Magic School uses AI to help kids learn, not cheat | E2196

Key Concepts

• Magic School: An AI platform designed to assist educators and students in K-12 education, focusing on responsible AI integration.
• Teacher Co-pilot: Magic School's core function for educators, aiding in daily tasks like lesson planning, rubric creation, and material differentiation.
• AI Literacy: Educating students on the responsible and effective use of AI tools.
• Formative Feedback: Providing students with immediate feedback on their work to allow for revision and improvement before final grading.
• District-wide AI Ecosystem: Enabling school districts to create customized, secure Generative AI tools integrated with their specific knowledge and priorities.
• Responsible AI Use: Emphasizing ethical and thoughtful implementation of AI in educational settings, particularly for students.
• Cognitive Offload: The risk of students relying too heavily on AI, potentially hindering their own cognitive development.
• Vesting Schedule: The timeline over which an employee earns ownership of their stock options or shares.
• On the Fly Energy: A startup developing flywheel-based energy storage solutions for critical infrastructure.
• Rotor Seed: The modular flywheel unit developed by On the Fly Energy.
• Flywheel Energy Storage: A method of storing energy by spinning a rotor at high speeds.
• Duck Curve: A graph illustrating the net load on an electricity grid over a day, showing a dip in demand during midday solar production and a sharp increase in the evening.
• Energy Independence: The goal of a nation to be self-reliant in its energy production and consumption.

Magic School: AI in the Classroom

Introduction to Magic School and its Mission

Magic School is an AI company aiming to bring AI into classrooms responsibly. Its core mission is to support educators and students by providing tools that enhance learning and teaching. The company has seen significant growth, raising $45 million in a Series B funding round. Alex, the host, notes that AI is increasingly prevalent in various sectors, including education, and Magic School is positioned to lead in this space.

Magic School's Offerings

• For Educators (Teacher Co-pilot):
  • Daily Task Assistance: Helps teachers with tasks such as building rubrics for projects, differentiating materials for students with varying needs (advanced learners, those needing extra support, English language learners, students with special education plans).
  • AI Experience Crafting: Teachers can create AI-powered learning tools for their students. These are teacher-led and include features like:
    • "ChatGPT for Kids" with Guardrails: A guided AI experience that helps students with academics without directly providing answers.
    • Writing Feedback Tool: Students can submit essays and receive AI-aligned formative feedback on areas of strength and growth, based on a teacher-defined rubric, allowing for revisions before final grading. This is presented as a significant improvement over traditional grading turnaround times.
• For Students:
  • AI Literacy: Tools designed to help students understand and use AI effectively and responsibly.
  • Personalized Learning: Access to AI-driven tools that provide tailored support and feedback.
  • Early Feedback on Papers: The writing feedback tool offers immediate insights for improvement.
• For School Districts:
  • Closed Sandbox Generative AI: Districts can create their own secure, customized Generative AI environment.
  • Infusion of District Knowledge: Using Retrieval-Augmented Generation (RAG) technology, districts can integrate their specific knowledge, priorities, and goals into the AI platform, ensuring alignment with district standards and policies. This creates a unique ecosystem of Gen AI tools for their educators.

Addressing Concerns about AI in Education

• Teacher Control: Adil Khan, CEO of Magic School, emphasizes that teachers remain central to the classroom. The slogan "Teachers are magic" reflects the company's ethos that AI should augment, not replace, teachers.
• Rubric Generation Example: Teachers provide specific input, such as state standards, assignment text, and additional instructions, to generate rubrics. The platform then allows for differentiation of these rubrics for various student needs (e.g., Spanish translation, versions for students behind or ahead of grade level). This addresses the challenge of meeting diverse learning needs in large, multilingual classrooms.
• Teacher Burnout and Shortage: Magic School aims to alleviate teacher burnout by providing tools that handle complex differentiation tasks, which are often overwhelming for educators. The significant teacher shortage in the US (412,000 unfilled positions or filled by uncertified teachers) highlights the need for such support.
• Ease of Use and Adoption: The platform is designed to be intuitive and familiar to educators, with many users commenting that it feels like it was built by someone who understands their job deeply. This ease of use has contributed to its rapid adoption.
• Support for Novice and Under-certified Teachers: Magic School assists teachers who may not have formal pedagogical training, such as long-term substitutes or career and technical education instructors, by helping them build lesson plans and content.

Distribution and Growth

• Organic Growth: Magic School experienced significant growth through word-of-mouth, with teachers sharing the tool with colleagues. They did not spend on marketing until they reached nearly a million educators.
• User Numbers: Over 6.5 million users have signed up for Magic School, with approximately half of those users in the United States, a remarkable penetration given the number of educators in the country.
• Market Share: A chart from Burbio shows Magic School leading in the percentage of school districts that have paid for an AI platform in the last 12 months, surpassing companies like School AI and OpenAI.
• Sales Model:
  • Free for Teachers: A generous free version is available to individual educators.
  • Paid for Districts: The primary customer is the school district, which pays for features like data security, single sign-on, and the ability to customize AI tools with their own data and policies.
  • Solutions Architects: Magic School employs "solutions architects" (akin to forward-deployed engineers) who work directly with districts to deeply integrate and customize the platform to their specific needs and regional priorities (e.g., customizing for Florida state standards).

Underlying Technology and Future Development

• AI Models: Magic School utilizes models from providers like Anthropic and OpenAI.
• Model Capabilities: Current models are considered extraordinary for existing tasks. Future improvements are sought in areas like student guardrails, text complexity evaluation, and audio models (e.g., better handling of speech-to-text for younger students).
• Student AI Usage: The College Board reports an increase in high school students using Generative AI for schoolwork (84% in May of the current year). Magic School advocates for thoughtful, scaffolded AI integration, with limited use for younger students and increasing, but still guardrailed, use for older students.
• Responsible AI Rubric (80/20 Rule): This rule is for teachers, encouraging them to use AI for the initial 80% of a task and apply their final 20% of human expertise and refinement. It is explicitly not for students, who are expected to do the full cognitive lifting.
• Future Product Vision:
  • Deep District Integration: Continued focus on deploying solutions architects to deeply integrate Magic School with district challenges.
  • Unified Knowledge Graph: Creating a comprehensive understanding of each student's learning profile (grade level, special designations, learning needs) to personalize AI support within their "zone of proximal development."
  • Personalized Education Assistant: AI tools that act as assistants for both students and teachers, understanding individual needs and contextualizing learning.
  • Enhanced Teacher Support: AI that auto-differentiates materials based on class-wide student knowledge and needs, making teachers more effective.
• Timeline: Some of these advanced features are expected to be available as early as January, with significant changes anticipated within the next year for partner districts.
• Partnerships: Magic School has strong relationships with foundation model providers, who offer support through dedicated education and beneficial deployment teams, even if not providing direct subsidies.

International Expansion

• Global Usage: Magic School sees significant international usage without direct marketing or sales, with inbound interest from schools worldwide.
• UAE Example: Approximately 80% of educators in the UAE are reportedly using Magic School, demonstrating massive adoption even in smaller nations.
• Low CAC: The organic growth and high adoption rates contribute to a low Customer Acquisition Cost (CAC), a key factor in their successful Series B funding.

Founder Question: Employee Share Vesting

The Question

A founder is asking about the appropriate vesting schedule for employees, specifically noting an offer of 0.023% of a company over a six-year period in the hardware sector. The question is whether this is standard for employees or more suited for founders.

Discussion and Perspectives

• Market Standard: The typical vesting schedule for employees is four years, with a one-year cliff (meaning no shares vest until the one-year anniversary).
• Six-Year Vesting: Six years is considered abnormal and is a long time for an employee to wait for full vesting.
• Exceptions and Variations:
  • Public Companies: Companies like Stripe may offer annual vesting over a 12-month period, but this is different as they are public entities with more established share values.
  • Founder Perspective: Founders might accept longer vesting periods, especially if they believe the company's valuation will significantly increase, making the later-vesting shares much more valuable.
  • Negotiation: The decision to accept a longer vesting schedule should depend on the employee's belief in the company and the overall compensation package (cash, stock, benefits).
• Innovation vs. Standardization:
  • Founder Tendency: Founders often want to innovate on legal documents like vesting schedules.
  • Advice: The advice from experienced figures like Ruof (Sequoia) is to keep things standard. This reduces friction in negotiations and allows founders to focus on product innovation rather than legal complexities.
  • Bezos Example: Jeff Bezos at Amazon used a back-loaded vesting schedule (5% in year one, 15% in year two, etc.), which is a strategy for retaining talent and ensuring employees are valuable to the company before significant equity vests. However, this is a strategy for exceptional founders.
• High Performers: A strategy for retaining high performers is to offer parallel grants, essentially layering on additional vesting schedules, which makes it more costly for them to leave.
• Severance: The discussion touches on severance, with the consensus being that in startups, severance is not standard unless there's a layoff or riff.

On the Fly Energy: Flywheel Energy Storage

The Problem

The electricity grid is fragile and susceptible to disruptions. Critical infrastructure, especially AI data centers with their rapidly fluctuating power demands (GPUs swinging from 0 to 100% instantly), destabilize the grid. Existing solutions like batteries (expensive, degrade, fire risk) and diesel generators (slow, maintenance-heavy) are insufficient. An example cited is a lightning arrestor failure in Virginia that took 60 data centers offline.

The Solution: Rotor Seed

On the Fly Energy offers the "Rotor Seed," a modular 10 kilowatt-hour (kWh) flywheel energy storage system that installs like a backup battery.

• Mechanism: It stores energy in kinetic form by spinning a carbon fiber composite disc at high speeds. This stored inertia can then be released to provide power.
• Scalability: Rotor Seeds can be stacked into "Rotor Pods" and scaled into "Rotor Farms" to deliver megawatt-hours of distributed protection.
• Key Advantages:
  • Instant Backup: Provides immediate power delivery.
  • Fire Safe: No fire risk associated with chemical batteries.
  • No Degradation: Unlike batteries, flywheels do not degrade over time.
  • Long Lifespan: Over two decades of life.
  • Built in America: Focus on energy independence.
  • No Lag: Instantaneous response compared to diesel generators.
  • Scalable: Can be scaled beyond capacitors for long-duration protection.
  • Better Spin on Backup Power: A more efficient and reliable alternative.

Business Model and Market

• Target Market: Over $100 billion energy market, including AI data centers, manufacturing hubs, and critical infrastructure where uptime is paramount.
• Revenue Streams:
  • Monthly Subscription: Rotor Seed units are offered as a monthly lease.
  • On-Demand Shifting: Revenue from energy delivered and demand shifting.
  • Utility Support & AI Optimization: Future offerings for grid regulation and AI-driven energy management.
• Path to Market:
  • Proof of Concept: Completed to validate core physics and design.
  • MVP Trials: Early 2026 at beta sites.
  • Small Commercial Launch: 2027, targeting telecom and small data centers.
  • Expansion: 2028, targeting large AI data centers and full grid-scale storage.

Technology and Physics

• Kinetic vs. Chemical Storage: Stores energy as rotational inertia (kinetic) rather than chemical bonds.
• Efficiency: Approximately 95% energy return from power put in to power retrieved.
• Materials: Uses carbon fiber composite for the rotor, allowing for higher rotational speeds and greater energy storage compared to heavier materials. Energy storage is proportional to the square of the rotational speed.
• Rotor Seed Size: The 10 kWh module is approximately 18 inches in diameter and 18 inches long.
• Vibration and Bearings: Operates on magnetic bearings, eliminating vibration.
• Cost: Prototype cost is under $10,000, with a target for the business model to achieve ROI in 1-2 years. This is competitive with 10 kWh battery costs ($10,000-$15,000).
• Scalability: Units can be stacked into ISO shipping containers or other form factors. Larger units can be made, but material stress (hoop stress) becomes a limiting factor at extreme speeds and diameters.

Use Cases and Applications

• AI Data Centers: Buffering power to level out spikes and drops in demand from AI workloads, creating a stable draw for the grid.
• Virtual Power Plants (VPPs): The network of Rotor Seeds can function as a VPP to help regulate the grid, not compete with utilities.
• Energy Arbitrage: Charging during off-peak hours when energy is cheap and discharging during peak demand.
• Grid Conditioning: Stabilizing power flow between the grid and high-demand facilities.
• Air Conditioning: In hot climates, units can charge at night and power AC during peak daytime demand, leveraging price differences.
• EV Fast Charging: Acting as a buffer between the grid and EV chargers, allowing for high-power charging even with lower grid connection capacities.
• Historical Precedent: Flywheel technology has existed for decades, used in applications like buses and roller coasters.

Energy Policy and Future Vision

• Short-Term: Natural gas and other immediate solutions are needed, alongside renewables that can be sorted and stored.
• Long-Term: America needs energy independence through self-reliance, generation, and storage.
• Renewable Integration: Combining renewables (wind, solar) with storage to mitigate production fluctuations and unpredictability.
• Energy Mix: A balanced approach including renewables, nuclear power for baseload, and storage solutions like On the Fly Energy's flywheels.
• Funding: Chris Kennedy is currently self-funding the company.

Pitch Competition and Gamma

The pitch was part of a pitch competition hosted by Gamma, a platform for creating AI-powered presentations. Chris Kennedy is a strong contender to win the competition.