous, self-directed acquisition for serious learners pursuing C1 – C2 mastery.
4.8x
Retention on nuanced syntactic forms
vs. baseline drill-based instruction
<800ms
Adaptive scaffolding response time
p95 formative feedback latency
Preliminary figures from internal pilot cohorts. Findings are not yet peer-reviewed and are shared for transparency with prospective research partners.