Company profile · 4 min read
Frizzle: Faster math feedback without opaque grading
Frizzle grades handwritten math work with AI and turns it into feedback and assessment analytics for teachers and tutors.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
Frizzle uses AI to grade handwritten math work and turn it into feedback and classroom analytics. It fits teachers and tutors who want to see student reasoning sooner without spending the week manually marking every worksheet.
What it does
Frizzle’s YC profile describes a workflow that reads handwritten math assignments, grades them and produces differentiated assessment. The intended change is from occasional end-of-unit tests to a more continuous picture of student thinking. The current homepage keeps the promise simple: grade math assignments in seconds and provide feedback that understands mathematical reasoning.
The useful product advantage is not merely saving keystrokes. If the system can identify the step where a student’s reasoning diverged, a teacher can adjust instruction before the next unit. That makes the product more valuable in classrooms, tutoring programs and intervention workflows where feedback speed matters. Frizzle’s about page says it builds with classroom teachers and private tutors, which is a better fit signal than a generic “AI for education” label.
The YC page says teachers spend more than 10 hours a week, or 25% of their work, grading and says Frizzle does it in minutes. Those are company-reported framing claims, not an independent time study. The same page says Frizzle was named an official White House AI Education Partner; a buyer should verify the current program record if that designation matters to procurement.
Pricing and founder context
The pricing page advertises a free plan with 50 worksheets per month in its page description, but the full plan table was loading in the retained observation. Paid tiers, student limits, export rules and school purchasing terms should be confirmed before adoption.
The founders met in high school. The YC record identifies Abhay Gupta as CEO, with product roles at Coinbase, Tesla and Meta, and Shyam Sai as CTO, with machine-learning roles at Microsoft, Meta and Google Life Sciences. That background supports the product’s software and education ambition, but it does not establish grading accuracy across handwriting styles, curricula or unusual solutions.
What could make it the wrong choice
Teachers still need a review path for ambiguous handwriting, partial credit and novel reasoning. A useful deployment should make disagreements easy to inspect and correct. The school buyer also needs clarity on student-data handling, roster integration and whether the product supports the specific grade levels and math formats in use.
Editorial take
I would shortlist Frizzle for a classroom or tutoring program with a repeatable math-assignment stream and a teacher willing to review edge cases. The product earns its place if faster feedback changes instruction, not if it only turns grading into an opaque score. Start with one assignment type and inspect the explanations before expanding.
Quick facts
| Field | Sourced detail |
|---|---|
| Product | Handwritten math grading, feedback and assessment analytics |
| Buyers | Teachers, tutors and education teams |
| Public pricing | Free plan description says 50 worksheets/month; full tiers not observed |
| Public claims | 10-plus hours/week and 25% grading framing; company-reported |
| Main gate | Accuracy, explainable partial credit, student-data handling and review workflow |
Sources checked
| Source | Checked |
|---|---|
| YC profile | 2026-09-19 |
| Frizzle homepage | 2026-09-19 |
| Frizzle about page | 2026-09-19 |
| Frizzle pricing | 2026-09-19 |
Cohort context
Frizzle is listed in Summer 2025. In our 2026-09-18 directory snapshot, 2 of 166 listed companies in that cohort have YC’s primary industry label Education (1.2%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.
Public website snapshot
Observed 2026-09-19T16:18:04.735Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.
| Signal | Homepage observation |
|---|---|
| Product description metadata | Observed |
| Canonical link | Observed |
| H1 or H2 heading | Observed |
| Typed structured data | Not observed in this response |
| Docs/developer link | Not observed in this response |
| Pricing link | Observed |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |
Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗