The encyclopedia

Chrome extension that gives K-12 teachers AI tooling inside Google Docs + Slides — rubrics, differentiated readings, feedback drafting, plagiarism / AI-writing detection.

Mika is Carnegie Learning's AI math tutor for grades 6-12, built on three decades of cognitive-science research at CMU. Adaptive practice + just-in-time hints.

Chegg's GPT-4-powered study assistant — homework help, conceptual explanations, step-by-step solutions. Layered over Chegg's existing textbook-solutions catalog.

Internal review email states Claude via Amazon Bedrock with OpenAthens-anonymized authentication and notes undisclosed retention policy on user queries. Internal library reply says vendor questionnaires had not yet all been sent at that point.

Duolingo's GPT-4-powered language-learning tier — Explain My Answer + Roleplay conversation features. Adds AI-driven personalized feedback to the gamified base product.

Tracking file states EBSCO Natural Language Search is an advanced AI search feature that powers EBSCO Discovery Service, no PII is involved, and user data is not used to update the LLMs.

Khan Academy's GPT-4-powered AI tutor for K-12 and college students. Socratic tutoring style; teacher-facing tooling for lesson planning and grading. Free for educators.

AI productivity suite for K-12 teachers — lesson planning, rubric generation, IEP drafting, student-feedback writing. Used by 4M+ teachers, with district-wide enterprise tiers.

Adaptive-learning system for K-12 + higher-ed math and chemistry. Identifies what each student knows + doesn't, then teaches only the gaps. AI-driven assessment loop.

SaaS solution used for voice-enabled, scenario-based interactions for medical training. Review specifically examined AI functionality enabling real-time conversations with virtual patients or team members generating context-aware responses.

Internal review email states OpenAI GPT-4o-mini disclosed, closed RAG over public metadata only (not full text), no training, and no session memory. Internal reply later states requested vendor questionnaire was sent.

Vendor ethics statement says models are trained exclusively on pseudonymized student data, historical grading records, pedagogical goals, conceptual labels, patterns, and rubrics, with no PII collected or required; Sense also states each grading model is benchmarked against human graders, audited for bias, and designed with human-in-the-loop oversight. Internal assessment summary later states Sense Instant Feedback uses a Claude-based LLM and SNHU data is not used to update models.