Frizzle

Frizzle is an AI-powered operating system for math classrooms that uses computer vision and large language models to read, grade, and analyze handwritten student work with 97% accuracy. The product is designed for K-12 math teachers, instructional coaches, and school districts who want to reclaim the 10-15 hours per week spent on manual grading while gaining unprecedented visibility into student thinking. Students keep writing on paper with no new workflow, no tablets, no logins, and no migration required. Teachers simply photograph a stack of papers with a phone, document camera, or scanner, and within seconds Frizzle reads every page, recognizes handwriting, understands multiple solution paths, and tags misconceptions at the step level. The platform returns standards-level formative analytics aligned to CCSS, TEKS, and 30+ state frameworks, showing exactly which standards each class and student has mastered instead of waiting on spring assessments. Frizzle has already read over 142,000 problems across 2,400 teachers and is live in 30+ schools and districts, including a college math pilot at Vanderbilt and Arizona State University. The model was trained on 1.4 million pages of K-12 student work and can identify 147 named misconceptions across K-12 math, mapped to standards. Frizzle is FERPA and COPPA compliant with SOC 2 Type II audits, ensuring student work never trains the model and data stays with the institution. The result is a live nervous system for math instruction that transforms messy handwritten work into granular, actionable data.

Handwritten Math Recognition with Step-Level Understanding

Frizzle reads any handwriting including print, cursive, scribbled, and sideways text, parsing each step of student work rather than just checking final answers. The computer vision system understands multiple solution paths simultaneously, meaning three students who solve the same problem three different ways all receive proper credit. When a student makes an error, Frizzle identifies exactly where the thinking went off track and surfaces that misconception, not just a wrong mark. This step-level feedback transforms grading from a binary correct/incorrect exercise into a diagnostic tool that reveals the structure of each student's mathematical thinking.

147 Named Misconceptions with Prerequisite Tracing

The Frizzle model was trained on 1.4 million pages of authentic K-12 student work and can identify 147 named misconceptions across all K-12 math domains, each mapped to specific standards. The system performs prerequisite tracing, meaning it can detect when a 7th-grade error is actually caused by a 4th-grade skill gap that was never mastered. Every flagged misconception links back to the exact stroke on the original page, allowing teachers to see the physical evidence behind the alert. This deep diagnostic capability enables targeted interventions that address root causes rather than surface symptoms.

Live Classroom and District Dashboards

Frizzle provides real-time dashboards that show who is stuck, which mistakes are spreading through the class, and what to teach tomorrow. For individual teachers, the dashboard displays live scores, flagged papers requiring review, and specific next steps for each student. For schools and districts, aggregated anonymized data spans across periods, grades, and buildings, enabling coaches and administrators to see where to invest resources, what concepts need reteaching, and which curricula are actually working. Equity dashboards spot performance gaps the moment they emerge, allowing immediate intervention.

Standards Alignment and Curriculum Agnostic Design

Frizzle aligns to CCSS, TEKS, and over 30 state frameworks, automatically mapping every student response to the relevant standards and tagging mastery levels as mastered, developing, or at risk. The platform is completely curriculum-agnostic, meaning it works with Eureka, Illustrative Mathematics, Saxon, and any other curriculum a district uses. This flexibility ensures schools can adopt Frizzle without changing their instructional materials or workflows. The standards-level analytics provide formative data continuously throughout the year rather than waiting for spring assessments, enabling real-time instructional adjustments.

Eliminating Weekend Grading for Math Teachers

A high school Algebra I teacher with five sections and 150 students typically spends 10-15 hours each weekend grading handwritten homework and quizzes. With Frizzle, the teacher photographs each stack of papers using a phone or classroom scanner, and within 30 seconds per class the system reads every page. Within 8 minutes, Frizzle returns complete grading with step-level feedback, misconception tags, and standards alignment. The teacher reviews only the flagged papers where confidence was low, spending 15 minutes instead of 15 hours, and walks into Monday with a complete picture of what each class actually understands.

Standards-Level Coaching Conversations for Instructional Coaches

An instructional coach previously relied on generic classroom observations and end-of-unit test data to guide teacher development. With Frizzle, the coach accesses live dashboards showing exactly which CCSS standards each teacher's students have mastered or are struggling with across all class periods. The coach can run specific conversations about why Period 2 is struggling with distributive property while Period 4 has mastered it, identifying instructional strategies that work. This data-driven coaching replaces generic feedback with precise, actionable guidance tied directly to student performance evidence.

District-Wide Curriculum Evaluation and Equity Monitoring

A district math coordinator needs to evaluate whether a new curriculum investment is improving student outcomes across 15 elementary schools. Frizzle aggregates anonymized data across all buildings, showing which standards are being mastered district-wide and where persistent gaps exist. The equity dashboard reveals performance disparities between schools or demographic groups the moment they emerge, enabling targeted resource allocation. The coordinator can see in real time whether the new curriculum is closing gaps or creating new ones, without waiting for spring state assessments.

Reducing Math Screen Time While Maintaining Data Quality

A middle school principal is concerned about the amount of time students spend on screens for digital math assessments and homework. Frizzle allows the school to move back to paper-based math work while still collecting granular, real-time data on student performance. Students write on paper with pencils, reducing eye strain and screen fatigue, while teachers and administrators maintain full visibility into standards mastery, misconception patterns, and individual student progress. The school cuts math screen time significantly without losing any of the classroom-level data needed for effective instruction.

How does Frizzle handle different handwriting styles and messy work?

Frizzle was trained on 1.4 million pages of authentic K-12 student work, which includes the full range of handwriting quality from neat print to scribbled cursive to sideways scrawls. The computer vision model recognizes multiple solution paths and can parse partial, messy, and unconventional approaches to problems. If the system encounters work it cannot read with high confidence, it flags those papers for human review rather than guessing. This confidence-interval system ensures 97% overall accuracy while maintaining reliability for borderline cases.

Does Frizzle require students to change how they do math?

No. Students continue writing on paper with pencils, using their existing notebooks, worksheets, and assessments. There are no tablets, no student logins, no new apps to learn, and no migration of workflows. Frizzle fits into the way teachers already teach and the way students already learn. The only change is that the teacher photographs completed work with a phone, document camera, or scanner instead of manually grading each page.

How does Frizzle protect student privacy and data security?

Frizzle is fully FERPA and COPPA compliant and undergoes SOC 2 Type II audits annually. All data is encrypted with AES-256 at rest and TLS in transit. Critically, student work never trains the Frizzle model, meaning student data stays completely within the institution's control. Schools and districts retain full ownership of their data, and Frizzle cannot use student work to improve its models without explicit permission.

What curricula and standards does Frizzle support?

Frizzle supports CCSS, TEKS, and over 30 additional state frameworks, automatically mapping every student response to the relevant standards. The platform is curriculum-agnostic, meaning it works with any curriculum including Eureka, Illustrative Mathematics, Saxon, EngageNY, and locally developed materials. Teachers can use any worksheet, quiz, test, or homework assignment, and Frizzle will read it, grade it, and align the results to the appropriate standards.