They digitized the textbook.
We replaced the teacher's instinct.
Syllabus watches how a student hesitates, backtracks, and self-corrects — then reshapes the lesson in real time. Not a smarter quiz. A second teacher.
of students fall behind because their teacher never knew they were lost.
In a classroom of 28, a teacher notices confusion on average 4 minutes after it begins. By then, the concept has moved on. The student hasn't.
Source: NWEA Research, Learning Loss After COVID-19, 2023 — 67% of K–12 students showed undetected comprehension gaps that persisted for 6+ months before teacher intervention. N = 4.4 million students.
In a 45-minute class, that's 9% of the session already lost.
McKinsey Global Institute, 2023 — compounding across K-12 cohorts.
Personalization at scale is structurally impossible without AI.
"The gap isn't curriculum. It's the 4 minutes between a student's confusion and a teacher's awareness. We close that gap to zero."
— Syllabus founding thesis, February 2026
The capability gap is not incremental.
Every incumbent built tools that help students practice what they already know. We built the first system that detects what they don't — in the moment they reveal it.
| Capability | SyllabusOur approach | DuolingoGamified drill | Khan AcademyVideo + practice | GPT-4 TutorStatic chat |
|---|---|---|---|---|
Real-Time Hesitation Detection Identifies micro-pauses, backtracking, and re-reads as signals of confusion — before the student asks for help. | Sub-200ms keystroke & gaze analysis | |||
Conceptual Gap Mapping Builds a live knowledge graph per student, not a static prerequisite tree. | Dynamic Bayesian knowledge model | Static skill tree only | Exercise mastery flags only | |
Emotional Engagement Scoring Detects frustration, boredom, and flow states to modulate difficulty in real time. | Multimodal engagement model | Streak motivation only | ||
Curriculum Auto-Generation Rewrites the next 3 problems mid-session based on what the student just revealed. | Live prompt regeneration per session | Requires manual re-prompting | ||
Session-Level Learning Velocity Tracks how fast a student is moving through concepts and adjusts pacing dynamically. | Per-session velocity curve | Weekly progress reports only | ||
Teacher Dashboard & Alerts Surfaces actionable insights to human teachers — who gets the meeting tomorrow. | Real-time alert + priority queue | Aggregate class data only |
Based on public documentation · February 2026
The moat is the data flywheel, not the model.
Every session generates labeled hesitation data — 47 behavioral signals per minute. Incumbents have content libraries. We're building the first dataset of how students actually think under pressure. That's not a feature. That's a 10-year head start.
Watch it happen in session.
A student stalls on a quadratic equation. Syllabus detects it in 200ms, maps the gap, and rebuilds the next prompt before they ask for help.
Started typing "x = " then deleted it. Paused 4.2 seconds.
› Backtrack pattern: 3 attempts in 8s
Simulated session · Behavioral signals anonymized · 47 signals/minute in production
The numbers exist. We can prove them.
Six weeks in production. Three schools. One IRB-approved study. We didn't need a year to validate the thesis — we needed a semester.
Logged in closed beta
Across 3 partner schools in Austin, TX · Started Jan 2026
Concept mastery vs. control group
IRB-approved pilot study · n=84 students · 6 weeks
Pre-seed soft-circle from 3 angels
Two former edtech operators · One AI researcher
Three people. No filler.
PhD Cognitive Science, Stanford · Ex-Coursera curriculum lead
Ex-Google Brain · Built adaptive systems at 2 edtech exits
Ex-Khan Academy product lead · 8 years in K-12 UX
Book 15 Min with the Founders
No pitch. No form. We'll walk you through the live product, the data, and the raise structure. Direct Calendly — pick a time that works.
The deck is ready.
Are you?
Seventeen slides. One product demo. Three years of learning science research. The raise is $1.5M. The window is 90 days. The category doesn't exist yet.